blob: 87de59517e15bfff029a0a2f250d9a6b1323cbe1 [file] [log] [blame]
Willy Tarreau6a06a402007-07-15 20:15:28 +02001 ----------------------
2 HAProxy
3 Configuration Manual
4 ----------------------
Willy Tarreau21475e32010-05-23 08:46:08 +02005 version 1.5
Willy Tarreau6a06a402007-07-15 20:15:28 +02006 willy tarreau
Willy Tarreaueab1dc62013-06-17 15:10:25 +02007 2013/06/17
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
727.3.1. Fetching samples from internal states
737.3.2. Fetching samples at Layer 4
747.3.3. Fetching samples at Layer 5
757.3.4. Fetching samples from buffer contents (Layer 6)
767.3.5. Fetching HTTP samples (Layer 7)
777.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020078
798. Logging
808.1. Log levels
818.2. Log formats
828.2.1. Default log format
838.2.2. TCP log format
848.2.3. HTTP log format
William Lallemand48940402012-01-30 16:47:22 +0100858.2.4. Custom log format
Willy Tarreau5f51e1a2012-12-03 18:40:10 +0100868.2.5. Error log format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200878.3. Advanced logging options
888.3.1. Disabling logging of external tests
898.3.2. Logging before waiting for the session to terminate
908.3.3. Raising log level upon errors
918.3.4. Disabling logging of successful connections
928.4. Timing events
938.5. Session state at disconnection
948.6. Non-printable characters
958.7. Capturing HTTP cookies
968.8. Capturing HTTP headers
978.9. Examples of logs
98
999. Statistics and monitoring
1009.1. CSV format
1019.2. Unix Socket commands
102
103
1041. Quick reminder about HTTP
105----------------------------
106
107When haproxy is running in HTTP mode, both the request and the response are
108fully analyzed and indexed, thus it becomes possible to build matching criteria
109on almost anything found in the contents.
110
111However, it is important to understand how HTTP requests and responses are
112formed, and how HAProxy decomposes them. It will then become easier to write
113correct rules and to debug existing configurations.
114
115
1161.1. The HTTP transaction model
117-------------------------------
118
119The HTTP protocol is transaction-driven. This means that each request will lead
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100120to one and only one response. Traditionally, a TCP connection is established
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200121from the client to the server, a request is sent by the client on the
122connection, the server responds and the connection is closed. A new request
123will involve a new connection :
124
125 [CON1] [REQ1] ... [RESP1] [CLO1] [CON2] [REQ2] ... [RESP2] [CLO2] ...
126
127In this mode, called the "HTTP close" mode, there are as many connection
128establishments as there are HTTP transactions. Since the connection is closed
129by the server after the response, the client does not need to know the content
130length.
131
132Due to the transactional nature of the protocol, it was possible to improve it
133to avoid closing a connection between two subsequent transactions. In this mode
134however, it is mandatory that the server indicates the content length for each
135response so that the client does not wait indefinitely. For this, a special
136header is used: "Content-length". This mode is called the "keep-alive" mode :
137
138 [CON] [REQ1] ... [RESP1] [REQ2] ... [RESP2] [CLO] ...
139
140Its advantages are a reduced latency between transactions, and less processing
141power required on the server side. It is generally better than the close mode,
142but not always because the clients often limit their concurrent connections to
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200143a smaller value.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200144
145A last improvement in the communications is the pipelining mode. It still uses
146keep-alive, but the client does not wait for the first response to send the
147second request. This is useful for fetching large number of images composing a
148page :
149
150 [CON] [REQ1] [REQ2] ... [RESP1] [RESP2] [CLO] ...
151
152This can obviously have a tremendous benefit on performance because the network
153latency is eliminated between subsequent requests. Many HTTP agents do not
154correctly support pipelining since there is no way to associate a response with
155the corresponding request in HTTP. For this reason, it is mandatory for the
Cyril Bonté78caf842010-03-10 22:41:43 +0100156server to reply in the exact same order as the requests were received.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200157
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200158By default HAProxy operates in a tunnel-like mode with regards to persistent
159connections: for each connection it processes the first request and forwards
160everything else (including additional requests) to selected server. Once
161established, the connection is persisted both on the client and server
162sides. Use "option http-server-close" to preserve client persistent connections
163while handling every incoming request individually, dispatching them one after
164another to servers, in HTTP close mode. Use "option httpclose" to switch both
165sides to HTTP close mode. "option forceclose" and "option
166http-pretend-keepalive" help working around servers misbehaving in HTTP close
167mode.
168
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200169
1701.2. HTTP request
171-----------------
172
173First, let's consider this HTTP request :
174
175 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100176 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200177 1 GET /serv/login.php?lang=en&profile=2 HTTP/1.1
178 2 Host: www.mydomain.com
179 3 User-agent: my small browser
180 4 Accept: image/jpeg, image/gif
181 5 Accept: image/png
182
183
1841.2.1. The Request line
185-----------------------
186
187Line 1 is the "request line". It is always composed of 3 fields :
188
189 - a METHOD : GET
190 - a URI : /serv/login.php?lang=en&profile=2
191 - a version tag : HTTP/1.1
192
193All of them are delimited by what the standard calls LWS (linear white spaces),
194which are commonly spaces, but can also be tabs or line feeds/carriage returns
195followed by spaces/tabs. The method itself cannot contain any colon (':') and
196is limited to alphabetic letters. All those various combinations make it
197desirable that HAProxy performs the splitting itself rather than leaving it to
198the user to write a complex or inaccurate regular expression.
199
200The URI itself can have several forms :
201
202 - A "relative URI" :
203
204 /serv/login.php?lang=en&profile=2
205
206 It is a complete URL without the host part. This is generally what is
207 received by servers, reverse proxies and transparent proxies.
208
209 - An "absolute URI", also called a "URL" :
210
211 http://192.168.0.12:8080/serv/login.php?lang=en&profile=2
212
213 It is composed of a "scheme" (the protocol name followed by '://'), a host
214 name or address, optionally a colon (':') followed by a port number, then
215 a relative URI beginning at the first slash ('/') after the address part.
216 This is generally what proxies receive, but a server supporting HTTP/1.1
217 must accept this form too.
218
219 - a star ('*') : this form is only accepted in association with the OPTIONS
220 method and is not relayable. It is used to inquiry a next hop's
221 capabilities.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100222
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200223 - an address:port combination : 192.168.0.12:80
224 This is used with the CONNECT method, which is used to establish TCP
225 tunnels through HTTP proxies, generally for HTTPS, but sometimes for
226 other protocols too.
227
228In a relative URI, two sub-parts are identified. The part before the question
229mark is called the "path". It is typically the relative path to static objects
230on the server. The part after the question mark is called the "query string".
231It is mostly used with GET requests sent to dynamic scripts and is very
232specific to the language, framework or application in use.
233
234
2351.2.2. The request headers
236--------------------------
237
238The headers start at the second line. They are composed of a name at the
239beginning of the line, immediately followed by a colon (':'). Traditionally,
240an LWS is added after the colon but that's not required. Then come the values.
241Multiple identical headers may be folded into one single line, delimiting the
242values with commas, provided that their order is respected. This is commonly
243encountered in the "Cookie:" field. A header may span over multiple lines if
244the subsequent lines begin with an LWS. In the example in 1.2, lines 4 and 5
245define a total of 3 values for the "Accept:" header.
246
247Contrary to a common mis-conception, header names are not case-sensitive, and
248their values are not either if they refer to other header names (such as the
249"Connection:" header).
250
251The end of the headers is indicated by the first empty line. People often say
252that it's a double line feed, which is not exact, even if a double line feed
253is one valid form of empty line.
254
255Fortunately, HAProxy takes care of all these complex combinations when indexing
256headers, checking values and counting them, so there is no reason to worry
257about the way they could be written, but it is important not to accuse an
258application of being buggy if it does unusual, valid things.
259
260Important note:
261 As suggested by RFC2616, HAProxy normalizes headers by replacing line breaks
262 in the middle of headers by LWS in order to join multi-line headers. This
263 is necessary for proper analysis and helps less capable HTTP parsers to work
264 correctly and not to be fooled by such complex constructs.
265
266
2671.3. HTTP response
268------------------
269
270An HTTP response looks very much like an HTTP request. Both are called HTTP
271messages. Let's consider this HTTP response :
272
273 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100274 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200275 1 HTTP/1.1 200 OK
276 2 Content-length: 350
277 3 Content-Type: text/html
278
Willy Tarreau816b9792009-09-15 21:25:21 +0200279As a special case, HTTP supports so called "Informational responses" as status
280codes 1xx. These messages are special in that they don't convey any part of the
281response, they're just used as sort of a signaling message to ask a client to
Willy Tarreau5843d1a2010-02-01 15:13:32 +0100282continue to post its request for instance. In the case of a status 100 response
283the requested information will be carried by the next non-100 response message
284following the informational one. This implies that multiple responses may be
285sent to a single request, and that this only works when keep-alive is enabled
286(1xx messages are HTTP/1.1 only). HAProxy handles these messages and is able to
287correctly forward and skip them, and only process the next non-100 response. As
288such, these messages are neither logged nor transformed, unless explicitly
289state otherwise. Status 101 messages indicate that the protocol is changing
290over the same connection and that haproxy must switch to tunnel mode, just as
291if a CONNECT had occurred. Then the Upgrade header would contain additional
292information about the type of protocol the connection is switching to.
Willy Tarreau816b9792009-09-15 21:25:21 +0200293
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200294
2951.3.1. The Response line
296------------------------
297
298Line 1 is the "response line". It is always composed of 3 fields :
299
300 - a version tag : HTTP/1.1
301 - a status code : 200
302 - a reason : OK
303
304The status code is always 3-digit. The first digit indicates a general status :
Willy Tarreau816b9792009-09-15 21:25:21 +0200305 - 1xx = informational message to be skipped (eg: 100, 101)
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200306 - 2xx = OK, content is following (eg: 200, 206)
307 - 3xx = OK, no content following (eg: 302, 304)
308 - 4xx = error caused by the client (eg: 401, 403, 404)
309 - 5xx = error caused by the server (eg: 500, 502, 503)
310
311Please refer to RFC2616 for the detailed meaning of all such codes. The
Willy Tarreaud72758d2010-01-12 10:42:19 +0100312"reason" field is just a hint, but is not parsed by clients. Anything can be
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200313found there, but it's a common practice to respect the well-established
314messages. It can be composed of one or multiple words, such as "OK", "Found",
315or "Authentication Required".
316
317Haproxy may emit the following status codes by itself :
318
319 Code When / reason
320 200 access to stats page, and when replying to monitoring requests
321 301 when performing a redirection, depending on the configured code
322 302 when performing a redirection, depending on the configured code
323 303 when performing a redirection, depending on the configured code
Willy Tarreaub67fdc42013-03-29 19:28:11 +0100324 307 when performing a redirection, depending on the configured code
325 308 when performing a redirection, depending on the configured code
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200326 400 for an invalid or too large request
327 401 when an authentication is required to perform the action (when
328 accessing the stats page)
329 403 when a request is forbidden by a "block" ACL or "reqdeny" filter
330 408 when the request timeout strikes before the request is complete
331 500 when haproxy encounters an unrecoverable internal error, such as a
332 memory allocation failure, which should never happen
333 502 when the server returns an empty, invalid or incomplete response, or
334 when an "rspdeny" filter blocks the response.
335 503 when no server was available to handle the request, or in response to
336 monitoring requests which match the "monitor fail" condition
337 504 when the response timeout strikes before the server responds
338
339The error 4xx and 5xx codes above may be customized (see "errorloc" in section
3404.2).
341
342
3431.3.2. The response headers
344---------------------------
345
346Response headers work exactly like request headers, and as such, HAProxy uses
347the same parsing function for both. Please refer to paragraph 1.2.2 for more
348details.
349
350
3512. Configuring HAProxy
352----------------------
353
3542.1. Configuration file format
355------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +0200356
357HAProxy's configuration process involves 3 major sources of parameters :
358
359 - the arguments from the command-line, which always take precedence
360 - the "global" section, which sets process-wide parameters
361 - the proxies sections which can take form of "defaults", "listen",
362 "frontend" and "backend".
363
Willy Tarreau0ba27502007-12-24 16:55:16 +0100364The configuration file syntax consists in lines beginning with a keyword
365referenced in this manual, optionally followed by one or several parameters
366delimited by spaces. If spaces have to be entered in strings, then they must be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100367preceded by a backslash ('\') to be escaped. Backslashes also have to be
Willy Tarreau0ba27502007-12-24 16:55:16 +0100368escaped by doubling them.
369
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200370
3712.2. Time format
372----------------
373
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100374Some parameters involve values representing time, such as timeouts. These
Willy Tarreau0ba27502007-12-24 16:55:16 +0100375values are generally expressed in milliseconds (unless explicitly stated
376otherwise) but may be expressed in any other unit by suffixing the unit to the
377numeric value. It is important to consider this because it will not be repeated
378for every keyword. Supported units are :
379
380 - us : microseconds. 1 microsecond = 1/1000000 second
381 - ms : milliseconds. 1 millisecond = 1/1000 second. This is the default.
382 - s : seconds. 1s = 1000ms
383 - m : minutes. 1m = 60s = 60000ms
384 - h : hours. 1h = 60m = 3600s = 3600000ms
385 - d : days. 1d = 24h = 1440m = 86400s = 86400000ms
386
387
Patrick Mezard35da19c2010-06-12 17:02:47 +02003882.3. Examples
389-------------
390
391 # Simple configuration for an HTTP proxy listening on port 80 on all
392 # interfaces and forwarding requests to a single backend "servers" with a
393 # single server "server1" listening on 127.0.0.1:8000
394 global
395 daemon
396 maxconn 256
397
398 defaults
399 mode http
400 timeout connect 5000ms
401 timeout client 50000ms
402 timeout server 50000ms
403
404 frontend http-in
405 bind *:80
406 default_backend servers
407
408 backend servers
409 server server1 127.0.0.1:8000 maxconn 32
410
411
412 # The same configuration defined with a single listen block. Shorter but
413 # less expressive, especially in HTTP mode.
414 global
415 daemon
416 maxconn 256
417
418 defaults
419 mode http
420 timeout connect 5000ms
421 timeout client 50000ms
422 timeout server 50000ms
423
424 listen http-in
425 bind *:80
426 server server1 127.0.0.1:8000 maxconn 32
427
428
429Assuming haproxy is in $PATH, test these configurations in a shell with:
430
Willy Tarreauccb289d2010-12-11 20:19:38 +0100431 $ sudo haproxy -f configuration.conf -c
Patrick Mezard35da19c2010-06-12 17:02:47 +0200432
433
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004343. Global parameters
Willy Tarreau6a06a402007-07-15 20:15:28 +0200435--------------------
436
437Parameters in the "global" section are process-wide and often OS-specific. They
438are generally set once for all and do not need being changed once correct. Some
439of them have command-line equivalents.
440
441The following keywords are supported in the "global" section :
442
443 * Process management and security
Emeric Brunc8e8d122012-10-02 18:42:10 +0200444 - ca-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200445 - chroot
Emeric Brunc8e8d122012-10-02 18:42:10 +0200446 - crt-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200447 - daemon
448 - gid
449 - group
450 - log
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100451 - log-send-hostname
Willy Tarreau6a06a402007-07-15 20:15:28 +0200452 - nbproc
453 - pidfile
454 - uid
455 - ulimit-n
456 - user
Willy Tarreaufbee7132007-10-18 13:53:22 +0200457 - stats
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +0200458 - node
459 - description
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100460 - unix-bind
Willy Tarreaud72758d2010-01-12 10:42:19 +0100461
Willy Tarreau6a06a402007-07-15 20:15:28 +0200462 * Performance tuning
463 - maxconn
Willy Tarreau81c25d02011-09-07 15:17:21 +0200464 - maxconnrate
William Lallemandd85f9172012-11-09 17:05:39 +0100465 - maxcomprate
William Lallemand072a2bf2012-11-20 17:01:01 +0100466 - maxcompcpuusage
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100467 - maxpipes
Willy Tarreau403edff2012-09-06 11:58:37 +0200468 - maxsslconn
Willy Tarreau6a06a402007-07-15 20:15:28 +0200469 - noepoll
470 - nokqueue
471 - nopoll
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100472 - nosplice
Willy Tarreaufe255b72007-10-14 23:09:26 +0200473 - spread-checks
Willy Tarreau27a674e2009-08-17 07:23:33 +0200474 - tune.bufsize
Willy Tarreau43961d52010-10-04 20:39:20 +0200475 - tune.chksize
William Lallemandf3747832012-11-09 12:33:10 +0100476 - tune.comp.maxlevel
Willy Tarreau193b8c62012-11-22 00:17:38 +0100477 - tune.http.cookielen
Willy Tarreauac1932d2011-10-24 19:14:41 +0200478 - tune.http.maxhdr
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100479 - tune.maxaccept
480 - tune.maxpollevents
Willy Tarreau27a674e2009-08-17 07:23:33 +0200481 - tune.maxrewrite
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200482 - tune.pipesize
Willy Tarreaue803de22010-01-21 17:43:04 +0100483 - tune.rcvbuf.client
484 - tune.rcvbuf.server
485 - tune.sndbuf.client
486 - tune.sndbuf.server
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100487 - tune.ssl.cachesize
Willy Tarreaubfd59462013-02-21 07:46:09 +0100488 - tune.ssl.lifetime
489 - tune.ssl.maxrecord
William Lallemanda509e4c2012-11-07 16:54:34 +0100490 - tune.zlib.memlevel
491 - tune.zlib.windowsize
Willy Tarreaud72758d2010-01-12 10:42:19 +0100492
Willy Tarreau6a06a402007-07-15 20:15:28 +0200493 * Debugging
494 - debug
495 - quiet
Willy Tarreau6a06a402007-07-15 20:15:28 +0200496
497
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004983.1. Process management and security
Willy Tarreau6a06a402007-07-15 20:15:28 +0200499------------------------------------
500
Emeric Brunc8e8d122012-10-02 18:42:10 +0200501ca-base <dir>
502 Assigns a default directory to fetch SSL CA certificates and CRLs from when a
Emeric Brunfd33a262012-10-11 16:28:27 +0200503 relative path is used with "ca-file" or "crl-file" directives. Absolute
504 locations specified in "ca-file" and "crl-file" prevail and ignore "ca-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +0200505
Willy Tarreau6a06a402007-07-15 20:15:28 +0200506chroot <jail dir>
507 Changes current directory to <jail dir> and performs a chroot() there before
508 dropping privileges. This increases the security level in case an unknown
509 vulnerability would be exploited, since it would make it very hard for the
510 attacker to exploit the system. This only works when the process is started
511 with superuser privileges. It is important to ensure that <jail_dir> is both
512 empty and unwritable to anyone.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100513
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100514cpu-map <"all"|"odd"|"even"|process_num> <cpu-set>...
515 On Linux 2.6 and above, it is possible to bind a process to a specific CPU
516 set. This means that the process will never run on other CPUs. The "cpu-map"
517 directive specifies CPU sets for process sets. The first argument is the
518 process number to bind. This process must have a number between 1 and 32,
519 and any process IDs above nbproc are ignored. It is possible to specify all
520 processes at once using "all", only odd numbers using "odd" or even numbers
521 using "even", just like with the "bind-process" directive. The second and
522 forthcoming arguments are CPU sets. Each CPU set is either a unique number
523 between 0 and 31 or a range with two such numbers delimited by a dash ('-').
524 Multiple CPU numbers or ranges may be specified, and the processes will be
525 allowed to bind to all of them. Obviously, multiple "cpu-map" directives may
526 be specified. Each "cpu-map" directive will replace the previous ones when
527 they overlap.
528
Emeric Brunc8e8d122012-10-02 18:42:10 +0200529crt-base <dir>
530 Assigns a default directory to fetch SSL certificates from when a relative
531 path is used with "crtfile" directives. Absolute locations specified after
532 "crtfile" prevail and ignore "crt-base".
533
Willy Tarreau6a06a402007-07-15 20:15:28 +0200534daemon
535 Makes the process fork into background. This is the recommended mode of
536 operation. It is equivalent to the command line "-D" argument. It can be
537 disabled by the command line "-db" argument.
538
539gid <number>
540 Changes the process' group ID to <number>. It is recommended that the group
541 ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
542 be started with a user belonging to this group, or with superuser privileges.
Michael Schererab012dd2013-01-12 18:35:19 +0100543 Note that if haproxy is started from a user having supplementary groups, it
544 will only be able to drop these groups if started with superuser privileges.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200545 See also "group" and "uid".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100546
Willy Tarreau6a06a402007-07-15 20:15:28 +0200547group <group name>
548 Similar to "gid" but uses the GID of group name <group name> from /etc/group.
549 See also "gid" and "user".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100550
Willy Tarreauf7edefa2009-05-10 17:20:05 +0200551log <address> <facility> [max level [min level]]
Willy Tarreau6a06a402007-07-15 20:15:28 +0200552 Adds a global syslog server. Up to two global servers can be defined. They
553 will receive logs for startups and exits, as well as all logs from proxies
Robert Tsai81ae1952007-12-05 10:47:29 +0100554 configured with "log global".
555
556 <address> can be one of:
557
Willy Tarreau2769aa02007-12-27 18:26:09 +0100558 - An IPv4 address optionally followed by a colon and a UDP port. If
Robert Tsai81ae1952007-12-05 10:47:29 +0100559 no port is specified, 514 is used by default (the standard syslog
560 port).
561
David du Colombier24bb5f52011-03-17 10:40:23 +0100562 - An IPv6 address followed by a colon and optionally a UDP port. If
563 no port is specified, 514 is used by default (the standard syslog
564 port).
565
Robert Tsai81ae1952007-12-05 10:47:29 +0100566 - A filesystem path to a UNIX domain socket, keeping in mind
567 considerations for chroot (be sure the path is accessible inside
568 the chroot) and uid/gid (be sure the path is appropriately
569 writeable).
570
Willy Tarreaudad36a32013-03-11 01:20:04 +0100571 Any part of the address string may reference any number of environment
572 variables by preceding their name with a dollar sign ('$') and
573 optionally enclosing them with braces ('{}'), similarly to what is done
574 in Bourne shell.
575
Robert Tsai81ae1952007-12-05 10:47:29 +0100576 <facility> must be one of the 24 standard syslog facilities :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200577
578 kern user mail daemon auth syslog lpr news
579 uucp cron auth2 ftp ntp audit alert cron2
580 local0 local1 local2 local3 local4 local5 local6 local7
581
582 An optional level can be specified to filter outgoing messages. By default,
Willy Tarreauf7edefa2009-05-10 17:20:05 +0200583 all messages are sent. If a maximum level is specified, only messages with a
584 severity at least as important as this level will be sent. An optional minimum
585 level can be specified. If it is set, logs emitted with a more severe level
586 than this one will be capped to this level. This is used to avoid sending
587 "emerg" messages on all terminals on some default syslog configurations.
588 Eight levels are known :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200589
Cyril Bontédc4d9032012-04-08 21:57:39 +0200590 emerg alert crit err warning notice info debug
Willy Tarreau6a06a402007-07-15 20:15:28 +0200591
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100592log-send-hostname [<string>]
593 Sets the hostname field in the syslog header. If optional "string" parameter
594 is set the header is set to the string contents, otherwise uses the hostname
595 of the system. Generally used if one is not relaying logs through an
596 intermediate syslog server or for simply customizing the hostname printed in
597 the logs.
598
Kevinm48936af2010-12-22 16:08:21 +0000599log-tag <string>
600 Sets the tag field in the syslog header to this string. It defaults to the
601 program name as launched from the command line, which usually is "haproxy".
602 Sometimes it can be useful to differentiate between multiple processes
603 running on the same host.
604
Willy Tarreau6a06a402007-07-15 20:15:28 +0200605nbproc <number>
606 Creates <number> processes when going daemon. This requires the "daemon"
607 mode. By default, only one process is created, which is the recommended mode
608 of operation. For systems limited to small sets of file descriptors per
609 process, it may be needed to fork multiple daemons. USING MULTIPLE PROCESSES
610 IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. See also "daemon".
611
612pidfile <pidfile>
613 Writes pids of all daemons into file <pidfile>. This option is equivalent to
614 the "-p" command line argument. The file must be accessible to the user
615 starting the process. See also "daemon".
616
Willy Tarreau110ecc12012-11-15 17:50:01 +0100617stats bind-process [ all | odd | even | <number 1-32>[-<number 1-32>] ] ...
Willy Tarreau35b7b162012-10-22 23:17:18 +0200618 Limits the stats socket to a certain set of processes numbers. By default the
619 stats socket is bound to all processes, causing a warning to be emitted when
620 nbproc is greater than 1 because there is no way to select the target process
621 when connecting. However, by using this setting, it becomes possible to pin
622 the stats socket to a specific set of processes, typically the first one. The
623 warning will automatically be disabled when this setting is used, whatever
624 the number of processes used.
625
Willy Tarreauabb175f2012-09-24 12:43:26 +0200626stats socket [<address:port>|<path>] [param*]
627 Binds a UNIX socket to <path> or a TCPv4/v6 address to <address:port>.
628 Connections to this socket will return various statistics outputs and even
629 allow some commands to be issued to change some runtime settings. Please
630 consult section 9.2 "Unix Socket commands" for more details.
Willy Tarreau6162db22009-10-10 17:13:00 +0200631
Willy Tarreauabb175f2012-09-24 12:43:26 +0200632 All parameters supported by "bind" lines are supported, for instance to
633 restrict access to some users or their access rights. Please consult
634 section 5.1 for more information.
Willy Tarreaufbee7132007-10-18 13:53:22 +0200635
636stats timeout <timeout, in milliseconds>
637 The default timeout on the stats socket is set to 10 seconds. It is possible
638 to change this value with "stats timeout". The value must be passed in
Willy Tarreaubefdff12007-12-02 22:27:38 +0100639 milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.
Willy Tarreaufbee7132007-10-18 13:53:22 +0200640
641stats maxconn <connections>
642 By default, the stats socket is limited to 10 concurrent connections. It is
643 possible to change this value with "stats maxconn".
644
Willy Tarreau6a06a402007-07-15 20:15:28 +0200645uid <number>
646 Changes the process' user ID to <number>. It is recommended that the user ID
647 is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
648 be started with superuser privileges in order to be able to switch to another
649 one. See also "gid" and "user".
650
651ulimit-n <number>
652 Sets the maximum number of per-process file-descriptors to <number>. By
653 default, it is automatically computed, so it is recommended not to use this
654 option.
655
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100656unix-bind [ prefix <prefix> ] [ mode <mode> ] [ user <user> ] [ uid <uid> ]
657 [ group <group> ] [ gid <gid> ]
658
659 Fixes common settings to UNIX listening sockets declared in "bind" statements.
660 This is mainly used to simplify declaration of those UNIX sockets and reduce
661 the risk of errors, since those settings are most commonly required but are
662 also process-specific. The <prefix> setting can be used to force all socket
663 path to be relative to that directory. This might be needed to access another
664 component's chroot. Note that those paths are resolved before haproxy chroots
665 itself, so they are absolute. The <mode>, <user>, <uid>, <group> and <gid>
666 all have the same meaning as their homonyms used by the "bind" statement. If
667 both are specified, the "bind" statement has priority, meaning that the
668 "unix-bind" settings may be seen as process-wide default settings.
669
Willy Tarreau6a06a402007-07-15 20:15:28 +0200670user <user name>
671 Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.
672 See also "uid" and "group".
673
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +0200674node <name>
675 Only letters, digits, hyphen and underscore are allowed, like in DNS names.
676
677 This statement is useful in HA configurations where two or more processes or
678 servers share the same IP address. By setting a different node-name on all
679 nodes, it becomes easy to immediately spot what server is handling the
680 traffic.
681
682description <text>
683 Add a text that describes the instance.
684
685 Please note that it is required to escape certain characters (# for example)
686 and this text is inserted into a html page so you should avoid using
687 "<" and ">" characters.
688
Willy Tarreau6a06a402007-07-15 20:15:28 +0200689
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006903.2. Performance tuning
Willy Tarreau6a06a402007-07-15 20:15:28 +0200691-----------------------
692
693maxconn <number>
694 Sets the maximum per-process number of concurrent connections to <number>. It
695 is equivalent to the command-line argument "-n". Proxies will stop accepting
696 connections when this limit is reached. The "ulimit-n" parameter is
697 automatically adjusted according to this value. See also "ulimit-n".
698
Willy Tarreau81c25d02011-09-07 15:17:21 +0200699maxconnrate <number>
700 Sets the maximum per-process number of connections per second to <number>.
701 Proxies will stop accepting connections when this limit is reached. It can be
702 used to limit the global capacity regardless of each frontend capacity. It is
703 important to note that this can only be used as a service protection measure,
704 as there will not necessarily be a fair share between frontends when the
705 limit is reached, so it's a good idea to also limit each frontend to some
706 value close to its expected share. Also, lowering tune.maxaccept can improve
707 fairness.
708
William Lallemandd85f9172012-11-09 17:05:39 +0100709maxcomprate <number>
710 Sets the maximum per-process input compression rate to <number> kilobytes
711 pers second. For each session, if the maximum is reached, the compression
712 level will be decreased during the session. If the maximum is reached at the
713 beginning of a session, the session will not compress at all. If the maximum
714 is not reached, the compression level will be increased up to
715 tune.comp.maxlevel. A value of zero means there is no limit, this is the
716 default value.
717
William Lallemand072a2bf2012-11-20 17:01:01 +0100718maxcompcpuusage <number>
719 Sets the maximum CPU usage HAProxy can reach before stopping the compression
720 for new requests or decreasing the compression level of current requests.
721 It works like 'maxcomprate' but measures CPU usage instead of incoming data
722 bandwidth. The value is expressed in percent of the CPU used by haproxy. In
723 case of multiple processes (nbproc > 1), each process manages its individual
724 usage. A value of 100 disable the limit. The default value is 100. Setting
725 a lower value will prevent the compression work from slowing the whole
726 process down and from introducing high latencies.
727
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100728maxpipes <number>
729 Sets the maximum per-process number of pipes to <number>. Currently, pipes
730 are only used by kernel-based tcp splicing. Since a pipe contains two file
731 descriptors, the "ulimit-n" value will be increased accordingly. The default
732 value is maxconn/4, which seems to be more than enough for most heavy usages.
733 The splice code dynamically allocates and releases pipes, and can fall back
734 to standard copy, so setting this value too low may only impact performance.
735
Willy Tarreau403edff2012-09-06 11:58:37 +0200736maxsslconn <number>
737 Sets the maximum per-process number of concurrent SSL connections to
738 <number>. By default there is no SSL-specific limit, which means that the
739 global maxconn setting will apply to all connections. Setting this limit
740 avoids having openssl use too much memory and crash when malloc returns NULL
741 (since it unfortunately does not reliably check for such conditions). Note
742 that the limit applies both to incoming and outgoing connections, so one
743 connection which is deciphered then ciphered accounts for 2 SSL connections.
744
William Lallemand9d5f5482012-11-07 16:12:57 +0100745maxzlibmem <number>
746 Sets the maximum amount of RAM in megabytes per process usable by the zlib.
747 When the maximum amount is reached, future sessions will not compress as long
748 as RAM is unavailable. When sets to 0, there is no limit.
William Lallemande3a7d992012-11-20 11:25:20 +0100749 The default value is 0. The value is available in bytes on the UNIX socket
750 with "show info" on the line "MaxZlibMemUsage", the memory used by zlib is
751 "ZlibMemUsage" in bytes.
752
Willy Tarreau6a06a402007-07-15 20:15:28 +0200753noepoll
754 Disables the use of the "epoll" event polling system on Linux. It is
755 equivalent to the command-line argument "-de". The next polling system
Willy Tarreaue9f49e72012-11-11 17:42:00 +0100756 used will generally be "poll". See also "nopoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +0200757
758nokqueue
759 Disables the use of the "kqueue" event polling system on BSD. It is
760 equivalent to the command-line argument "-dk". The next polling system
761 used will generally be "poll". See also "nopoll".
762
763nopoll
764 Disables the use of the "poll" event polling system. It is equivalent to the
765 command-line argument "-dp". The next polling system used will be "select".
Willy Tarreau0ba27502007-12-24 16:55:16 +0100766 It should never be needed to disable "poll" since it's available on all
Willy Tarreaue9f49e72012-11-11 17:42:00 +0100767 platforms supported by HAProxy. See also "nokqueue" and "noepoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +0200768
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100769nosplice
770 Disables the use of kernel tcp splicing between sockets on Linux. It is
771 equivalent to the command line argument "-dS". Data will then be copied
772 using conventional and more portable recv/send calls. Kernel tcp splicing is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100773 limited to some very recent instances of kernel 2.6. Most versions between
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100774 2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not
775 be used. This option makes it easier to globally disable kernel splicing in
776 case of doubt. See also "option splice-auto", "option splice-request" and
777 "option splice-response".
778
Willy Tarreaufe255b72007-10-14 23:09:26 +0200779spread-checks <0..50, in percent>
780 Sometimes it is desirable to avoid sending health checks to servers at exact
781 intervals, for instance when many logical servers are located on the same
782 physical server. With the help of this parameter, it becomes possible to add
783 some randomness in the check interval between 0 and +/- 50%. A value between
784 2 and 5 seems to show good results. The default value remains at 0.
785
Willy Tarreau27a674e2009-08-17 07:23:33 +0200786tune.bufsize <number>
787 Sets the buffer size to this size (in bytes). Lower values allow more
788 sessions to coexist in the same amount of RAM, and higher values allow some
789 applications with very large cookies to work. The default value is 16384 and
790 can be changed at build time. It is strongly recommended not to change this
791 from the default value, as very low values will break some services such as
792 statistics, and values larger than default size will increase memory usage,
793 possibly causing the system to run out of memory. At least the global maxconn
794 parameter should be decreased by the same factor as this one is increased.
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +0400795 If HTTP request is larger than (tune.bufsize - tune.maxrewrite), haproxy will
796 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
797 than this size, haproxy will return HTTP 502 (Bad Gateway).
Willy Tarreau27a674e2009-08-17 07:23:33 +0200798
Willy Tarreau43961d52010-10-04 20:39:20 +0200799tune.chksize <number>
800 Sets the check buffer size to this size (in bytes). Higher values may help
801 find string or regex patterns in very large pages, though doing so may imply
802 more memory and CPU usage. The default value is 16384 and can be changed at
803 build time. It is not recommended to change this value, but to use better
804 checks whenever possible.
805
William Lallemandf3747832012-11-09 12:33:10 +0100806tune.comp.maxlevel <number>
807 Sets the maximum compression level. The compression level affects CPU
808 usage during compression. This value affects CPU usage during compression.
809 Each session using compression initializes the compression algorithm with
810 this value. The default value is 1.
811
Willy Tarreau193b8c62012-11-22 00:17:38 +0100812tune.http.cookielen <number>
813 Sets the maximum length of captured cookies. This is the maximum value that
814 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
815 will automatically be truncated to this one. It is important not to set too
816 high a value because all cookie captures still allocate this size whatever
817 their configured value (they share a same pool). This value is per request
818 per response, so the memory allocated is twice this value per connection.
819 When not specified, the limit is set to 63 characters. It is recommended not
820 to change this value.
821
Willy Tarreauac1932d2011-10-24 19:14:41 +0200822tune.http.maxhdr <number>
823 Sets the maximum number of headers in a request. When a request comes with a
824 number of headers greater than this value (including the first line), it is
825 rejected with a "400 Bad Request" status code. Similarly, too large responses
826 are blocked with "502 Bad Gateway". The default value is 101, which is enough
827 for all usages, considering that the widely deployed Apache server uses the
828 same limit. It can be useful to push this limit further to temporarily allow
829 a buggy application to work by the time it gets fixed. Keep in mind that each
830 new header consumes 32bits of memory for each session, so don't push this
831 limit too high.
832
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100833tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +0100834 Sets the maximum number of consecutive connections a process may accept in a
835 row before switching to other work. In single process mode, higher numbers
836 give better performance at high connection rates. However in multi-process
837 modes, keeping a bit of fairness between processes generally is better to
838 increase performance. This value applies individually to each listener, so
839 that the number of processes a listener is bound to is taken into account.
840 This value defaults to 64. In multi-process mode, it is divided by twice
841 the number of processes the listener is bound to. Setting this value to -1
842 completely disables the limitation. It should normally not be needed to tweak
843 this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100844
845tune.maxpollevents <number>
846 Sets the maximum amount of events that can be processed at once in a call to
847 the polling system. The default value is adapted to the operating system. It
848 has been noticed that reducing it below 200 tends to slightly decrease
849 latency at the expense of network bandwidth, and increasing it above 200
850 tends to trade latency for slightly increased bandwidth.
851
Willy Tarreau27a674e2009-08-17 07:23:33 +0200852tune.maxrewrite <number>
853 Sets the reserved buffer space to this size in bytes. The reserved space is
854 used for header rewriting or appending. The first reads on sockets will never
855 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
856 bufsize, though that does not make much sense since there are rarely large
857 numbers of headers to add. Setting it too high prevents processing of large
858 requests or responses. Setting it too low prevents addition of new headers
859 to already large requests or to POST requests. It is generally wise to set it
860 to about 1024. It is automatically readjusted to half of bufsize if it is
861 larger than that. This means you don't have to worry about it when changing
862 bufsize.
863
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200864tune.pipesize <number>
865 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
866 are the default size for the system. But sometimes when using TCP splicing,
867 it can improve performance to increase pipe sizes, especially if it is
868 suspected that pipes are not filled and that many calls to splice() are
869 performed. This has an impact on the kernel's memory footprint, so this must
870 not be changed if impacts are not understood.
871
Willy Tarreaue803de22010-01-21 17:43:04 +0100872tune.rcvbuf.client <number>
873tune.rcvbuf.server <number>
874 Forces the kernel socket receive buffer size on the client or the server side
875 to the specified value in bytes. This value applies to all TCP/HTTP frontends
876 and backends. It should normally never be set, and the default size (0) lets
877 the kernel autotune this value depending on the amount of available memory.
878 However it can sometimes help to set it to very low values (eg: 4096) in
879 order to save kernel memory by preventing it from buffering too large amounts
880 of received data. Lower values will significantly increase CPU usage though.
881
882tune.sndbuf.client <number>
883tune.sndbuf.server <number>
884 Forces the kernel socket send buffer size on the client or the server side to
885 the specified value in bytes. This value applies to all TCP/HTTP frontends
886 and backends. It should normally never be set, and the default size (0) lets
887 the kernel autotune this value depending on the amount of available memory.
888 However it can sometimes help to set it to very low values (eg: 4096) in
889 order to save kernel memory by preventing it from buffering too large amounts
890 of received data. Lower values will significantly increase CPU usage though.
891 Another use case is to prevent write timeouts with extremely slow clients due
892 to the kernel waiting for a large part of the buffer to be read before
893 notifying haproxy again.
894
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100895tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +0100896 Sets the size of the global SSL session cache, in a number of blocks. A block
897 is large enough to contain an encoded session without peer certificate.
898 An encoded session with peer certificate is stored in multiple blocks
899 depending on the size of the peer certificate. A block use approximatively
900 200 bytes of memory. The default value may be forced at build time, otherwise
901 defaults to 20000. When the cache is full, the most idle entries are purged
902 and reassigned. Higher values reduce the occurrence of such a purge, hence
903 the number of CPU-intensive SSL handshakes by ensuring that all users keep
904 their session as long as possible. All entries are pre-allocated upon startup
Emeric Brun22890a12012-12-28 14:41:32 +0100905 and are shared between all processes if "nbproc" is greater than 1. Setting
906 this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100907
Emeric Brun4f65bff2012-11-16 15:11:00 +0100908tune.ssl.lifetime <timeout>
909 Sets how long a cached SSL session may remain valid. This time is expressed
910 in seconds and defaults to 300 (5 mn). It is important to understand that it
911 does not guarantee that sessions will last that long, because if the cache is
912 full, the longest idle sessions will be purged despite their configured
913 lifetime. The real usefulness of this setting is to prevent sessions from
914 being used for too long.
915
Willy Tarreaubfd59462013-02-21 07:46:09 +0100916tune.ssl.maxrecord <number>
917 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
918 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
919 data only once it has received a full record. With large records, it means
920 that clients might have to download up to 16kB of data before starting to
921 process them. Limiting the value can improve page load times on browsers
922 located over high latency or low bandwidth networks. It is suggested to find
923 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
924 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
925 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
926 2859 gave good results during tests. Use "strace -e trace=write" to find the
927 best value.
928
William Lallemanda509e4c2012-11-07 16:54:34 +0100929tune.zlib.memlevel <number>
930 Sets the memLevel parameter in zlib initialization for each session. It
931 defines how much memory should be allocated for the intenal compression
932 state. A value of 1 uses minimum memory but is slow and reduces compression
933 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
934 between 1 and 9. The default value is 8.
935
936tune.zlib.windowsize <number>
937 Sets the window size (the size of the history buffer) as a parameter of the
938 zlib initialization for each session. Larger values of this parameter result
939 in better compression at the expense of memory usage. Can be a value between
940 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200941
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009423.3. Debugging
943--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +0200944
945debug
946 Enables debug mode which dumps to stdout all exchanges, and disables forking
947 into background. It is the equivalent of the command-line argument "-d". It
948 should never be used in a production configuration since it may prevent full
949 system startup.
950
951quiet
952 Do not display any message during startup. It is equivalent to the command-
953 line argument "-q".
954
Emeric Brunf099e792010-09-27 12:05:28 +0200955
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01009563.4. Userlists
957--------------
958It is possible to control access to frontend/backend/listen sections or to
959http stats by allowing only authenticated and authorized users. To do this,
960it is required to create at least one userlist and to define users.
961
962userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +0100963 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100964 used to store authentication & authorization data for independent customers.
965
966group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +0100967 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100968 attach users to this group by using a comma separated list of names
969 proceeded by "users" keyword.
970
Cyril Bontéf0c60612010-02-06 14:44:47 +0100971user <username> [password|insecure-password <password>]
972 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100973 Adds user <username> to the current userlist. Both secure (encrypted) and
974 insecure (unencrypted) passwords can be used. Encrypted passwords are
Cyril Bonté78caf842010-03-10 22:41:43 +0100975 evaluated using the crypt(3) function so depending of the system's
976 capabilities, different algorithms are supported. For example modern Glibc
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100977 based Linux system supports MD5, SHA-256, SHA-512 and of course classic,
978 DES-based method of crypting passwords.
979
980
981 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +0100982 userlist L1
983 group G1 users tiger,scott
984 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100985
Cyril Bontéf0c60612010-02-06 14:44:47 +0100986 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
987 user scott insecure-password elgato
988 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100989
Cyril Bontéf0c60612010-02-06 14:44:47 +0100990 userlist L2
991 group G1
992 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100993
Cyril Bontéf0c60612010-02-06 14:44:47 +0100994 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
995 user scott insecure-password elgato groups G1,G2
996 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100997
998 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200999
Emeric Brunf099e792010-09-27 12:05:28 +02001000
10013.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02001002----------
Emeric Brunf099e792010-09-27 12:05:28 +02001003It is possible to synchronize server entries in stick tables between several
1004haproxy instances over TCP connections in a multi-master fashion. Each instance
1005pushes its local updates and insertions to remote peers. Server IDs are used to
1006identify servers remotely, so it is important that configurations look similar
1007or at least that the same IDs are forced on each server on all participants.
1008Interrupted exchanges are automatically detected and recovered from the last
1009known point. In addition, during a soft restart, the old process connects to
1010the new one using such a TCP connection to push all its entries before the new
1011process tries to connect to other peers. That ensures very fast replication
1012during a reload, it typically takes a fraction of a second even for large
1013tables.
1014
1015peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001016 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02001017 which is referenced by one or more stick-tables.
1018
1019peer <peername> <ip>:<port>
1020 Defines a peer inside a peers section.
1021 If <peername> is set to the local peer name (by default hostname, or forced
1022 using "-L" command line option), haproxy will listen for incoming remote peer
1023 connection on <ip>:<port>. Otherwise, <ip>:<port> defines where to connect to
1024 to join the remote peer, and <peername> is used at the protocol level to
1025 identify and validate the remote peer on the server side.
1026
1027 During a soft restart, local peer <ip>:<port> is used by the old instance to
1028 connect the new one and initiate a complete replication (teaching process).
1029
1030 It is strongly recommended to have the exact same peers declaration on all
1031 peers and to only rely on the "-L" command line argument to change the local
1032 peer name. This makes it easier to maintain coherent configuration files
1033 across all peers.
1034
Willy Tarreaudad36a32013-03-11 01:20:04 +01001035 Any part of the address string may reference any number of environment
1036 variables by preceding their name with a dollar sign ('$') and optionally
1037 enclosing them with braces ('{}'), similarly to what is done in Bourne shell.
1038
Cyril Bontédc4d9032012-04-08 21:57:39 +02001039 Example:
Emeric Brunf099e792010-09-27 12:05:28 +02001040 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001041 peer haproxy1 192.168.0.1:1024
1042 peer haproxy2 192.168.0.2:1024
1043 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02001044
1045 backend mybackend
1046 mode tcp
1047 balance roundrobin
1048 stick-table type ip size 20k peers mypeers
1049 stick on src
1050
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001051 server srv1 192.168.0.30:80
1052 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02001053
1054
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010554. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02001056----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001057
Willy Tarreau6a06a402007-07-15 20:15:28 +02001058Proxy configuration can be located in a set of sections :
1059 - defaults <name>
1060 - frontend <name>
1061 - backend <name>
1062 - listen <name>
1063
1064A "defaults" section sets default parameters for all other sections following
1065its declaration. Those default parameters are reset by the next "defaults"
1066section. See below for the list of parameters which can be set in a "defaults"
Willy Tarreau0ba27502007-12-24 16:55:16 +01001067section. The name is optional but its use is encouraged for better readability.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001068
1069A "frontend" section describes a set of listening sockets accepting client
1070connections.
1071
1072A "backend" section describes a set of servers to which the proxy will connect
1073to forward incoming connections.
1074
1075A "listen" section defines a complete proxy with its frontend and backend
1076parts combined in one section. It is generally useful for TCP-only traffic.
1077
Willy Tarreau0ba27502007-12-24 16:55:16 +01001078All proxy names must be formed from upper and lower case letters, digits,
1079'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
1080case-sensitive, which means that "www" and "WWW" are two different proxies.
1081
1082Historically, all proxy names could overlap, it just caused troubles in the
1083logs. Since the introduction of content switching, it is mandatory that two
1084proxies with overlapping capabilities (frontend/backend) have different names.
1085However, it is still permitted that a frontend and a backend share the same
1086name, as this configuration seems to be commonly encountered.
1087
1088Right now, two major proxy modes are supported : "tcp", also known as layer 4,
1089and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001090bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01001091protocol, and can interact with it by allowing, blocking, switching, adding,
1092modifying, or removing arbitrary contents in requests or responses, based on
1093arbitrary criteria.
1094
Willy Tarreau0ba27502007-12-24 16:55:16 +01001095
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010964.1. Proxy keywords matrix
1097--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001098
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001099The following list of keywords is supported. Most of them may only be used in a
1100limited set of section types. Some of them are marked as "deprecated" because
1101they are inherited from an old syntax which may be confusing or functionally
1102limited, and there are new recommended keywords to replace them. Keywords
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001103marked with "(*)" can be optionally inverted using the "no" prefix, eg. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001104option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02001105and must be disabled for a specific instance. Such options may also be prefixed
1106with "default" in order to restore default settings regardless of what has been
1107specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001108
Willy Tarreau6a06a402007-07-15 20:15:28 +02001109
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001110 keyword defaults frontend listen backend
1111------------------------------------+----------+----------+---------+---------
1112acl - X X X
1113appsession - - X X
1114backlog X X X -
1115balance X - X X
1116bind - X X -
1117bind-process X X X X
1118block - X X X
1119capture cookie - X X -
1120capture request header - X X -
1121capture response header - X X -
1122clitimeout (deprecated) X X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02001123compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001124contimeout (deprecated) X - X X
1125cookie X - X X
1126default-server X - X X
1127default_backend X X X -
1128description - X X X
1129disabled X X X X
1130dispatch - - X X
1131enabled X X X X
1132errorfile X X X X
1133errorloc X X X X
1134errorloc302 X X X X
1135-- keyword -------------------------- defaults - frontend - listen -- backend -
1136errorloc303 X X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001137force-persist - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001138fullconn X - X X
1139grace X X X X
1140hash-type X - X X
1141http-check disable-on-404 X - X X
Willy Tarreaubd741542010-03-16 18:46:54 +01001142http-check expect - - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02001143http-check send-state X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001144http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02001145http-response - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001146id - X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001147ignore-persist - X X X
William Lallemand0f99e342011-10-12 17:50:54 +02001148log (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001149maxconn X X X -
1150mode X X X X
1151monitor fail - X X -
1152monitor-net X X X -
1153monitor-uri X X X -
1154option abortonclose (*) X - X X
1155option accept-invalid-http-request (*) X X X -
1156option accept-invalid-http-response (*) X - X X
1157option allbackups (*) X - X X
1158option checkcache (*) X - X X
1159option clitcpka (*) X X X -
1160option contstats (*) X X X -
1161option dontlog-normal (*) X X X -
1162option dontlognull (*) X X X -
1163option forceclose (*) X X X X
1164-- keyword -------------------------- defaults - frontend - listen -- backend -
1165option forwardfor X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02001166option http-no-delay (*) X X X X
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02001167option http-pretend-keepalive (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001168option http-server-close (*) X X X X
1169option http-use-proxy-header (*) X X X -
1170option httpchk X - X X
1171option httpclose (*) X X X X
1172option httplog X X X X
1173option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001174option independent-streams (*) X X X X
Simon Hormana2b9dad2013-02-12 10:45:54 +09001175option lb-agent-chk X - X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02001176option ldap-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001177option log-health-checks (*) X - X X
1178option log-separate-errors (*) X X X -
1179option logasap (*) X X X -
1180option mysql-check X - X X
Rauf Kuliyev38b41562011-01-04 15:14:13 +01001181option pgsql-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001182option nolinger (*) X X X X
1183option originalto X X X X
1184option persist (*) X - X X
1185option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02001186option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001187option smtpchk X - X X
1188option socket-stats (*) X X X -
1189option splice-auto (*) X X X X
1190option splice-request (*) X X X X
1191option splice-response (*) X X X X
1192option srvtcpka (*) X - X X
1193option ssl-hello-chk X - X X
1194-- keyword -------------------------- defaults - frontend - listen -- backend -
1195option tcp-smart-accept (*) X X X -
1196option tcp-smart-connect (*) X - X X
1197option tcpka X X X X
1198option tcplog X X X X
1199option transparent (*) X - X X
1200persist rdp-cookie X - X X
1201rate-limit sessions X X X -
1202redirect - X X X
1203redisp (deprecated) X - X X
1204redispatch (deprecated) X - X X
1205reqadd - X X X
1206reqallow - X X X
1207reqdel - X X X
1208reqdeny - X X X
1209reqiallow - X X X
1210reqidel - X X X
1211reqideny - X X X
1212reqipass - X X X
1213reqirep - X X X
1214reqisetbe - X X X
1215reqitarpit - X X X
1216reqpass - X X X
1217reqrep - X X X
1218-- keyword -------------------------- defaults - frontend - listen -- backend -
1219reqsetbe - X X X
1220reqtarpit - X X X
1221retries X - X X
1222rspadd - X X X
1223rspdel - X X X
1224rspdeny - X X X
1225rspidel - X X X
1226rspideny - X X X
1227rspirep - X X X
1228rsprep - X X X
1229server - - X X
1230source X - X X
1231srvtimeout (deprecated) X - X X
Cyril Bonté66c327d2010-10-12 00:14:37 +02001232stats admin - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001233stats auth X - X X
1234stats enable X - X X
1235stats hide-version X - X X
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02001236stats http-request - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001237stats realm X - X X
1238stats refresh X - X X
1239stats scope X - X X
1240stats show-desc X - X X
1241stats show-legends X - X X
1242stats show-node X - X X
1243stats uri X - X X
1244-- keyword -------------------------- defaults - frontend - listen -- backend -
1245stick match - - X X
1246stick on - - X X
1247stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02001248stick store-response - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001249stick-table - - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02001250tcp-request connection - X X -
1251tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02001252tcp-request inspect-delay - X X X
Emeric Brun0a3b67f2010-09-24 15:34:53 +02001253tcp-response content - - X X
1254tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001255timeout check X - X X
1256timeout client X X X -
1257timeout clitimeout (deprecated) X X X -
1258timeout connect X - X X
1259timeout contimeout (deprecated) X - X X
1260timeout http-keep-alive X X X X
1261timeout http-request X X X X
1262timeout queue X - X X
1263timeout server X - X X
1264timeout srvtimeout (deprecated) X - X X
1265timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02001266timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001267transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01001268unique-id-format X X X -
1269unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001270use_backend - X X -
Willy Tarreau4a5cade2012-04-05 21:09:48 +02001271use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001272------------------------------------+----------+----------+---------+---------
1273 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02001274
Willy Tarreau0ba27502007-12-24 16:55:16 +01001275
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012764.2. Alphabetically sorted keywords reference
1277---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001278
1279This section provides a description of each keyword and its usage.
1280
1281
1282acl <aclname> <criterion> [flags] [operator] <value> ...
1283 Declare or complete an access list.
1284 May be used in sections : defaults | frontend | listen | backend
1285 no | yes | yes | yes
1286 Example:
1287 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
1288 acl invalid_src src_port 0:1023
1289 acl local_dst hdr(host) -i localhost
1290
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001291 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001292
1293
Cyril Bontéb21570a2009-11-29 20:04:48 +01001294appsession <cookie> len <length> timeout <holdtime>
1295 [request-learn] [prefix] [mode <path-parameters|query-string>]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001296 Define session stickiness on an existing application cookie.
1297 May be used in sections : defaults | frontend | listen | backend
1298 no | no | yes | yes
1299 Arguments :
1300 <cookie> this is the name of the cookie used by the application and which
1301 HAProxy will have to learn for each new session.
1302
Cyril Bontéb21570a2009-11-29 20:04:48 +01001303 <length> this is the max number of characters that will be memorized and
Willy Tarreau0ba27502007-12-24 16:55:16 +01001304 checked in each cookie value.
1305
1306 <holdtime> this is the time after which the cookie will be removed from
1307 memory if unused. If no unit is specified, this time is in
1308 milliseconds.
1309
Cyril Bontébf47aeb2009-10-15 00:15:40 +02001310 request-learn
1311 If this option is specified, then haproxy will be able to learn
1312 the cookie found in the request in case the server does not
1313 specify any in response. This is typically what happens with
1314 PHPSESSID cookies, or when haproxy's session expires before
1315 the application's session and the correct server is selected.
1316 It is recommended to specify this option to improve reliability.
1317
Cyril Bontéb21570a2009-11-29 20:04:48 +01001318 prefix When this option is specified, haproxy will match on the cookie
1319 prefix (or URL parameter prefix). The appsession value is the
1320 data following this prefix.
1321
1322 Example :
1323 appsession ASPSESSIONID len 64 timeout 3h prefix
1324
1325 This will match the cookie ASPSESSIONIDXXXX=XXXXX,
1326 the appsession value will be XXXX=XXXXX.
1327
1328 mode This option allows to change the URL parser mode.
1329 2 modes are currently supported :
1330 - path-parameters :
1331 The parser looks for the appsession in the path parameters
1332 part (each parameter is separated by a semi-colon), which is
1333 convenient for JSESSIONID for example.
1334 This is the default mode if the option is not set.
1335 - query-string :
1336 In this mode, the parser will look for the appsession in the
1337 query string.
1338
Willy Tarreau0ba27502007-12-24 16:55:16 +01001339 When an application cookie is defined in a backend, HAProxy will check when
1340 the server sets such a cookie, and will store its value in a table, and
1341 associate it with the server's identifier. Up to <length> characters from
1342 the value will be retained. On each connection, haproxy will look for this
Cyril Bontéb21570a2009-11-29 20:04:48 +01001343 cookie both in the "Cookie:" headers, and as a URL parameter (depending on
1344 the mode used). If a known value is found, the client will be directed to the
1345 server associated with this value. Otherwise, the load balancing algorithm is
Willy Tarreau0ba27502007-12-24 16:55:16 +01001346 applied. Cookies are automatically removed from memory when they have been
1347 unused for a duration longer than <holdtime>.
1348
1349 The definition of an application cookie is limited to one per backend.
1350
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01001351 Note : Consider not using this feature in multi-process mode (nbproc > 1)
1352 unless you know what you do : memory is not shared between the
1353 processes, which can result in random behaviours.
1354
Willy Tarreau0ba27502007-12-24 16:55:16 +01001355 Example :
1356 appsession JSESSIONID len 52 timeout 3h
1357
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01001358 See also : "cookie", "capture cookie", "balance", "stick", "stick-table",
1359 "ignore-persist", "nbproc" and "bind-process".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001360
1361
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01001362backlog <conns>
1363 Give hints to the system about the approximate listen backlog desired size
1364 May be used in sections : defaults | frontend | listen | backend
1365 yes | yes | yes | no
1366 Arguments :
1367 <conns> is the number of pending connections. Depending on the operating
1368 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02001369 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01001370
1371 In order to protect against SYN flood attacks, one solution is to increase
1372 the system's SYN backlog size. Depending on the system, sometimes it is just
1373 tunable via a system parameter, sometimes it is not adjustable at all, and
1374 sometimes the system relies on hints given by the application at the time of
1375 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
1376 to the listen() syscall. On systems which can make use of this value, it can
1377 sometimes be useful to be able to specify a different value, hence this
1378 backlog parameter.
1379
1380 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
1381 used as a hint and the system accepts up to the smallest greater power of
1382 two, and never more than some limits (usually 32768).
1383
1384 See also : "maxconn" and the target operating system's tuning guide.
1385
1386
Willy Tarreau0ba27502007-12-24 16:55:16 +01001387balance <algorithm> [ <arguments> ]
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001388balance url_param <param> [check_post [<max_wait>]]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001389 Define the load balancing algorithm to be used in a backend.
1390 May be used in sections : defaults | frontend | listen | backend
1391 yes | no | yes | yes
1392 Arguments :
1393 <algorithm> is the algorithm used to select a server when doing load
1394 balancing. This only applies when no persistence information
1395 is available, or when a connection is redispatched to another
1396 server. <algorithm> may be one of the following :
1397
1398 roundrobin Each server is used in turns, according to their weights.
1399 This is the smoothest and fairest algorithm when the server's
1400 processing time remains equally distributed. This algorithm
1401 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02001402 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08001403 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02001404 large farms, when a server becomes up after having been down
1405 for a very short time, it may sometimes take a few hundreds
1406 requests for it to be re-integrated into the farm and start
1407 receiving traffic. This is normal, though very rare. It is
1408 indicated here in case you would have the chance to observe
1409 it, so that you don't worry.
1410
1411 static-rr Each server is used in turns, according to their weights.
1412 This algorithm is as similar to roundrobin except that it is
1413 static, which means that changing a server's weight on the
1414 fly will have no effect. On the other hand, it has no design
1415 limitation on the number of servers, and when a server goes
1416 up, it is always immediately reintroduced into the farm, once
1417 the full map is recomputed. It also uses slightly less CPU to
1418 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01001419
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01001420 leastconn The server with the lowest number of connections receives the
1421 connection. Round-robin is performed within groups of servers
1422 of the same load to ensure that all servers will be used. Use
1423 of this algorithm is recommended where very long sessions are
1424 expected, such as LDAP, SQL, TSE, etc... but is not very well
1425 suited for protocols using short sessions such as HTTP. This
1426 algorithm is dynamic, which means that server weights may be
1427 adjusted on the fly for slow starts for instance.
1428
Willy Tarreauf09c6602012-02-13 17:12:08 +01001429 first The first server with available connection slots receives the
1430 connection. The servers are choosen from the lowest numeric
1431 identifier to the highest (see server parameter "id"), which
1432 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02001433 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01001434 not make sense to use this algorithm without setting maxconn.
1435 The purpose of this algorithm is to always use the smallest
1436 number of servers so that extra servers can be powered off
1437 during non-intensive hours. This algorithm ignores the server
1438 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02001439 or IMAP than HTTP, though it can be useful there too. In
1440 order to use this algorithm efficiently, it is recommended
1441 that a cloud controller regularly checks server usage to turn
1442 them off when unused, and regularly checks backend queue to
1443 turn new servers on when the queue inflates. Alternatively,
1444 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01001445
Willy Tarreau0ba27502007-12-24 16:55:16 +01001446 source The source IP address is hashed and divided by the total
1447 weight of the running servers to designate which server will
1448 receive the request. This ensures that the same client IP
1449 address will always reach the same server as long as no
1450 server goes down or up. If the hash result changes due to the
1451 number of running servers changing, many clients will be
1452 directed to a different server. This algorithm is generally
1453 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001454 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01001455 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001456 static by default, which means that changing a server's
1457 weight on the fly will have no effect, but this can be
1458 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001459
Oskar Stolc8dc41842012-05-19 10:19:54 +01001460 uri This algorithm hashes either the left part of the URI (before
1461 the question mark) or the whole URI (if the "whole" parameter
1462 is present) and divides the hash value by the total weight of
1463 the running servers. The result designates which server will
1464 receive the request. This ensures that the same URI will
1465 always be directed to the same server as long as no server
1466 goes up or down. This is used with proxy caches and
1467 anti-virus proxies in order to maximize the cache hit rate.
1468 Note that this algorithm may only be used in an HTTP backend.
1469 This algorithm is static by default, which means that
1470 changing a server's weight on the fly will have no effect,
1471 but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001472
Oskar Stolc8dc41842012-05-19 10:19:54 +01001473 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02001474 "depth", both followed by a positive integer number. These
1475 options may be helpful when it is needed to balance servers
1476 based on the beginning of the URI only. The "len" parameter
1477 indicates that the algorithm should only consider that many
1478 characters at the beginning of the URI to compute the hash.
1479 Note that having "len" set to 1 rarely makes sense since most
1480 URIs start with a leading "/".
1481
1482 The "depth" parameter indicates the maximum directory depth
1483 to be used to compute the hash. One level is counted for each
1484 slash in the request. If both parameters are specified, the
1485 evaluation stops when either is reached.
1486
Willy Tarreau0ba27502007-12-24 16:55:16 +01001487 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001488 the query string of each HTTP GET request.
1489
1490 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02001491 request entity will be searched for the parameter argument,
1492 when it is not found in a query string after a question mark
1493 ('?') in the URL. Optionally, specify a number of octets to
1494 wait for before attempting to search the message body. If the
1495 entity can not be searched, then round robin is used for each
1496 request. For instance, if your clients always send the LB
1497 parameter in the first 128 bytes, then specify that. The
1498 default is 48. The entity data will not be scanned until the
1499 required number of octets have arrived at the gateway, this
1500 is the minimum of: (default/max_wait, Content-Length or first
1501 chunk length). If Content-Length is missing or zero, it does
1502 not need to wait for more data than the client promised to
1503 send. When Content-Length is present and larger than
1504 <max_wait>, then waiting is limited to <max_wait> and it is
1505 assumed that this will be enough data to search for the
1506 presence of the parameter. In the unlikely event that
1507 Transfer-Encoding: chunked is used, only the first chunk is
1508 scanned. Parameter values separated by a chunk boundary, may
1509 be randomly balanced if at all.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001510
1511 If the parameter is found followed by an equal sign ('=') and
1512 a value, then the value is hashed and divided by the total
1513 weight of the running servers. The result designates which
1514 server will receive the request.
1515
1516 This is used to track user identifiers in requests and ensure
1517 that a same user ID will always be sent to the same server as
1518 long as no server goes up or down. If no value is found or if
1519 the parameter is not found, then a round robin algorithm is
1520 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001521 backend. This algorithm is static by default, which means
1522 that changing a server's weight on the fly will have no
1523 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001524
Cyril Bontédc4d9032012-04-08 21:57:39 +02001525 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
1526 request. Just as with the equivalent ACL 'hdr()' function,
1527 the header name in parenthesis is not case sensitive. If the
1528 header is absent or if it does not contain any value, the
1529 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01001530
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001531 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01001532 reducing the hash algorithm to the main domain part with some
1533 specific headers such as 'Host'. For instance, in the Host
1534 value "haproxy.1wt.eu", only "1wt" will be considered.
1535
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001536 This algorithm is static by default, which means that
1537 changing a server's weight on the fly will have no effect,
1538 but this can be changed using "hash-type".
1539
Emeric Brun736aa232009-06-30 17:56:00 +02001540 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02001541 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02001542 The RDP cookie <name> (or "mstshash" if omitted) will be
1543 looked up and hashed for each incoming TCP request. Just as
1544 with the equivalent ACL 'req_rdp_cookie()' function, the name
1545 is not case-sensitive. This mechanism is useful as a degraded
1546 persistence mode, as it makes it possible to always send the
1547 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001548 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02001549 used instead.
1550
1551 Note that for this to work, the frontend must ensure that an
1552 RDP cookie is already present in the request buffer. For this
1553 you must use 'tcp-request content accept' rule combined with
1554 a 'req_rdp_cookie_cnt' ACL.
1555
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001556 This algorithm is static by default, which means that
1557 changing a server's weight on the fly will have no effect,
1558 but this can be changed using "hash-type".
1559
Cyril Bontédc4d9032012-04-08 21:57:39 +02001560 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09001561
Willy Tarreau0ba27502007-12-24 16:55:16 +01001562 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02001563 algorithms. Right now, only "url_param" and "uri" support an
1564 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001565
Marek Majkowski9c30fc12008-04-27 23:25:55 +02001566 balance uri [len <len>] [depth <depth>]
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001567 balance url_param <param> [check_post [<max_wait>]]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001568
Willy Tarreau3cd9af22009-03-15 14:06:41 +01001569 The load balancing algorithm of a backend is set to roundrobin when no other
1570 algorithm, mode nor option have been set. The algorithm may only be set once
1571 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001572
1573 Examples :
1574 balance roundrobin
1575 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001576 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01001577 balance hdr(User-Agent)
1578 balance hdr(host)
1579 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001580
1581 Note: the following caveats and limitations on using the "check_post"
1582 extension with "url_param" must be considered :
1583
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001584 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001585 to determine if the parameters will be found in the body or entity which
1586 may contain binary data. Therefore another method may be required to
1587 restrict consideration of POST requests that have no URL parameters in
1588 the body. (see acl reqideny http_end)
1589
1590 - using a <max_wait> value larger than the request buffer size does not
1591 make sense and is useless. The buffer size is set at build time, and
1592 defaults to 16 kB.
1593
1594 - Content-Encoding is not supported, the parameter search will probably
1595 fail; and load balancing will fall back to Round Robin.
1596
1597 - Expect: 100-continue is not supported, load balancing will fall back to
1598 Round Robin.
1599
1600 - Transfer-Encoding (RFC2616 3.6.1) is only supported in the first chunk.
1601 If the entire parameter value is not present in the first chunk, the
1602 selection of server is undefined (actually, defined by how little
1603 actually appeared in the first chunk).
1604
1605 - This feature does not support generation of a 100, 411 or 501 response.
1606
1607 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001608 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001609 white space or control characters are found, indicating the end of what
1610 might be a URL parameter list. This is probably not a concern with SGML
1611 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001612
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001613 See also : "dispatch", "cookie", "appsession", "transparent", "hash-type" and
1614 "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001615
1616
Willy Tarreaub6205fd2012-09-24 12:27:33 +02001617bind [<address>]:<port_range> [, ...] [param*]
1618bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001619 Define one or several listening addresses and/or ports in a frontend.
1620 May be used in sections : defaults | frontend | listen | backend
1621 no | yes | yes | no
1622 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01001623 <address> is optional and can be a host name, an IPv4 address, an IPv6
1624 address, or '*'. It designates the address the frontend will
1625 listen on. If unset, all IPv4 addresses of the system will be
1626 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01001627 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01001628 Optionally, an address family prefix may be used before the
1629 address to force the family regardless of the address format,
1630 which can be useful to specify a path to a unix socket with
1631 no slash ('/'). Currently supported prefixes are :
1632 - 'ipv4@' -> address is always IPv4
1633 - 'ipv6@' -> address is always IPv6
1634 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau40aa0702013-03-10 23:51:38 +01001635 - 'fd@<n>' -> use file descriptor <n> inherited from the
1636 parent. The fd must be bound and may or may not already
1637 be listening.
Willy Tarreaudad36a32013-03-11 01:20:04 +01001638 Any part of the address string may reference any number of
1639 environment variables by preceding their name with a dollar
1640 sign ('$') and optionally enclosing them with braces ('{}'),
1641 similarly to what is done in Bourne shell.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01001642
Willy Tarreauc5011ca2010-03-22 11:53:56 +01001643 <port_range> is either a unique TCP port, or a port range for which the
1644 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001645 above. The port is mandatory for TCP listeners. Note that in
1646 the case of an IPv6 address, the port is always the number
1647 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01001648 - a numerical port (ex: '80')
1649 - a dash-delimited ports range explicitly stating the lower
1650 and upper bounds (ex: '2000-2100') which are included in
1651 the range.
1652
1653 Particular care must be taken against port ranges, because
1654 every <address:port> couple consumes one socket (= a file
1655 descriptor), so it's easy to consume lots of descriptors
1656 with a simple range, and to run out of sockets. Also, each
1657 <address:port> couple must be used only once among all
1658 instances running on a same system. Please note that binding
1659 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001660 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01001661 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001662
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001663 <path> is a UNIX socket path beginning with a slash ('/'). This is
1664 alternative to the TCP listening port. Haproxy will then
1665 receive UNIX connections on the socket located at this place.
1666 The path must begin with a slash and by default is absolute.
1667 It can be relative to the prefix defined by "unix-bind" in
1668 the global section. Note that the total length of the prefix
1669 followed by the socket path cannot exceed some system limits
1670 for UNIX sockets, which commonly are set to 107 characters.
1671
Willy Tarreaub6205fd2012-09-24 12:27:33 +02001672 <param*> is a list of parameters common to all sockets declared on the
1673 same line. These numerous parameters depend on OS and build
1674 options and have a complete section dedicated to them. Please
1675 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02001676
Willy Tarreau0ba27502007-12-24 16:55:16 +01001677 It is possible to specify a list of address:port combinations delimited by
1678 commas. The frontend will then listen on all of these addresses. There is no
1679 fixed limit to the number of addresses and ports which can be listened on in
1680 a frontend, as well as there is no limit to the number of "bind" statements
1681 in a frontend.
1682
1683 Example :
1684 listen http_proxy
1685 bind :80,:443
1686 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001687 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01001688
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02001689 listen http_https_proxy
1690 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02001691 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02001692
Willy Tarreau24709282013-03-10 21:32:12 +01001693 listen http_https_proxy_explicit
1694 bind ipv6@:80
1695 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
1696 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
1697
Willy Tarreaudad36a32013-03-11 01:20:04 +01001698 listen external_bind_app1
1699 bind fd@${FD_APP1}
1700
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001701 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02001702 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001703
1704
Willy Tarreau110ecc12012-11-15 17:50:01 +01001705bind-process [ all | odd | even | <number 1-32>[-<number 1-32>] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001706 Limit visibility of an instance to a certain set of processes numbers.
1707 May be used in sections : defaults | frontend | listen | backend
1708 yes | yes | yes | yes
1709 Arguments :
1710 all All process will see this instance. This is the default. It
1711 may be used to override a default value.
1712
1713 odd This instance will be enabled on processes 1,3,5,...31. This
1714 option may be combined with other numbers.
1715
1716 even This instance will be enabled on processes 2,4,6,...32. This
1717 option may be combined with other numbers. Do not use it
1718 with less than 2 processes otherwise some instances might be
1719 missing from all processes.
1720
Willy Tarreau110ecc12012-11-15 17:50:01 +01001721 number The instance will be enabled on this process number or range,
1722 whose values must all be between 1 and 32. You must be
1723 careful not to reference a process number greater than the
1724 configured global.nbproc, otherwise some instances might be
1725 missing from all processes.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001726
1727 This keyword limits binding of certain instances to certain processes. This
1728 is useful in order not to have too many processes listening to the same
1729 ports. For instance, on a dual-core machine, it might make sense to set
1730 'nbproc 2' in the global section, then distributes the listeners among 'odd'
1731 and 'even' instances.
1732
1733 At the moment, it is not possible to reference more than 32 processes using
1734 this keyword, but this should be more than enough for most setups. Please
1735 note that 'all' really means all processes and is not limited to the first
1736 32.
1737
1738 If some backends are referenced by frontends bound to other processes, the
1739 backend automatically inherits the frontend's processes.
1740
1741 Example :
1742 listen app_ip1
1743 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02001744 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001745
1746 listen app_ip2
1747 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02001748 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001749
1750 listen management
1751 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02001752 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001753
Willy Tarreau110ecc12012-11-15 17:50:01 +01001754 listen management
1755 bind 10.0.0.4:80
1756 bind-process 1-4
1757
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001758 See also : "nbproc" in global section.
1759
1760
Willy Tarreau0ba27502007-12-24 16:55:16 +01001761block { if | unless } <condition>
1762 Block a layer 7 request if/unless a condition is matched
1763 May be used in sections : defaults | frontend | listen | backend
1764 no | yes | yes | yes
1765
1766 The HTTP request will be blocked very early in the layer 7 processing
1767 if/unless <condition> is matched. A 403 error will be returned if the request
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001768 is blocked. The condition has to reference ACLs (see section 7). This is
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02001769 typically used to deny access to certain sensitive resources if some
Willy Tarreau0ba27502007-12-24 16:55:16 +01001770 conditions are met or not met. There is no fixed limit to the number of
1771 "block" statements per instance.
1772
1773 Example:
1774 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
1775 acl invalid_src src_port 0:1023
1776 acl local_dst hdr(host) -i localhost
1777 block if invalid_src || local_dst
1778
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001779 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001780
1781
1782capture cookie <name> len <length>
1783 Capture and log a cookie in the request and in the response.
1784 May be used in sections : defaults | frontend | listen | backend
1785 no | yes | yes | no
1786 Arguments :
1787 <name> is the beginning of the name of the cookie to capture. In order
1788 to match the exact name, simply suffix the name with an equal
1789 sign ('='). The full name will appear in the logs, which is
1790 useful with application servers which adjust both the cookie name
1791 and value (eg: ASPSESSIONXXXXX).
1792
1793 <length> is the maximum number of characters to report in the logs, which
1794 include the cookie name, the equal sign and the value, all in the
1795 standard "name=value" form. The string will be truncated on the
1796 right if it exceeds <length>.
1797
1798 Only the first cookie is captured. Both the "cookie" request headers and the
1799 "set-cookie" response headers are monitored. This is particularly useful to
1800 check for application bugs causing session crossing or stealing between
1801 users, because generally the user's cookies can only change on a login page.
1802
1803 When the cookie was not presented by the client, the associated log column
1804 will report "-". When a request does not cause a cookie to be assigned by the
1805 server, a "-" is reported in the response column.
1806
1807 The capture is performed in the frontend only because it is necessary that
1808 the log format does not change for a given frontend depending on the
1809 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01001810 "capture cookie" statement in a frontend. The maximum capture length is set
1811 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
1812 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001813
1814 Example:
1815 capture cookie ASPSESSION len 32
1816
1817 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001818 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001819
1820
1821capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01001822 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001823 May be used in sections : defaults | frontend | listen | backend
1824 no | yes | yes | no
1825 Arguments :
1826 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01001827 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01001828 appear in the requests, with the first letter of each word in
1829 upper case. The header name will not appear in the logs, only the
1830 value is reported, but the position in the logs is respected.
1831
1832 <length> is the maximum number of characters to extract from the value and
1833 report in the logs. The string will be truncated on the right if
1834 it exceeds <length>.
1835
Willy Tarreau4460d032012-11-21 23:37:37 +01001836 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01001837 value will be added to the logs between braces ('{}'). If multiple headers
1838 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01001839 in the same order they were declared in the configuration. Non-existent
1840 headers will be logged just as an empty string. Common uses for request
1841 header captures include the "Host" field in virtual hosting environments, the
1842 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001843 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01001844 environments to find where the request came from.
1845
1846 Note that when capturing headers such as "User-agent", some spaces may be
1847 logged, making the log analysis more difficult. Thus be careful about what
1848 you log if you know your log parser is not smart enough to rely on the
1849 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001850
Willy Tarreau0900abb2012-11-22 00:21:46 +01001851 There is no limit to the number of captured request headers nor to their
1852 length, though it is wise to keep them low to limit memory usage per session.
1853 In order to keep log format consistent for a same frontend, header captures
1854 can only be declared in a frontend. It is not possible to specify a capture
1855 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001856
1857 Example:
1858 capture request header Host len 15
1859 capture request header X-Forwarded-For len 15
1860 capture request header Referrer len 15
1861
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001862 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01001863 about logging.
1864
1865
1866capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01001867 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001868 May be used in sections : defaults | frontend | listen | backend
1869 no | yes | yes | no
1870 Arguments :
1871 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01001872 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01001873 appear in the response, with the first letter of each word in
1874 upper case. The header name will not appear in the logs, only the
1875 value is reported, but the position in the logs is respected.
1876
1877 <length> is the maximum number of characters to extract from the value and
1878 report in the logs. The string will be truncated on the right if
1879 it exceeds <length>.
1880
Willy Tarreau4460d032012-11-21 23:37:37 +01001881 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01001882 result will be added to the logs between braces ('{}') after the captured
1883 request headers. If multiple headers are captured, they will be delimited by
1884 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01001885 the configuration. Non-existent headers will be logged just as an empty
1886 string. Common uses for response header captures include the "Content-length"
1887 header which indicates how many bytes are expected to be returned, the
1888 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001889
Willy Tarreau0900abb2012-11-22 00:21:46 +01001890 There is no limit to the number of captured response headers nor to their
1891 length, though it is wise to keep them low to limit memory usage per session.
1892 In order to keep log format consistent for a same frontend, header captures
1893 can only be declared in a frontend. It is not possible to specify a capture
1894 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001895
1896 Example:
1897 capture response header Content-length len 9
1898 capture response header Location len 15
1899
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001900 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01001901 about logging.
1902
1903
Cyril Bontéf0c60612010-02-06 14:44:47 +01001904clitimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01001905 Set the maximum inactivity time on the client side.
1906 May be used in sections : defaults | frontend | listen | backend
1907 yes | yes | yes | no
1908 Arguments :
1909 <timeout> is the timeout value is specified in milliseconds by default, but
1910 can be in any other unit if the number is suffixed by the unit,
1911 as explained at the top of this document.
1912
1913 The inactivity timeout applies when the client is expected to acknowledge or
1914 send data. In HTTP mode, this timeout is particularly important to consider
1915 during the first phase, when the client sends the request, and during the
1916 response while it is reading data sent by the server. The value is specified
1917 in milliseconds by default, but can be in any other unit if the number is
1918 suffixed by the unit, as specified at the top of this document. In TCP mode
1919 (and to a lesser extent, in HTTP mode), it is highly recommended that the
1920 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01001921 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01001922 losses by specifying timeouts that are slightly above multiples of 3 seconds
1923 (eg: 4 or 5 seconds).
1924
1925 This parameter is specific to frontends, but can be specified once for all in
1926 "defaults" sections. This is in fact one of the easiest solutions not to
1927 forget about it. An unspecified timeout results in an infinite timeout, which
1928 is not recommended. Such a usage is accepted and works but reports a warning
1929 during startup because it may results in accumulation of expired sessions in
1930 the system if the system's timeouts are not configured either.
1931
1932 This parameter is provided for compatibility but is currently deprecated.
1933 Please use "timeout client" instead.
1934
Willy Tarreau036fae02008-01-06 13:24:40 +01001935 See also : "timeout client", "timeout http-request", "timeout server", and
1936 "srvtimeout".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001937
Cyril Bonté316a8cf2012-11-11 13:38:27 +01001938compression algo <algorithm> ...
1939compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02001940compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02001941 Enable HTTP compression.
1942 May be used in sections : defaults | frontend | listen | backend
1943 yes | yes | yes | yes
1944 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01001945 algo is followed by the list of supported compression algorithms.
1946 type is followed by the list of MIME types that will be compressed.
1947 offload makes haproxy work as a compression offloader only (see notes).
1948
1949 The currently supported algorithms are :
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04001950 identity this is mostly for debugging, and it was useful for developing
Cyril Bonté316a8cf2012-11-11 13:38:27 +01001951 the compression feature. Identity does not apply any change on
1952 data.
1953
1954 gzip applies gzip compression. This setting is only available when
1955 support for zlib was built in.
1956
1957 deflate same as gzip, but with deflate algorithm and zlib format.
1958 Note that this algorithm has ambiguous support on many browsers
1959 and no support at all from recent ones. It is strongly
1960 recommended not to use it for anything else than experimentation.
1961 This setting is only available when support for zlib was built
1962 in.
1963
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04001964 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01001965 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04001966 If backend servers support HTTP compression, these directives
1967 will be no-op: haproxy will see the compressed response and will not
1968 compress again. If backend servers do not support HTTP compression and
1969 there is Accept-Encoding header in request, haproxy will compress the
1970 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02001971
1972 The "offload" setting makes haproxy remove the Accept-Encoding header to
1973 prevent backend servers from compressing responses. It is strongly
1974 recommended not to do this because this means that all the compression work
1975 will be done on the single point where haproxy is located. However in some
1976 deployment scenarios, haproxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04001977 with broken HTTP compression implementation which can't be turned off.
1978 In that case haproxy can be used to prevent that gateway from emitting
1979 invalid payloads. In this case, simply removing the header in the
1980 configuration does not work because it applies before the header is parsed,
1981 so that prevents haproxy from compressing. The "offload" setting should
1982 then be used for such scenarios.
William Lallemand82fe75c2012-10-23 10:25:10 +02001983
William Lallemand05097442012-11-20 12:14:28 +01001984 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01001985 * the request does not advertise a supported compression algorithm in the
1986 "Accept-Encoding" header
1987 * the response message is not HTTP/1.1
William Lallemandd3002612012-11-26 14:34:47 +01001988 * HTTP status code is not 200
Baptiste Assmann650d53d2013-01-05 15:44:44 +01001989 * response contain neither a "Content-Length" header nor a
1990 "Transfer-Encoding" whose last value is "chunked"
1991 * response contains a "Content-Type" header whose first value starts with
1992 "multipart"
1993 * the response contains the "no-transform" value in the "Cache-control"
1994 header
1995 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
1996 and later
1997 * The response contains a "Content-Encoding" header, indicating that the
1998 response is already compressed (see compression offload)
William Lallemand05097442012-11-20 12:14:28 +01001999
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002000 Note: The compression does not rewrite Etag headers, and does not emit the
2001 Warning header.
William Lallemand05097442012-11-20 12:14:28 +01002002
William Lallemand82fe75c2012-10-23 10:25:10 +02002003 Examples :
2004 compression algo gzip
2005 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01002006
Cyril Bontéf0c60612010-02-06 14:44:47 +01002007contimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002008 Set the maximum time to wait for a connection attempt to a server to succeed.
2009 May be used in sections : defaults | frontend | listen | backend
2010 yes | no | yes | yes
2011 Arguments :
2012 <timeout> is the timeout value is specified in milliseconds by default, but
2013 can be in any other unit if the number is suffixed by the unit,
2014 as explained at the top of this document.
2015
2016 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002017 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01002018 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01002019 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
2020 connect timeout also presets the queue timeout to the same value if this one
2021 has not been specified. Historically, the contimeout was also used to set the
2022 tarpit timeout in a listen section, which is not possible in a pure frontend.
2023
2024 This parameter is specific to backends, but can be specified once for all in
2025 "defaults" sections. This is in fact one of the easiest solutions not to
2026 forget about it. An unspecified timeout results in an infinite timeout, which
2027 is not recommended. Such a usage is accepted and works but reports a warning
2028 during startup because it may results in accumulation of failed sessions in
2029 the system if the system's timeouts are not configured either.
2030
2031 This parameter is provided for backwards compatibility but is currently
2032 deprecated. Please use "timeout connect", "timeout queue" or "timeout tarpit"
2033 instead.
2034
2035 See also : "timeout connect", "timeout queue", "timeout tarpit",
2036 "timeout server", "contimeout".
2037
2038
Willy Tarreau55165fe2009-05-10 12:02:55 +02002039cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02002040 [ postonly ] [ preserve ] [ httponly ] [ secure ]
2041 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002042 Enable cookie-based persistence in a backend.
2043 May be used in sections : defaults | frontend | listen | backend
2044 yes | no | yes | yes
2045 Arguments :
2046 <name> is the name of the cookie which will be monitored, modified or
2047 inserted in order to bring persistence. This cookie is sent to
2048 the client via a "Set-Cookie" header in the response, and is
2049 brought back by the client in a "Cookie" header in all requests.
2050 Special care should be taken to choose a name which does not
2051 conflict with any likely application cookie. Also, if the same
2052 backends are subject to be used by the same clients (eg:
2053 HTTP/HTTPS), care should be taken to use different cookie names
2054 between all backends if persistence between them is not desired.
2055
2056 rewrite This keyword indicates that the cookie will be provided by the
2057 server and that haproxy will have to modify its value to set the
2058 server's identifier in it. This mode is handy when the management
2059 of complex combinations of "Set-cookie" and "Cache-control"
2060 headers is left to the application. The application can then
2061 decide whether or not it is appropriate to emit a persistence
2062 cookie. Since all responses should be monitored, this mode only
2063 works in HTTP close mode. Unless the application behaviour is
2064 very complex and/or broken, it is advised not to start with this
2065 mode for new deployments. This keyword is incompatible with
2066 "insert" and "prefix".
2067
2068 insert This keyword indicates that the persistence cookie will have to
Willy Tarreaua79094d2010-08-31 22:54:15 +02002069 be inserted by haproxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002070
Willy Tarreaua79094d2010-08-31 22:54:15 +02002071 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002072 server. When used without the "preserve" option, if the server
2073 emits a cookie with the same name, it will be remove before
2074 processing. For this reason, this mode can be used to upgrade
2075 existing configurations running in the "rewrite" mode. The cookie
2076 will only be a session cookie and will not be stored on the
2077 client's disk. By default, unless the "indirect" option is added,
2078 the server will see the cookies emitted by the client. Due to
2079 caching effects, it is generally wise to add the "nocache" or
2080 "postonly" keywords (see below). The "insert" keyword is not
2081 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002082
2083 prefix This keyword indicates that instead of relying on a dedicated
2084 cookie for the persistence, an existing one will be completed.
2085 This may be needed in some specific environments where the client
2086 does not support more than one single cookie and the application
2087 already needs it. In this case, whenever the server sets a cookie
2088 named <name>, it will be prefixed with the server's identifier
2089 and a delimiter. The prefix will be removed from all client
2090 requests so that the server still finds the cookie it emitted.
2091 Since all requests and responses are subject to being modified,
2092 this mode requires the HTTP close mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02002093 not compatible with "rewrite" and "insert". Note: it is highly
2094 recommended not to use "indirect" with "prefix", otherwise server
2095 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002096
Willy Tarreaua79094d2010-08-31 22:54:15 +02002097 indirect When this option is specified, no cookie will be emitted to a
2098 client which already has a valid one for the server which has
2099 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002100 it will be removed, unless the "preserve" option is also set. In
2101 "insert" mode, this will additionally remove cookies from the
2102 requests transmitted to the server, making the persistence
2103 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02002104 Note: it is highly recommended not to use "indirect" with
2105 "prefix", otherwise server cookie updates would not be sent to
2106 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002107
2108 nocache This option is recommended in conjunction with the insert mode
2109 when there is a cache between the client and HAProxy, as it
2110 ensures that a cacheable response will be tagged non-cacheable if
2111 a cookie needs to be inserted. This is important because if all
2112 persistence cookies are added on a cacheable home page for
2113 instance, then all customers will then fetch the page from an
2114 outer cache and will all share the same persistence cookie,
2115 leading to one server receiving much more traffic than others.
2116 See also the "insert" and "postonly" options.
2117
2118 postonly This option ensures that cookie insertion will only be performed
2119 on responses to POST requests. It is an alternative to the
2120 "nocache" option, because POST responses are not cacheable, so
2121 this ensures that the persistence cookie will never get cached.
2122 Since most sites do not need any sort of persistence before the
2123 first POST which generally is a login request, this is a very
2124 efficient method to optimize caching without risking to find a
2125 persistence cookie in the cache.
2126 See also the "insert" and "nocache" options.
2127
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002128 preserve This option may only be used with "insert" and/or "indirect". It
2129 allows the server to emit the persistence cookie itself. In this
2130 case, if a cookie is found in the response, haproxy will leave it
2131 untouched. This is useful in order to end persistence after a
2132 logout request for instance. For this, the server just has to
2133 emit a cookie with an invalid value (eg: empty) or with a date in
2134 the past. By combining this mechanism with the "disable-on-404"
2135 check option, it is possible to perform a completely graceful
2136 shutdown because users will definitely leave the server after
2137 they logout.
2138
Willy Tarreau4992dd22012-05-31 21:02:17 +02002139 httponly This option tells haproxy to add an "HttpOnly" cookie attribute
2140 when a cookie is inserted. This attribute is used so that a
2141 user agent doesn't share the cookie with non-HTTP components.
2142 Please check RFC6265 for more information on this attribute.
2143
2144 secure This option tells haproxy to add a "Secure" cookie attribute when
2145 a cookie is inserted. This attribute is used so that a user agent
2146 never emits this cookie over non-secure channels, which means
2147 that a cookie learned with this flag will be presented only over
2148 SSL/TLS connections. Please check RFC6265 for more information on
2149 this attribute.
2150
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002151 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002152 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01002153 name. If the domain begins with a dot, the browser is allowed to
2154 use it for any host ending with that name. It is also possible to
2155 specify several domain names by invoking this option multiple
2156 times. Some browsers might have small limits on the number of
2157 domains, so be careful when doing that. For the record, sending
2158 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002159
Willy Tarreau996a92c2010-10-13 19:30:47 +02002160 maxidle This option allows inserted cookies to be ignored after some idle
2161 time. It only works with insert-mode cookies. When a cookie is
2162 sent to the client, the date this cookie was emitted is sent too.
2163 Upon further presentations of this cookie, if the date is older
2164 than the delay indicated by the parameter (in seconds), it will
2165 be ignored. Otherwise, it will be refreshed if needed when the
2166 response is sent to the client. This is particularly useful to
2167 prevent users who never close their browsers from remaining for
2168 too long on the same server (eg: after a farm size change). When
2169 this option is set and a cookie has no date, it is always
2170 accepted, but gets refreshed in the response. This maintains the
2171 ability for admins to access their sites. Cookies that have a
2172 date in the future further than 24 hours are ignored. Doing so
2173 lets admins fix timezone issues without risking kicking users off
2174 the site.
2175
2176 maxlife This option allows inserted cookies to be ignored after some life
2177 time, whether they're in use or not. It only works with insert
2178 mode cookies. When a cookie is first sent to the client, the date
2179 this cookie was emitted is sent too. Upon further presentations
2180 of this cookie, if the date is older than the delay indicated by
2181 the parameter (in seconds), it will be ignored. If the cookie in
2182 the request has no date, it is accepted and a date will be set.
2183 Cookies that have a date in the future further than 24 hours are
2184 ignored. Doing so lets admins fix timezone issues without risking
2185 kicking users off the site. Contrary to maxidle, this value is
2186 not refreshed, only the first visit date counts. Both maxidle and
2187 maxlife may be used at the time. This is particularly useful to
2188 prevent users who never close their browsers from remaining for
2189 too long on the same server (eg: after a farm size change). This
2190 is stronger than the maxidle method in that it forces a
2191 redispatch after some absolute delay.
2192
Willy Tarreau0ba27502007-12-24 16:55:16 +01002193 There can be only one persistence cookie per HTTP backend, and it can be
2194 declared in a defaults section. The value of the cookie will be the value
2195 indicated after the "cookie" keyword in a "server" statement. If no cookie
2196 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002197
Willy Tarreau0ba27502007-12-24 16:55:16 +01002198 Examples :
2199 cookie JSESSIONID prefix
2200 cookie SRV insert indirect nocache
2201 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02002202 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01002203
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02002204 See also : "appsession", "balance source", "capture cookie", "server"
Cyril Bonté0d4bf012010-04-25 23:21:46 +02002205 and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002206
Willy Tarreau983e01e2010-01-11 18:42:06 +01002207
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002208default-server [param*]
2209 Change default options for a server in a backend
2210 May be used in sections : defaults | frontend | listen | backend
2211 yes | no | yes | yes
2212 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01002213 <param*> is a list of parameters for this server. The "default-server"
2214 keyword accepts an important number of options and has a complete
2215 section dedicated to it. Please refer to section 5 for more
2216 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002217
Willy Tarreau983e01e2010-01-11 18:42:06 +01002218 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002219 default-server inter 1000 weight 13
2220
2221 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01002222
Willy Tarreau983e01e2010-01-11 18:42:06 +01002223
Willy Tarreau0ba27502007-12-24 16:55:16 +01002224default_backend <backend>
2225 Specify the backend to use when no "use_backend" rule has been matched.
2226 May be used in sections : defaults | frontend | listen | backend
2227 yes | yes | yes | no
2228 Arguments :
2229 <backend> is the name of the backend to use.
2230
2231 When doing content-switching between frontend and backends using the
2232 "use_backend" keyword, it is often useful to indicate which backend will be
2233 used when no rule has matched. It generally is the dynamic backend which
2234 will catch all undetermined requests.
2235
Willy Tarreau0ba27502007-12-24 16:55:16 +01002236 Example :
2237
2238 use_backend dynamic if url_dyn
2239 use_backend static if url_css url_img extension_img
2240 default_backend dynamic
2241
Willy Tarreau2769aa02007-12-27 18:26:09 +01002242 See also : "use_backend", "reqsetbe", "reqisetbe"
2243
Willy Tarreau0ba27502007-12-24 16:55:16 +01002244
2245disabled
2246 Disable a proxy, frontend or backend.
2247 May be used in sections : defaults | frontend | listen | backend
2248 yes | yes | yes | yes
2249 Arguments : none
2250
2251 The "disabled" keyword is used to disable an instance, mainly in order to
2252 liberate a listening port or to temporarily disable a service. The instance
2253 will still be created and its configuration will be checked, but it will be
2254 created in the "stopped" state and will appear as such in the statistics. It
2255 will not receive any traffic nor will it send any health-checks or logs. It
2256 is possible to disable many instances at once by adding the "disabled"
2257 keyword in a "defaults" section.
2258
2259 See also : "enabled"
2260
2261
Willy Tarreau5ce94572010-06-07 14:35:41 +02002262dispatch <address>:<port>
2263 Set a default server address
2264 May be used in sections : defaults | frontend | listen | backend
2265 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02002266 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02002267
2268 <address> is the IPv4 address of the default server. Alternatively, a
2269 resolvable hostname is supported, but this name will be resolved
2270 during start-up.
2271
2272 <ports> is a mandatory port specification. All connections will be sent
2273 to this port, and it is not permitted to use port offsets as is
2274 possible with normal servers.
2275
Willy Tarreau787aed52011-04-15 06:45:37 +02002276 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02002277 server can take the connection. In the past it was used to forward non
2278 persistent connections to an auxiliary load balancer. Due to its simple
2279 syntax, it has also been used for simple TCP relays. It is recommended not to
2280 use it for more clarity, and to use the "server" directive instead.
2281
2282 See also : "server"
2283
2284
Willy Tarreau0ba27502007-12-24 16:55:16 +01002285enabled
2286 Enable a proxy, frontend or backend.
2287 May be used in sections : defaults | frontend | listen | backend
2288 yes | yes | yes | yes
2289 Arguments : none
2290
2291 The "enabled" keyword is used to explicitly enable an instance, when the
2292 defaults has been set to "disabled". This is very rarely used.
2293
2294 See also : "disabled"
2295
2296
2297errorfile <code> <file>
2298 Return a file contents instead of errors generated by HAProxy
2299 May be used in sections : defaults | frontend | listen | backend
2300 yes | yes | yes | yes
2301 Arguments :
2302 <code> is the HTTP status code. Currently, HAProxy is capable of
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002303 generating codes 200, 400, 403, 408, 500, 502, 503, and 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002304
2305 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002306 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01002307 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01002308 error pages, and to use absolute paths, since files are read
2309 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002310
2311 It is important to understand that this keyword is not meant to rewrite
2312 errors returned by the server, but errors detected and returned by HAProxy.
2313 This is why the list of supported errors is limited to a small set.
2314
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002315 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2316
Willy Tarreau0ba27502007-12-24 16:55:16 +01002317 The files are returned verbatim on the TCP socket. This allows any trick such
2318 as redirections to another URL or site, as well as tricks to clean cookies,
2319 force enable or disable caching, etc... The package provides default error
2320 files returning the same contents as default errors.
2321
Willy Tarreau59140a22009-02-22 12:02:30 +01002322 The files should not exceed the configured buffer size (BUFSIZE), which
2323 generally is 8 or 16 kB, otherwise they will be truncated. It is also wise
2324 not to put any reference to local contents (eg: images) in order to avoid
2325 loops between the client and HAProxy when all servers are down, causing an
2326 error to be returned instead of an image. For better HTTP compliance, it is
2327 recommended that all header lines end with CR-LF and not LF alone.
2328
Willy Tarreau0ba27502007-12-24 16:55:16 +01002329 The files are read at the same time as the configuration and kept in memory.
2330 For this reason, the errors continue to be returned even when the process is
2331 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01002332 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01002333 403 status code and interrogating a blocked URL.
2334
2335 See also : "errorloc", "errorloc302", "errorloc303"
2336
Willy Tarreau59140a22009-02-22 12:02:30 +01002337 Example :
2338 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
2339 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
2340 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
2341
Willy Tarreau2769aa02007-12-27 18:26:09 +01002342
2343errorloc <code> <url>
2344errorloc302 <code> <url>
2345 Return an HTTP redirection to a URL instead of errors generated by HAProxy
2346 May be used in sections : defaults | frontend | listen | backend
2347 yes | yes | yes | yes
2348 Arguments :
2349 <code> is the HTTP status code. Currently, HAProxy is capable of
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002350 generating codes 200, 400, 403, 408, 500, 502, 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002351
2352 <url> it is the exact contents of the "Location" header. It may contain
2353 either a relative URI to an error page hosted on the same site,
2354 or an absolute URI designating an error page on another site.
2355 Special care should be given to relative URIs to avoid redirect
2356 loops if the URI itself may generate the same error (eg: 500).
2357
2358 It is important to understand that this keyword is not meant to rewrite
2359 errors returned by the server, but errors detected and returned by HAProxy.
2360 This is why the list of supported errors is limited to a small set.
2361
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002362 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2363
Willy Tarreau2769aa02007-12-27 18:26:09 +01002364 Note that both keyword return the HTTP 302 status code, which tells the
2365 client to fetch the designated URL using the same HTTP method. This can be
2366 quite problematic in case of non-GET methods such as POST, because the URL
2367 sent to the client might not be allowed for something other than GET. To
2368 workaround this problem, please use "errorloc303" which send the HTTP 303
2369 status code, indicating to the client that the URL must be fetched with a GET
2370 request.
2371
2372 See also : "errorfile", "errorloc303"
2373
2374
2375errorloc303 <code> <url>
2376 Return an HTTP redirection to a URL instead of errors generated by HAProxy
2377 May be used in sections : defaults | frontend | listen | backend
2378 yes | yes | yes | yes
2379 Arguments :
2380 <code> is the HTTP status code. Currently, HAProxy is capable of
2381 generating codes 400, 403, 408, 500, 502, 503, and 504.
2382
2383 <url> it is the exact contents of the "Location" header. It may contain
2384 either a relative URI to an error page hosted on the same site,
2385 or an absolute URI designating an error page on another site.
2386 Special care should be given to relative URIs to avoid redirect
2387 loops if the URI itself may generate the same error (eg: 500).
2388
2389 It is important to understand that this keyword is not meant to rewrite
2390 errors returned by the server, but errors detected and returned by HAProxy.
2391 This is why the list of supported errors is limited to a small set.
2392
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002393 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2394
Willy Tarreau2769aa02007-12-27 18:26:09 +01002395 Note that both keyword return the HTTP 303 status code, which tells the
2396 client to fetch the designated URL using the same HTTP GET method. This
2397 solves the usual problems associated with "errorloc" and the 302 code. It is
2398 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002399 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002400
2401 See also : "errorfile", "errorloc", "errorloc302"
2402
2403
Willy Tarreau4de91492010-01-22 19:10:05 +01002404force-persist { if | unless } <condition>
2405 Declare a condition to force persistence on down servers
2406 May be used in sections: defaults | frontend | listen | backend
2407 no | yes | yes | yes
2408
2409 By default, requests are not dispatched to down servers. It is possible to
2410 force this using "option persist", but it is unconditional and redispatches
2411 to a valid server if "option redispatch" is set. That leaves with very little
2412 possibilities to force some requests to reach a server which is artificially
2413 marked down for maintenance operations.
2414
2415 The "force-persist" statement allows one to declare various ACL-based
2416 conditions which, when met, will cause a request to ignore the down status of
2417 a server and still try to connect to it. That makes it possible to start a
2418 server, still replying an error to the health checks, and run a specially
2419 configured browser to test the service. Among the handy methods, one could
2420 use a specific source IP address, or a specific cookie. The cookie also has
2421 the advantage that it can easily be added/removed on the browser from a test
2422 page. Once the service is validated, it is then possible to open the service
2423 to the world by returning a valid response to health checks.
2424
2425 The forced persistence is enabled when an "if" condition is met, or unless an
2426 "unless" condition is met. The final redispatch is always disabled when this
2427 is used.
2428
Cyril Bonté0d4bf012010-04-25 23:21:46 +02002429 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02002430 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01002431
2432
Willy Tarreau2769aa02007-12-27 18:26:09 +01002433fullconn <conns>
2434 Specify at what backend load the servers will reach their maxconn
2435 May be used in sections : defaults | frontend | listen | backend
2436 yes | no | yes | yes
2437 Arguments :
2438 <conns> is the number of connections on the backend which will make the
2439 servers use the maximal number of connections.
2440
Willy Tarreau198a7442008-01-17 12:05:32 +01002441 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01002442 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01002443 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01002444 load. The server will then always accept at least <minconn> connections,
2445 never more than <maxconn>, and the limit will be on the ramp between both
2446 values when the backend has less than <conns> concurrent connections. This
2447 makes it possible to limit the load on the servers during normal loads, but
2448 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002449 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002450
Willy Tarreaufbb78422011-06-05 15:38:35 +02002451 Since it's hard to get this value right, haproxy automatically sets it to
2452 10% of the sum of the maxconns of all frontends that may branch to this
2453 backend. That way it's safe to leave it unset.
2454
Willy Tarreau2769aa02007-12-27 18:26:09 +01002455 Example :
2456 # The servers will accept between 100 and 1000 concurrent connections each
2457 # and the maximum of 1000 will be reached when the backend reaches 10000
2458 # connections.
2459 backend dynamic
2460 fullconn 10000
2461 server srv1 dyn1:80 minconn 100 maxconn 1000
2462 server srv2 dyn2:80 minconn 100 maxconn 1000
2463
2464 See also : "maxconn", "server"
2465
2466
2467grace <time>
2468 Maintain a proxy operational for some time after a soft stop
2469 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01002470 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01002471 Arguments :
2472 <time> is the time (by default in milliseconds) for which the instance
2473 will remain operational with the frontend sockets still listening
2474 when a soft-stop is received via the SIGUSR1 signal.
2475
2476 This may be used to ensure that the services disappear in a certain order.
2477 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002478 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01002479 needed by the equipment to detect the failure.
2480
2481 Note that currently, there is very little benefit in using this parameter,
2482 and it may in fact complicate the soft-reconfiguration process more than
2483 simplify it.
2484
Willy Tarreau0ba27502007-12-24 16:55:16 +01002485
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002486hash-type <method>
2487 Specify a method to use for mapping hashes to servers
2488 May be used in sections : defaults | frontend | listen | backend
2489 yes | no | yes | yes
2490 Arguments :
2491 map-based the hash table is a static array containing all alive servers.
2492 The hashes will be very smooth, will consider weights, but will
2493 be static in that weight changes while a server is up will be
2494 ignored. This means that there will be no slow start. Also,
2495 since a server is selected by its position in the array, most
2496 mappings are changed when the server count changes. This means
2497 that when a server goes up or down, or when a server is added
2498 to a farm, most connections will be redistributed to different
2499 servers. This can be inconvenient with caches for instance.
2500
Willy Tarreau798a39c2010-11-24 15:04:29 +01002501 avalanche this mechanism uses the default map-based hashing described
2502 above but applies a full avalanche hash before performing the
2503 mapping. The result is a slightly less smooth hash for most
2504 situations, but the hash becomes better than pure map-based
2505 hashes when the number of servers is a multiple of the size of
2506 the input set. When using URI hash with a number of servers
2507 multiple of 64, it's desirable to change the hash type to
2508 this value.
2509
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002510 consistent the hash table is a tree filled with many occurrences of each
2511 server. The hash key is looked up in the tree and the closest
2512 server is chosen. This hash is dynamic, it supports changing
2513 weights while the servers are up, so it is compatible with the
2514 slow start feature. It has the advantage that when a server
2515 goes up or down, only its associations are moved. When a server
2516 is added to the farm, only a few part of the mappings are
2517 redistributed, making it an ideal algorithm for caches.
2518 However, due to its principle, the algorithm will never be very
2519 smooth and it may sometimes be necessary to adjust a server's
2520 weight or its ID to get a more balanced distribution. In order
2521 to get the same distribution on multiple load balancers, it is
2522 important that all servers have the same IDs.
2523
2524 The default hash type is "map-based" and is recommended for most usages.
2525
2526 See also : "balance", "server"
2527
2528
Willy Tarreau0ba27502007-12-24 16:55:16 +01002529http-check disable-on-404
2530 Enable a maintenance mode upon HTTP/404 response to health-checks
2531 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01002532 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01002533 Arguments : none
2534
2535 When this option is set, a server which returns an HTTP code 404 will be
2536 excluded from further load-balancing, but will still receive persistent
2537 connections. This provides a very convenient method for Web administrators
2538 to perform a graceful shutdown of their servers. It is also important to note
2539 that a server which is detected as failed while it was in this mode will not
2540 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
2541 will immediately be reinserted into the farm. The status on the stats page
2542 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01002543 option only works in conjunction with the "httpchk" option. If this option
2544 is used with "http-check expect", then it has precedence over it so that 404
2545 responses will still be considered as soft-stop.
2546
2547 See also : "option httpchk", "http-check expect"
2548
2549
2550http-check expect [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002551 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01002552 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02002553 yes | no | yes | yes
Willy Tarreaubd741542010-03-16 18:46:54 +01002554 Arguments :
2555 <match> is a keyword indicating how to look for a specific pattern in the
2556 response. The keyword may be one of "status", "rstatus",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002557 "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01002558 exclamation mark ("!") to negate the match. Spaces are allowed
2559 between the exclamation mark and the keyword. See below for more
2560 details on the supported keywords.
2561
2562 <pattern> is the pattern to look for. It may be a string or a regular
2563 expression. If the pattern contains spaces, they must be escaped
2564 with the usual backslash ('\').
2565
2566 By default, "option httpchk" considers that response statuses 2xx and 3xx
2567 are valid, and that others are invalid. When "http-check expect" is used,
2568 it defines what is considered valid or invalid. Only one "http-check"
2569 statement is supported in a backend. If a server fails to respond or times
2570 out, the check obviously fails. The available matches are :
2571
2572 status <string> : test the exact string match for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002573 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01002574 response's status code is exactly this string. If the
2575 "status" keyword is prefixed with "!", then the response
2576 will be considered invalid if the status code matches.
2577
2578 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002579 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01002580 response's status code matches the expression. If the
2581 "rstatus" keyword is prefixed with "!", then the response
2582 will be considered invalid if the status code matches.
2583 This is mostly used to check for multiple codes.
2584
2585 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002586 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01002587 response's body contains this exact string. If the
2588 "string" keyword is prefixed with "!", then the response
2589 will be considered invalid if the body contains this
2590 string. This can be used to look for a mandatory word at
2591 the end of a dynamic page, or to detect a failure when a
2592 specific error appears on the check page (eg: a stack
2593 trace).
2594
2595 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002596 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01002597 response's body matches this expression. If the "rstring"
2598 keyword is prefixed with "!", then the response will be
2599 considered invalid if the body matches the expression.
2600 This can be used to look for a mandatory word at the end
2601 of a dynamic page, or to detect a failure when a specific
2602 error appears on the check page (eg: a stack trace).
2603
2604 It is important to note that the responses will be limited to a certain size
2605 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
2606 Thus, too large responses may not contain the mandatory pattern when using
2607 "string" or "rstring". If a large response is absolutely required, it is
2608 possible to change the default max size by setting the global variable.
2609 However, it is worth keeping in mind that parsing very large responses can
2610 waste some CPU cycles, especially when regular expressions are used, and that
2611 it is always better to focus the checks on smaller resources.
2612
2613 Last, if "http-check expect" is combined with "http-check disable-on-404",
2614 then this last one has precedence when the server responds with 404.
2615
2616 Examples :
2617 # only accept status 200 as valid
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01002618 http-check expect status 200
Willy Tarreaubd741542010-03-16 18:46:54 +01002619
2620 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01002621 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01002622
2623 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01002624 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01002625
2626 # check that we have a correct hexadecimal tag before /html
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01002627 http-check expect rstring <!--tag:[0-9a-f]*</html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01002628
Willy Tarreaubd741542010-03-16 18:46:54 +01002629 See also : "option httpchk", "http-check disable-on-404"
Willy Tarreau2769aa02007-12-27 18:26:09 +01002630
2631
Willy Tarreauef781042010-01-27 11:53:01 +01002632http-check send-state
2633 Enable emission of a state header with HTTP health checks
2634 May be used in sections : defaults | frontend | listen | backend
2635 yes | no | yes | yes
2636 Arguments : none
2637
2638 When this option is set, haproxy will systematically send a special header
2639 "X-Haproxy-Server-State" with a list of parameters indicating to each server
2640 how they are seen by haproxy. This can be used for instance when a server is
2641 manipulated without access to haproxy and the operator needs to know whether
2642 haproxy still sees it up or not, or if the server is the last one in a farm.
2643
2644 The header is composed of fields delimited by semi-colons, the first of which
2645 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
2646 checks on the total number before transition, just as appears in the stats
2647 interface. Next headers are in the form "<variable>=<value>", indicating in
2648 no specific order some values available in the stats interface :
2649 - a variable "name", containing the name of the backend followed by a slash
2650 ("/") then the name of the server. This can be used when a server is
2651 checked in multiple backends.
2652
2653 - a variable "node" containing the name of the haproxy node, as set in the
2654 global "node" variable, otherwise the system's hostname if unspecified.
2655
2656 - a variable "weight" indicating the weight of the server, a slash ("/")
2657 and the total weight of the farm (just counting usable servers). This
2658 helps to know if other servers are available to handle the load when this
2659 one fails.
2660
2661 - a variable "scur" indicating the current number of concurrent connections
2662 on the server, followed by a slash ("/") then the total number of
2663 connections on all servers of the same backend.
2664
2665 - a variable "qcur" indicating the current number of requests in the
2666 server's queue.
2667
2668 Example of a header received by the application server :
2669 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
2670 scur=13/22; qcur=0
2671
2672 See also : "option httpchk", "http-check disable-on-404"
2673
Willy Tarreauccbcc372012-12-27 12:37:57 +01002674http-request { allow | deny | tarpit | auth [realm <realm>] | redirect <rule> |
Willy Tarreauf4c43c12013-06-11 17:01:13 +02002675 add-header <name> <fmt> | set-header <name> <fmt> |
Willy Tarreau51347ed2013-06-11 19:34:13 +02002676 set-nice <nice> | set-log-level <level> | set-tos <tos> |
2677 set-mark <mark> }
Cyril Bontéf0c60612010-02-06 14:44:47 +01002678 [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002679 Access control for Layer 7 requests
2680
2681 May be used in sections: defaults | frontend | listen | backend
2682 no | yes | yes | yes
2683
Willy Tarreau20b0de52012-12-24 15:45:22 +01002684 The http-request statement defines a set of rules which apply to layer 7
2685 processing. The rules are evaluated in their declaration order when they are
2686 met in a frontend, listen or backend section. Any rule may optionally be
2687 followed by an ACL-based condition, in which case it will only be evaluated
2688 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002689
Willy Tarreau20b0de52012-12-24 15:45:22 +01002690 The first keyword is the rule's action. Currently supported actions include :
2691 - "allow" : this stops the evaluation of the rules and lets the request
2692 pass the check. No further "http-request" rules are evaluated.
2693
2694 - "deny" : this stops the evaluation of the rules and immediately rejects
2695 the request and emits an HTTP 403 error. No further "http-request" rules
2696 are evaluated.
2697
Willy Tarreauccbcc372012-12-27 12:37:57 +01002698 - "tarpit" : this stops the evaluation of the rules and immediately blocks
2699 the request without responding for a delay specified by "timeout tarpit"
2700 or "timeout connect" if the former is not set. After that delay, if the
2701 client is still connected, an HTTP error 500 is returned so that the
2702 client does not suspect it has been tarpitted. Logs will report the flags
2703 "PT". The goal of the tarpit rule is to slow down robots during an attack
2704 when they're limited on the number of concurrent requests. It can be very
2705 efficient against very dumb robots, and will significantly reduce the
2706 load on firewalls compared to a "deny" rule. But when facing "correctly"
2707 developped robots, it can make things worse by forcing haproxy and the
2708 front firewall to support insane number of concurrent connections.
2709
Willy Tarreau20b0de52012-12-24 15:45:22 +01002710 - "auth" : this stops the evaluation of the rules and immediately responds
2711 with an HTTP 401 or 407 error code to invite the user to present a valid
2712 user name and password. No further "http-request" rules are evaluated. An
2713 optional "realm" parameter is supported, it sets the authentication realm
2714 that is returned with the response (typically the application's name).
2715
Willy Tarreau81499eb2012-12-27 12:19:02 +01002716 - "redirect" : this performs an HTTP redirection based on a redirect rule.
2717 This is exactly the same as the "redirect" statement except that it
2718 inserts a redirect rule which can be processed in the middle of other
2719 "http-request" rules. See the "redirect" keyword for the rule's syntax.
2720
Willy Tarreau20b0de52012-12-24 15:45:22 +01002721 - "add-header" appends an HTTP header field whose name is specified in
2722 <name> and whose value is defined by <fmt> which follows the log-format
2723 rules (see Custom Log Format in section 8.2.4). This is particularly
2724 useful to pass connection-specific information to the server (eg: the
2725 client's SSL certificate), or to combine several headers into one. This
2726 rule is not final, so it is possible to add other similar rules. Note
2727 that header addition is performed immediately, so one rule might reuse
2728 the resulting header from a previous rule.
2729
2730 - "set-header" does the same as "add-header" except that the header name
2731 is first removed if it existed. This is useful when passing security
2732 information to the server, where the header must not be manipulated by
2733 external users.
2734
Willy Tarreauf4c43c12013-06-11 17:01:13 +02002735 - "set-nice" sets the "nice" factor of the current request being processed.
2736 It only has effect against the other requests being processed at the same
2737 time. The default value is 0, unless altered by the "nice" setting on the
2738 "bind" line. The accepted range is -1024..1024. The higher the value, the
2739 nicest the request will be. Lower values will make the request more
2740 important than other ones. This can be useful to improve the speed of
2741 some requests, or lower the priority of non-important requests. Using
2742 this setting without prior experimentation can cause some major slowdown.
2743
Willy Tarreau9a355ec2013-06-11 17:45:46 +02002744 - "set-log-level" is used to change the log level of the current request
2745 when a certain condition is met. Valid levels are the 8 syslog levels
2746 (see the "log" keyword) plus the special level "silent" which disables
2747 logging for this request. This rule is not final so the last matching
2748 rule wins. This rule can be useful to disable health checks coming from
2749 another equipment.
2750
Willy Tarreau42cf39e2013-06-11 18:51:32 +02002751 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
2752 the client to the value passed in <tos> on platforms which support this.
2753 This value represents the whole 8 bits of the IP TOS field, and can be
2754 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
2755 that only the 6 higher bits are used in DSCP or TOS, and the two lower
2756 bits are always 0. This can be used to adjust some routing behaviour on
2757 border routers based on some information from the request. See RFC 2474,
2758 2597, 3260 and 4594 for more information.
2759
Willy Tarreau51347ed2013-06-11 19:34:13 +02002760 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
2761 client to the value passed in <mark> on platforms which support it. This
2762 value is an unsigned 32 bit value which can be matched by netfilter and
2763 by the routing table. It can be expressed both in decimal or hexadecimal
2764 format (prefixed by "0x"). This can be useful to force certain packets to
2765 take a different route (for example a cheaper network path for bulk
2766 downloads). This works on Linux kernels 2.6.32 and above and requires
2767 admin privileges.
2768
Willy Tarreau20b0de52012-12-24 15:45:22 +01002769 There is no limit to the number of http-request statements per instance.
2770
2771 It is important to know that http-request rules are processed very early in
2772 the HTTP processing, just after "block" rules and before "reqdel" or "reqrep"
2773 rules. That way, headers added by "add-header"/"set-header" are visible by
2774 almost all further ACL rules.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002775
2776 Example:
Cyril Bonté78caf842010-03-10 22:41:43 +01002777 acl nagios src 192.168.129.3
2778 acl local_net src 192.168.0.0/16
2779 acl auth_ok http_auth(L1)
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002780
Cyril Bonté78caf842010-03-10 22:41:43 +01002781 http-request allow if nagios
2782 http-request allow if local_net auth_ok
2783 http-request auth realm Gimme if local_net auth_ok
2784 http-request deny
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002785
Cyril Bonté78caf842010-03-10 22:41:43 +01002786 Example:
2787 acl auth_ok http_auth_group(L1) G1
Cyril Bonté78caf842010-03-10 22:41:43 +01002788 http-request auth unless auth_ok
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002789
Willy Tarreau20b0de52012-12-24 15:45:22 +01002790 Example:
2791 http-request set-header X-Haproxy-Current-Date %T
2792 http-request set-header X-SSL %[ssl_fc]
2793 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id]
2794 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
2795 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
2796 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
2797 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
2798 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
2799 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
2800
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02002801 See also : "stats http-request", section 3.4 about userlists and section 7
2802 about ACL usage.
Willy Tarreauef781042010-01-27 11:53:01 +01002803
Willy Tarreauf4c43c12013-06-11 17:01:13 +02002804http-response { allow | deny | add-header <name> <fmt> | set-nice <nice> |
Willy Tarreau51347ed2013-06-11 19:34:13 +02002805 set-header <name> <fmt> | set-log-level <level> |
Lukas Tribus2dd1d1a2013-06-19 23:34:41 +02002806 set-mark <mark> | set-tos <tos> }
2807 [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02002808 Access control for Layer 7 responses
2809
2810 May be used in sections: defaults | frontend | listen | backend
2811 no | yes | yes | yes
2812
2813 The http-response statement defines a set of rules which apply to layer 7
2814 processing. The rules are evaluated in their declaration order when they are
2815 met in a frontend, listen or backend section. Any rule may optionally be
2816 followed by an ACL-based condition, in which case it will only be evaluated
2817 if the condition is true. Since these rules apply on responses, the backend
2818 rules are applied first, followed by the frontend's rules.
2819
2820 The first keyword is the rule's action. Currently supported actions include :
2821 - "allow" : this stops the evaluation of the rules and lets the response
2822 pass the check. No further "http-response" rules are evaluated for the
2823 current section.
2824
2825 - "deny" : this stops the evaluation of the rules and immediately rejects
2826 the response and emits an HTTP 502 error. No further "http-response"
2827 rules are evaluated.
2828
2829 - "add-header" appends an HTTP header field whose name is specified in
2830 <name> and whose value is defined by <fmt> which follows the log-format
2831 rules (see Custom Log Format in section 8.2.4). This may be used to send
2832 a cookie to a client for example, or to pass some internal information.
2833 This rule is not final, so it is possible to add other similar rules.
2834 Note that header addition is performed immediately, so one rule might
2835 reuse the resulting header from a previous rule.
2836
2837 - "set-header" does the same as "add-header" except that the header name
2838 is first removed if it existed. This is useful when passing security
2839 information to the server, where the header must not be manipulated by
2840 external users.
2841
Willy Tarreauf4c43c12013-06-11 17:01:13 +02002842 - "set-nice" sets the "nice" factor of the current request being processed.
2843 It only has effect against the other requests being processed at the same
2844 time. The default value is 0, unless altered by the "nice" setting on the
2845 "bind" line. The accepted range is -1024..1024. The higher the value, the
2846 nicest the request will be. Lower values will make the request more
2847 important than other ones. This can be useful to improve the speed of
2848 some requests, or lower the priority of non-important requests. Using
2849 this setting without prior experimentation can cause some major slowdown.
2850
Willy Tarreau9a355ec2013-06-11 17:45:46 +02002851 - "set-log-level" is used to change the log level of the current request
2852 when a certain condition is met. Valid levels are the 8 syslog levels
2853 (see the "log" keyword) plus the special level "silent" which disables
2854 logging for this request. This rule is not final so the last matching
2855 rule wins. This rule can be useful to disable health checks coming from
2856 another equipment.
2857
Willy Tarreau42cf39e2013-06-11 18:51:32 +02002858 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
2859 the client to the value passed in <tos> on platforms which support this.
2860 This value represents the whole 8 bits of the IP TOS field, and can be
2861 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
2862 that only the 6 higher bits are used in DSCP or TOS, and the two lower
2863 bits are always 0. This can be used to adjust some routing behaviour on
2864 border routers based on some information from the request. See RFC 2474,
2865 2597, 3260 and 4594 for more information.
2866
Willy Tarreau51347ed2013-06-11 19:34:13 +02002867 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
2868 client to the value passed in <mark> on platforms which support it. This
2869 value is an unsigned 32 bit value which can be matched by netfilter and
2870 by the routing table. It can be expressed both in decimal or hexadecimal
2871 format (prefixed by "0x"). This can be useful to force certain packets to
2872 take a different route (for example a cheaper network path for bulk
2873 downloads). This works on Linux kernels 2.6.32 and above and requires
2874 admin privileges.
2875
Willy Tarreaue365c0b2013-06-11 16:06:12 +02002876 There is no limit to the number of http-response statements per instance.
2877
Godbach09250262013-07-02 01:19:15 +08002878 It is important to know that http-response rules are processed very early in
Willy Tarreaue365c0b2013-06-11 16:06:12 +02002879 the HTTP processing, before "reqdel" or "reqrep" rules. That way, headers
2880 added by "add-header"/"set-header" are visible by almost all further ACL
2881 rules.
2882
2883 See also : "http-request", section 3.4 about userlists and section 7 about
2884 ACL usage.
2885
Mark Lamourinec2247f02012-01-04 13:02:01 -05002886http-send-name-header [<header>]
2887 Add the server name to a request. Use the header string given by <header>
2888
2889 May be used in sections: defaults | frontend | listen | backend
2890 yes | no | yes | yes
2891
2892 Arguments :
2893
2894 <header> The header string to use to send the server name
2895
2896 The "http-send-name-header" statement causes the name of the target
2897 server to be added to the headers of an HTTP request. The name
2898 is added with the header string proved.
2899
2900 See also : "server"
2901
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01002902id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02002903 Set a persistent ID to a proxy.
2904 May be used in sections : defaults | frontend | listen | backend
2905 no | yes | yes | yes
2906 Arguments : none
2907
2908 Set a persistent ID for the proxy. This ID must be unique and positive.
2909 An unused ID will automatically be assigned if unset. The first assigned
2910 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01002911
2912
Cyril Bonté0d4bf012010-04-25 23:21:46 +02002913ignore-persist { if | unless } <condition>
2914 Declare a condition to ignore persistence
2915 May be used in sections: defaults | frontend | listen | backend
2916 no | yes | yes | yes
2917
2918 By default, when cookie persistence is enabled, every requests containing
2919 the cookie are unconditionally persistent (assuming the target server is up
2920 and running).
2921
2922 The "ignore-persist" statement allows one to declare various ACL-based
2923 conditions which, when met, will cause a request to ignore persistence.
2924 This is sometimes useful to load balance requests for static files, which
2925 oftenly don't require persistence. This can also be used to fully disable
2926 persistence for a specific User-Agent (for example, some web crawler bots).
2927
2928 Combined with "appsession", it can also help reduce HAProxy memory usage, as
2929 the appsession table won't grow if persistence is ignored.
2930
2931 The persistence is ignored when an "if" condition is met, or unless an
2932 "unless" condition is met.
2933
2934 See also : "force-persist", "cookie", and section 7 about ACL usage.
2935
2936
Willy Tarreau2769aa02007-12-27 18:26:09 +01002937log global
Willy Tarreauf7edefa2009-05-10 17:20:05 +02002938log <address> <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02002939no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01002940 Enable per-instance logging of events and traffic.
2941 May be used in sections : defaults | frontend | listen | backend
2942 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02002943
2944 Prefix :
2945 no should be used when the logger list must be flushed. For example,
2946 if you don't want to inherit from the default logger list. This
2947 prefix does not allow arguments.
2948
Willy Tarreau2769aa02007-12-27 18:26:09 +01002949 Arguments :
2950 global should be used when the instance's logging parameters are the
2951 same as the global ones. This is the most common usage. "global"
2952 replaces <address>, <facility> and <level> with those of the log
2953 entries found in the "global" section. Only one "log global"
2954 statement may be used per instance, and this form takes no other
2955 parameter.
2956
2957 <address> indicates where to send the logs. It takes the same format as
2958 for the "global" section's logs, and can be one of :
2959
2960 - An IPv4 address optionally followed by a colon (':') and a UDP
2961 port. If no port is specified, 514 is used by default (the
2962 standard syslog port).
2963
David du Colombier24bb5f52011-03-17 10:40:23 +01002964 - An IPv6 address followed by a colon (':') and optionally a UDP
2965 port. If no port is specified, 514 is used by default (the
2966 standard syslog port).
2967
Willy Tarreau2769aa02007-12-27 18:26:09 +01002968 - A filesystem path to a UNIX domain socket, keeping in mind
2969 considerations for chroot (be sure the path is accessible
2970 inside the chroot) and uid/gid (be sure the path is
2971 appropriately writeable).
2972
Willy Tarreaudad36a32013-03-11 01:20:04 +01002973 Any part of the address string may reference any number of
2974 environment variables by preceding their name with a dollar
2975 sign ('$') and optionally enclosing them with braces ('{}'),
2976 similarly to what is done in Bourne shell.
2977
Willy Tarreau2769aa02007-12-27 18:26:09 +01002978 <facility> must be one of the 24 standard syslog facilities :
2979
2980 kern user mail daemon auth syslog lpr news
2981 uucp cron auth2 ftp ntp audit alert cron2
2982 local0 local1 local2 local3 local4 local5 local6 local7
2983
2984 <level> is optional and can be specified to filter outgoing messages. By
2985 default, all messages are sent. If a level is specified, only
2986 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02002987 will be sent. An optional minimum level can be specified. If it
2988 is set, logs emitted with a more severe level than this one will
2989 be capped to this level. This is used to avoid sending "emerg"
2990 messages on all terminals on some default syslog configurations.
2991 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01002992
2993 emerg alert crit err warning notice info debug
2994
William Lallemand0f99e342011-10-12 17:50:54 +02002995 It is important to keep in mind that it is the frontend which decides what to
2996 log from a connection, and that in case of content switching, the log entries
2997 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002998
2999 However, backend log declaration define how and where servers status changes
3000 will be logged. Level "notice" will be used to indicate a server going up,
3001 "warning" will be used for termination signals and definitive service
3002 termination, and "alert" will be used for when a server goes down.
3003
3004 Note : According to RFC3164, messages are truncated to 1024 bytes before
3005 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003006
3007 Example :
3008 log global
Willy Tarreauf7edefa2009-05-10 17:20:05 +02003009 log 127.0.0.1:514 local0 notice # only send important events
3010 log 127.0.0.1:514 local0 notice notice # same but limit output level
Willy Tarreaudad36a32013-03-11 01:20:04 +01003011 log ${LOCAL_SYSLOG}:514 local0 notice # send to local server
3012
Willy Tarreau2769aa02007-12-27 18:26:09 +01003013
William Lallemand48940402012-01-30 16:47:22 +01003014log-format <string>
3015 Allows you to custom a log line.
3016
3017 See also : Custom Log Format (8.2.4)
3018
Willy Tarreau2769aa02007-12-27 18:26:09 +01003019
3020maxconn <conns>
3021 Fix the maximum number of concurrent connections on a frontend
3022 May be used in sections : defaults | frontend | listen | backend
3023 yes | yes | yes | no
3024 Arguments :
3025 <conns> is the maximum number of concurrent connections the frontend will
3026 accept to serve. Excess connections will be queued by the system
3027 in the socket's listen queue and will be served once a connection
3028 closes.
3029
3030 If the system supports it, it can be useful on big sites to raise this limit
3031 very high so that haproxy manages connection queues, instead of leaving the
3032 clients with unanswered connection attempts. This value should not exceed the
3033 global maxconn. Also, keep in mind that a connection contains two buffers
3034 of 8kB each, as well as some other data resulting in about 17 kB of RAM being
3035 consumed per established connection. That means that a medium system equipped
3036 with 1GB of RAM can withstand around 40000-50000 concurrent connections if
3037 properly tuned.
3038
3039 Also, when <conns> is set to large values, it is possible that the servers
3040 are not sized to accept such loads, and for this reason it is generally wise
3041 to assign them some reasonable connection limits.
3042
Vincent Bernat6341be52012-06-27 17:18:30 +02003043 By default, this value is set to 2000.
3044
Willy Tarreau2769aa02007-12-27 18:26:09 +01003045 See also : "server", global section's "maxconn", "fullconn"
3046
3047
3048mode { tcp|http|health }
3049 Set the running mode or protocol of the instance
3050 May be used in sections : defaults | frontend | listen | backend
3051 yes | yes | yes | yes
3052 Arguments :
3053 tcp The instance will work in pure TCP mode. A full-duplex connection
3054 will be established between clients and servers, and no layer 7
3055 examination will be performed. This is the default mode. It
3056 should be used for SSL, SSH, SMTP, ...
3057
3058 http The instance will work in HTTP mode. The client request will be
3059 analyzed in depth before connecting to any server. Any request
3060 which is not RFC-compliant will be rejected. Layer 7 filtering,
3061 processing and switching will be possible. This is the mode which
3062 brings HAProxy most of its value.
3063
3064 health The instance will work in "health" mode. It will just reply "OK"
Willy Tarreau82569f92012-09-27 23:48:56 +02003065 to incoming connections and close the connection. Alternatively,
3066 If the "httpchk" option is set, "HTTP/1.0 200 OK" will be sent
3067 instead. Nothing will be logged in either case. This mode is used
3068 to reply to external components health checks. This mode is
3069 deprecated and should not be used anymore as it is possible to do
3070 the same and even better by combining TCP or HTTP modes with the
3071 "monitor" keyword.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003072
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003073 When doing content switching, it is mandatory that the frontend and the
3074 backend are in the same mode (generally HTTP), otherwise the configuration
3075 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003076
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003077 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01003078 defaults http_instances
3079 mode http
3080
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003081 See also : "monitor", "monitor-net"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003082
Willy Tarreau0ba27502007-12-24 16:55:16 +01003083
Cyril Bontéf0c60612010-02-06 14:44:47 +01003084monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01003085 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003086 May be used in sections : defaults | frontend | listen | backend
3087 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01003088 Arguments :
3089 if <cond> the monitor request will fail if the condition is satisfied,
3090 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003091 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01003092 are met, for instance a low number of servers both in a
3093 backend and its backup.
3094
3095 unless <cond> the monitor request will succeed only if the condition is
3096 satisfied, and will fail otherwise. Such a condition may be
3097 based on a test on the presence of a minimum number of active
3098 servers in a list of backends.
3099
3100 This statement adds a condition which can force the response to a monitor
3101 request to report a failure. By default, when an external component queries
3102 the URI dedicated to monitoring, a 200 response is returned. When one of the
3103 conditions above is met, haproxy will return 503 instead of 200. This is
3104 very useful to report a site failure to an external component which may base
3105 routing advertisements between multiple sites on the availability reported by
3106 haproxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003107 criterion. Note that "monitor fail" only works in HTTP mode. Both status
3108 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003109
3110 Example:
3111 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01003112 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01003113 acl site_dead nbsrv(dynamic) lt 2
3114 acl site_dead nbsrv(static) lt 2
3115 monitor-uri /site_alive
3116 monitor fail if site_dead
3117
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003118 See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003119
3120
3121monitor-net <source>
3122 Declare a source network which is limited to monitor requests
3123 May be used in sections : defaults | frontend | listen | backend
3124 yes | yes | yes | no
3125 Arguments :
3126 <source> is the source IPv4 address or network which will only be able to
3127 get monitor responses to any request. It can be either an IPv4
3128 address, a host name, or an address followed by a slash ('/')
3129 followed by a mask.
3130
3131 In TCP mode, any connection coming from a source matching <source> will cause
3132 the connection to be immediately closed without any log. This allows another
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003133 equipment to probe the port and verify that it is still listening, without
Willy Tarreau2769aa02007-12-27 18:26:09 +01003134 forwarding the connection to a remote server.
3135
3136 In HTTP mode, a connection coming from a source matching <source> will be
3137 accepted, the following response will be sent without waiting for a request,
3138 then the connection will be closed : "HTTP/1.0 200 OK". This is normally
3139 enough for any front-end HTTP probe to detect that the service is UP and
Willy Tarreau82569f92012-09-27 23:48:56 +02003140 running without forwarding the request to a backend server. Note that this
3141 response is sent in raw format, without any transformation. This is important
3142 as it means that it will not be SSL-encrypted on SSL listeners.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003143
Willy Tarreau82569f92012-09-27 23:48:56 +02003144 Monitor requests are processed very early, just after tcp-request connection
3145 ACLs which are the only ones able to block them. These connections are short
3146 lived and never wait for any data from the client. They cannot be logged, and
3147 it is the intended purpose. They are only used to report HAProxy's health to
3148 an upper component, nothing more. Please note that "monitor fail" rules do
3149 not apply to connections intercepted by "monitor-net".
Willy Tarreau2769aa02007-12-27 18:26:09 +01003150
Willy Tarreau95cd2832010-03-04 23:36:33 +01003151 Last, please note that only one "monitor-net" statement can be specified in
3152 a frontend. If more than one is found, only the last one will be considered.
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003153
Willy Tarreau2769aa02007-12-27 18:26:09 +01003154 Example :
3155 # addresses .252 and .253 are just probing us.
3156 frontend www
3157 monitor-net 192.168.0.252/31
3158
3159 See also : "monitor fail", "monitor-uri"
3160
3161
3162monitor-uri <uri>
3163 Intercept a URI used by external components' monitor requests
3164 May be used in sections : defaults | frontend | listen | backend
3165 yes | yes | yes | no
3166 Arguments :
3167 <uri> is the exact URI which we want to intercept to return HAProxy's
3168 health status instead of forwarding the request.
3169
3170 When an HTTP request referencing <uri> will be received on a frontend,
3171 HAProxy will not forward it nor log it, but instead will return either
3172 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
3173 conditions defined with "monitor fail". This is normally enough for any
3174 front-end HTTP probe to detect that the service is UP and running without
3175 forwarding the request to a backend server. Note that the HTTP method, the
3176 version and all headers are ignored, but the request must at least be valid
3177 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
3178
3179 Monitor requests are processed very early. It is not possible to block nor
3180 divert them using ACLs. They cannot be logged either, and it is the intended
3181 purpose. They are only used to report HAProxy's health to an upper component,
3182 nothing more. However, it is possible to add any number of conditions using
3183 "monitor fail" and ACLs so that the result can be adjusted to whatever check
3184 can be imagined (most often the number of available servers in a backend).
3185
3186 Example :
3187 # Use /haproxy_test to report haproxy's status
3188 frontend www
3189 mode http
3190 monitor-uri /haproxy_test
3191
3192 See also : "monitor fail", "monitor-net"
3193
Willy Tarreau0ba27502007-12-24 16:55:16 +01003194
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003195option abortonclose
3196no option abortonclose
3197 Enable or disable early dropping of aborted requests pending in queues.
3198 May be used in sections : defaults | frontend | listen | backend
3199 yes | no | yes | yes
3200 Arguments : none
3201
3202 In presence of very high loads, the servers will take some time to respond.
3203 The per-instance connection queue will inflate, and the response time will
3204 increase respective to the size of the queue times the average per-session
3205 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01003206 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003207 the queue, and slowing down other users, and the servers as well, because the
3208 request will eventually be served, then aborted at the first error
3209 encountered while delivering the response.
3210
3211 As there is no way to distinguish between a full STOP and a simple output
3212 close on the client side, HTTP agents should be conservative and consider
3213 that the client might only have closed its output channel while waiting for
3214 the response. However, this introduces risks of congestion when lots of users
3215 do the same, and is completely useless nowadays because probably no client at
3216 all will close the session while waiting for the response. Some HTTP agents
Willy Tarreaud72758d2010-01-12 10:42:19 +01003217 support this behaviour (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003218 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01003219 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003220 of being the single component to break rare but valid traffic is extremely
3221 low, which adds to the temptation to be able to abort a session early while
3222 still not served and not pollute the servers.
3223
3224 In HAProxy, the user can choose the desired behaviour using the option
3225 "abortonclose". By default (without the option) the behaviour is HTTP
3226 compliant and aborted requests will be served. But when the option is
3227 specified, a session with an incoming channel closed will be aborted while
3228 it is still possible, either pending in the queue for a connection slot, or
3229 during the connection establishment if the server has not yet acknowledged
3230 the connection request. This considerably reduces the queue size and the load
3231 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01003232 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003233
3234 If this option has been enabled in a "defaults" section, it can be disabled
3235 in a specific instance by prepending the "no" keyword before it.
3236
3237 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
3238
3239
Willy Tarreau4076a152009-04-02 15:18:36 +02003240option accept-invalid-http-request
3241no option accept-invalid-http-request
3242 Enable or disable relaxing of HTTP request parsing
3243 May be used in sections : defaults | frontend | listen | backend
3244 yes | yes | yes | no
3245 Arguments : none
3246
3247 By default, HAProxy complies with RFC2616 in terms of message parsing. This
3248 means that invalid characters in header names are not permitted and cause an
3249 error to be returned to the client. This is the desired behaviour as such
3250 forbidden characters are essentially used to build attacks exploiting server
3251 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
3252 server will emit invalid header names for whatever reason (configuration,
3253 implementation) and the issue will not be immediately fixed. In such a case,
3254 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01003255 even if that does not make sense, by specifying this option. Similarly, the
3256 list of characters allowed to appear in a URI is well defined by RFC3986, and
3257 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
3258 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
3259 not allowed at all. Haproxy always blocks a number of them (0..32, 127). The
3260 remaining ones are blocked by default unless this option is enabled.
Willy Tarreau4076a152009-04-02 15:18:36 +02003261
3262 This option should never be enabled by default as it hides application bugs
3263 and open security breaches. It should only be deployed after a problem has
3264 been confirmed.
3265
3266 When this option is enabled, erroneous header names will still be accepted in
3267 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01003268 analysis using the "show errors" request on the UNIX stats socket. Similarly,
3269 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02003270 also helps confirming that the issue has been solved.
3271
3272 If this option has been enabled in a "defaults" section, it can be disabled
3273 in a specific instance by prepending the "no" keyword before it.
3274
3275 See also : "option accept-invalid-http-response" and "show errors" on the
3276 stats socket.
3277
3278
3279option accept-invalid-http-response
3280no option accept-invalid-http-response
3281 Enable or disable relaxing of HTTP response parsing
3282 May be used in sections : defaults | frontend | listen | backend
3283 yes | no | yes | yes
3284 Arguments : none
3285
3286 By default, HAProxy complies with RFC2616 in terms of message parsing. This
3287 means that invalid characters in header names are not permitted and cause an
3288 error to be returned to the client. This is the desired behaviour as such
3289 forbidden characters are essentially used to build attacks exploiting server
3290 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
3291 server will emit invalid header names for whatever reason (configuration,
3292 implementation) and the issue will not be immediately fixed. In such a case,
3293 it is possible to relax HAProxy's header name parser to accept any character
3294 even if that does not make sense, by specifying this option.
3295
3296 This option should never be enabled by default as it hides application bugs
3297 and open security breaches. It should only be deployed after a problem has
3298 been confirmed.
3299
3300 When this option is enabled, erroneous header names will still be accepted in
3301 responses, but the complete response will be captured in order to permit
3302 later analysis using the "show errors" request on the UNIX stats socket.
3303 Doing this also helps confirming that the issue has been solved.
3304
3305 If this option has been enabled in a "defaults" section, it can be disabled
3306 in a specific instance by prepending the "no" keyword before it.
3307
3308 See also : "option accept-invalid-http-request" and "show errors" on the
3309 stats socket.
3310
3311
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003312option allbackups
3313no option allbackups
3314 Use either all backup servers at a time or only the first one
3315 May be used in sections : defaults | frontend | listen | backend
3316 yes | no | yes | yes
3317 Arguments : none
3318
3319 By default, the first operational backup server gets all traffic when normal
3320 servers are all down. Sometimes, it may be preferred to use multiple backups
3321 at once, because one will not be enough. When "option allbackups" is enabled,
3322 the load balancing will be performed among all backup servers when all normal
3323 ones are unavailable. The same load balancing algorithm will be used and the
3324 servers' weights will be respected. Thus, there will not be any priority
3325 order between the backup servers anymore.
3326
3327 This option is mostly used with static server farms dedicated to return a
3328 "sorry" page when an application is completely offline.
3329
3330 If this option has been enabled in a "defaults" section, it can be disabled
3331 in a specific instance by prepending the "no" keyword before it.
3332
3333
3334option checkcache
3335no option checkcache
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003336 Analyze all server responses and block requests with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003337 May be used in sections : defaults | frontend | listen | backend
3338 yes | no | yes | yes
3339 Arguments : none
3340
3341 Some high-level frameworks set application cookies everywhere and do not
3342 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003343 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003344 high risk of session crossing or stealing between users traversing the same
3345 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02003346 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003347
3348 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003349 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01003350 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003351 response to check if there's a risk of caching a cookie on a client-side
3352 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01003353 to the client are :
3354 - all those without "Set-Cookie" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003355 - all those with a return code other than 200, 203, 206, 300, 301, 410,
Willy Tarreau198a7442008-01-17 12:05:32 +01003356 provided that the server has not set a "Cache-control: public" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003357 - all those that come from a POST request, provided that the server has not
3358 set a 'Cache-Control: public' header ;
3359 - those with a 'Pragma: no-cache' header
3360 - those with a 'Cache-control: private' header
3361 - those with a 'Cache-control: no-store' header
3362 - those with a 'Cache-control: max-age=0' header
3363 - those with a 'Cache-control: s-maxage=0' header
3364 - those with a 'Cache-control: no-cache' header
3365 - those with a 'Cache-control: no-cache="set-cookie"' header
3366 - those with a 'Cache-control: no-cache="set-cookie,' header
3367 (allowing other fields after set-cookie)
3368
3369 If a response doesn't respect these requirements, then it will be blocked
Willy Tarreau198a7442008-01-17 12:05:32 +01003370 just as if it was from an "rspdeny" filter, with an "HTTP 502 bad gateway".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003371 The session state shows "PH--" meaning that the proxy blocked the response
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003372 during headers processing. Additionally, an alert will be sent in the logs so
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003373 that admins are informed that there's something to be fixed.
3374
3375 Due to the high impact on the application, the application should be tested
3376 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003377 good practice to always activate it during tests, even if it is not used in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003378 production, as it will report potentially dangerous application behaviours.
3379
3380 If this option has been enabled in a "defaults" section, it can be disabled
3381 in a specific instance by prepending the "no" keyword before it.
3382
3383
3384option clitcpka
3385no option clitcpka
3386 Enable or disable the sending of TCP keepalive packets on the client side
3387 May be used in sections : defaults | frontend | listen | backend
3388 yes | yes | yes | no
3389 Arguments : none
3390
3391 When there is a firewall or any session-aware component between a client and
3392 a server, and when the protocol involves very long sessions with long idle
3393 periods (eg: remote desktops), there is a risk that one of the intermediate
3394 components decides to expire a session which has remained idle for too long.
3395
3396 Enabling socket-level TCP keep-alives makes the system regularly send packets
3397 to the other end of the connection, leaving it active. The delay between
3398 keep-alive probes is controlled by the system only and depends both on the
3399 operating system and its tuning parameters.
3400
3401 It is important to understand that keep-alive packets are neither emitted nor
3402 received at the application level. It is only the network stacks which sees
3403 them. For this reason, even if one side of the proxy already uses keep-alives
3404 to maintain its connection alive, those keep-alive packets will not be
3405 forwarded to the other side of the proxy.
3406
3407 Please note that this has nothing to do with HTTP keep-alive.
3408
3409 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
3410 client side of a connection, which should help when session expirations are
3411 noticed between HAProxy and a client.
3412
3413 If this option has been enabled in a "defaults" section, it can be disabled
3414 in a specific instance by prepending the "no" keyword before it.
3415
3416 See also : "option srvtcpka", "option tcpka"
3417
3418
Willy Tarreau0ba27502007-12-24 16:55:16 +01003419option contstats
3420 Enable continuous traffic statistics updates
3421 May be used in sections : defaults | frontend | listen | backend
3422 yes | yes | yes | no
3423 Arguments : none
3424
3425 By default, counters used for statistics calculation are incremented
3426 only when a session finishes. It works quite well when serving small
3427 objects, but with big ones (for example large images or archives) or
3428 with A/V streaming, a graph generated from haproxy counters looks like
3429 a hedgehog. With this option enabled counters get incremented continuously,
3430 during a whole session. Recounting touches a hotpath directly so
3431 it is not enabled by default, as it has small performance impact (~0.5%).
3432
3433
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02003434option dontlog-normal
3435no option dontlog-normal
3436 Enable or disable logging of normal, successful connections
3437 May be used in sections : defaults | frontend | listen | backend
3438 yes | yes | yes | no
3439 Arguments : none
3440
3441 There are large sites dealing with several thousand connections per second
3442 and for which logging is a major pain. Some of them are even forced to turn
3443 logs off and cannot debug production issues. Setting this option ensures that
3444 normal connections, those which experience no error, no timeout, no retry nor
3445 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
3446 mode, the response status code is checked and return codes 5xx will still be
3447 logged.
3448
3449 It is strongly discouraged to use this option as most of the time, the key to
3450 complex issues is in the normal logs which will not be logged here. If you
3451 need to separate logs, see the "log-separate-errors" option instead.
3452
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003453 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02003454 logging.
3455
3456
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003457option dontlognull
3458no option dontlognull
3459 Enable or disable logging of null connections
3460 May be used in sections : defaults | frontend | listen | backend
3461 yes | yes | yes | no
3462 Arguments : none
3463
3464 In certain environments, there are components which will regularly connect to
3465 various systems to ensure that they are still alive. It can be the case from
3466 another load balancer as well as from monitoring systems. By default, even a
3467 simple port probe or scan will produce a log. If those connections pollute
3468 the logs too much, it is possible to enable option "dontlognull" to indicate
3469 that a connection on which no data has been transferred will not be logged,
3470 which typically corresponds to those probes.
3471
3472 It is generally recommended not to use this option in uncontrolled
3473 environments (eg: internet), otherwise scans and other malicious activities
3474 would not be logged.
3475
3476 If this option has been enabled in a "defaults" section, it can be disabled
3477 in a specific instance by prepending the "no" keyword before it.
3478
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003479 See also : "log", "monitor-net", "monitor-uri" and section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003480
3481
3482option forceclose
3483no option forceclose
3484 Enable or disable active connection closing after response is transferred.
3485 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaua31e5df2009-12-30 01:10:35 +01003486 yes | yes | yes | yes
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003487 Arguments : none
3488
3489 Some HTTP servers do not necessarily close the connections when they receive
3490 the "Connection: close" set by "option httpclose", and if the client does not
3491 close either, then the connection remains open till the timeout expires. This
3492 causes high number of simultaneous connections on the servers and shows high
3493 global session times in the logs.
3494
3495 When this happens, it is possible to use "option forceclose". It will
Willy Tarreau82eeaf22009-12-29 12:09:05 +01003496 actively close the outgoing server channel as soon as the server has finished
Willy Tarreau0dfdf192010-01-05 11:33:11 +01003497 to respond. This option implicitly enables the "httpclose" option. Note that
3498 this option also enables the parsing of the full request and response, which
3499 means we can close the connection to the server very quickly, releasing some
3500 resources earlier than with httpclose.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003501
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02003502 This option may also be combined with "option http-pretend-keepalive", which
3503 will disable sending of the "Connection: close" header, but will still cause
3504 the connection to be closed once the whole response is received.
3505
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003506 If this option has been enabled in a "defaults" section, it can be disabled
3507 in a specific instance by prepending the "no" keyword before it.
3508
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02003509 See also : "option httpclose" and "option http-pretend-keepalive"
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003510
3511
Willy Tarreau87cf5142011-08-19 22:57:24 +02003512option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01003513 Enable insertion of the X-Forwarded-For header to requests sent to servers
3514 May be used in sections : defaults | frontend | listen | backend
3515 yes | yes | yes | yes
3516 Arguments :
3517 <network> is an optional argument used to disable this option for sources
3518 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02003519 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01003520 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01003521
3522 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
3523 their client address. This is sometimes annoying when the client's IP address
3524 is expected in server logs. To solve this problem, the well-known HTTP header
3525 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
3526 This header contains a value representing the client's IP address. Since this
3527 header is always appended at the end of the existing header list, the server
3528 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02003529 the server's manual to find how to enable use of this standard header. Note
3530 that only the last occurrence of the header must be used, since it is really
3531 possible that the client has already brought one.
3532
Willy Tarreaud72758d2010-01-12 10:42:19 +01003533 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02003534 the default "X-Forwarded-For". This can be useful where you might already
Willy Tarreaud72758d2010-01-12 10:42:19 +01003535 have a "X-Forwarded-For" header from a different application (eg: stunnel),
3536 and you need preserve it. Also if your backend server doesn't use the
Ross Westaf72a1d2008-08-03 10:51:45 +02003537 "X-Forwarded-For" header and requires different one (eg: Zeus Web Servers
3538 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01003539
3540 Sometimes, a same HAProxy instance may be shared between a direct client
3541 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
3542 used to decrypt HTTPS traffic). It is possible to disable the addition of the
3543 header for a known source address or network by adding the "except" keyword
3544 followed by the network address. In this case, any source IP matching the
3545 network will not cause an addition of this header. Most common uses are with
3546 private networks or 127.0.0.1.
3547
Willy Tarreau87cf5142011-08-19 22:57:24 +02003548 Alternatively, the keyword "if-none" states that the header will only be
3549 added if it is not present. This should only be used in perfectly trusted
3550 environment, as this might cause a security issue if headers reaching haproxy
3551 are under the control of the end-user.
3552
Willy Tarreauc27debf2008-01-06 08:57:02 +01003553 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02003554 least one of them uses it, the header will be added. Note that the backend's
3555 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02003556 both are defined. In the case of the "if-none" argument, if at least one of
3557 the frontend or the backend does not specify it, it wants the addition to be
3558 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01003559
Willy Tarreau87cf5142011-08-19 22:57:24 +02003560 It is important to note that by default, HAProxy works in tunnel mode and
3561 only inspects the first request of a connection, meaning that only the first
3562 request will have the header appended, which is certainly not what you want.
3563 In order to fix this, ensure that any of the "httpclose", "forceclose" or
3564 "http-server-close" options is set when using this option.
Willy Tarreauc27debf2008-01-06 08:57:02 +01003565
Ross Westaf72a1d2008-08-03 10:51:45 +02003566 Examples :
Willy Tarreauc27debf2008-01-06 08:57:02 +01003567 # Public HTTP address also used by stunnel on the same machine
3568 frontend www
3569 mode http
3570 option forwardfor except 127.0.0.1 # stunnel already adds the header
3571
Ross Westaf72a1d2008-08-03 10:51:45 +02003572 # Those servers want the IP Address in X-Client
3573 backend www
3574 mode http
3575 option forwardfor header X-Client
3576
Willy Tarreau87cf5142011-08-19 22:57:24 +02003577 See also : "option httpclose", "option http-server-close",
3578 "option forceclose"
Willy Tarreauc27debf2008-01-06 08:57:02 +01003579
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02003580
Willy Tarreau96e31212011-05-30 18:10:30 +02003581option http-no-delay
3582no option http-no-delay
3583 Instruct the system to favor low interactive delays over performance in HTTP
3584 May be used in sections : defaults | frontend | listen | backend
3585 yes | yes | yes | yes
3586 Arguments : none
3587
3588 In HTTP, each payload is unidirectional and has no notion of interactivity.
3589 Any agent is expected to queue data somewhat for a reasonably low delay.
3590 There are some very rare server-to-server applications that abuse the HTTP
3591 protocol and expect the payload phase to be highly interactive, with many
3592 interleaved data chunks in both directions within a single request. This is
3593 absolutely not supported by the HTTP specification and will not work across
3594 most proxies or servers. When such applications attempt to do this through
3595 haproxy, it works but they will experience high delays due to the network
3596 optimizations which favor performance by instructing the system to wait for
3597 enough data to be available in order to only send full packets. Typical
3598 delays are around 200 ms per round trip. Note that this only happens with
3599 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
3600 affected.
3601
3602 When "option http-no-delay" is present in either the frontend or the backend
3603 used by a connection, all such optimizations will be disabled in order to
3604 make the exchanges as fast as possible. Of course this offers no guarantee on
3605 the functionality, as it may break at any other place. But if it works via
3606 HAProxy, it will work as fast as possible. This option should never be used
3607 by default, and should never be used at all unless such a buggy application
3608 is discovered. The impact of using this option is an increase of bandwidth
3609 usage and CPU usage, which may significantly lower performance in high
3610 latency environments.
3611
3612
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02003613option http-pretend-keepalive
3614no option http-pretend-keepalive
3615 Define whether haproxy will announce keepalive to the server or not
3616 May be used in sections : defaults | frontend | listen | backend
3617 yes | yes | yes | yes
3618 Arguments : none
3619
3620 When running with "option http-server-close" or "option forceclose", haproxy
3621 adds a "Connection: close" header to the request forwarded to the server.
3622 Unfortunately, when some servers see this header, they automatically refrain
3623 from using the chunked encoding for responses of unknown length, while this
3624 is totally unrelated. The immediate effect is that this prevents haproxy from
3625 maintaining the client connection alive. A second effect is that a client or
3626 a cache could receive an incomplete response without being aware of it, and
3627 consider the response complete.
3628
3629 By setting "option http-pretend-keepalive", haproxy will make the server
3630 believe it will keep the connection alive. The server will then not fall back
3631 to the abnormal undesired above. When haproxy gets the whole response, it
3632 will close the connection with the server just as it would do with the
3633 "forceclose" option. That way the client gets a normal response and the
3634 connection is correctly closed on the server side.
3635
3636 It is recommended not to enable this option by default, because most servers
3637 will more efficiently close the connection themselves after the last packet,
3638 and release its buffers slightly earlier. Also, the added packet on the
3639 network could slightly reduce the overall peak performance. However it is
3640 worth noting that when this option is enabled, haproxy will have slightly
3641 less work to do. So if haproxy is the bottleneck on the whole architecture,
3642 enabling this option might save a few CPU cycles.
3643
3644 This option may be set both in a frontend and in a backend. It is enabled if
3645 at least one of the frontend or backend holding a connection has it enabled.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003646 This option may be combined with "option httpclose", which will cause
Willy Tarreau22a95342010-09-29 14:31:41 +02003647 keepalive to be announced to the server and close to be announced to the
3648 client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02003649
3650 If this option has been enabled in a "defaults" section, it can be disabled
3651 in a specific instance by prepending the "no" keyword before it.
3652
3653 See also : "option forceclose" and "option http-server-close"
3654
Willy Tarreauc27debf2008-01-06 08:57:02 +01003655
Willy Tarreaub608feb2010-01-02 22:47:18 +01003656option http-server-close
3657no option http-server-close
3658 Enable or disable HTTP connection closing on the server side
3659 May be used in sections : defaults | frontend | listen | backend
3660 yes | yes | yes | yes
3661 Arguments : none
3662
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02003663 By default, when a client communicates with a server, HAProxy will only
3664 analyze, log, and process the first request of each connection. Setting
3665 "option http-server-close" enables HTTP connection-close mode on the server
3666 side while keeping the ability to support HTTP keep-alive and pipelining on
3667 the client side. This provides the lowest latency on the client side (slow
3668 network) and the fastest session reuse on the server side to save server
3669 resources, similarly to "option forceclose". It also permits non-keepalive
3670 capable servers to be served in keep-alive mode to the clients if they
3671 conform to the requirements of RFC2616. Please note that some servers do not
3672 always conform to those requirements when they see "Connection: close" in the
3673 request. The effect will be that keep-alive will never be used. A workaround
3674 consists in enabling "option http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01003675
3676 At the moment, logs will not indicate whether requests came from the same
3677 session or not. The accept date reported in the logs corresponds to the end
3678 of the previous request, and the request time corresponds to the time spent
3679 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01003680 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
3681 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01003682
3683 This option may be set both in a frontend and in a backend. It is enabled if
3684 at least one of the frontend or backend holding a connection has it enabled.
Willy Tarreau0dfdf192010-01-05 11:33:11 +01003685 It is worth noting that "option forceclose" has precedence over "option
3686 http-server-close" and that combining "http-server-close" with "httpclose"
3687 basically achieve the same result as "forceclose".
Willy Tarreaub608feb2010-01-02 22:47:18 +01003688
3689 If this option has been enabled in a "defaults" section, it can be disabled
3690 in a specific instance by prepending the "no" keyword before it.
3691
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02003692 See also : "option forceclose", "option http-pretend-keepalive",
3693 "option httpclose" and "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01003694
3695
Willy Tarreau88d349d2010-01-25 12:15:43 +01003696option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01003697no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01003698 Make use of non-standard Proxy-Connection header instead of Connection
3699 May be used in sections : defaults | frontend | listen | backend
3700 yes | yes | yes | no
3701 Arguments : none
3702
3703 While RFC2616 explicitly states that HTTP/1.1 agents must use the
3704 Connection header to indicate their wish of persistent or non-persistent
3705 connections, both browsers and proxies ignore this header for proxied
3706 connections and make use of the undocumented, non-standard Proxy-Connection
3707 header instead. The issue begins when trying to put a load balancer between
3708 browsers and such proxies, because there will be a difference between what
3709 haproxy understands and what the client and the proxy agree on.
3710
3711 By setting this option in a frontend, haproxy can automatically switch to use
3712 that non-standard header if it sees proxied requests. A proxied request is
3713 defined here as one where the URI begins with neither a '/' nor a '*'. The
3714 choice of header only affects requests passing through proxies making use of
3715 one of the "httpclose", "forceclose" and "http-server-close" options. Note
3716 that this option can only be specified in a frontend and will affect the
3717 request along its whole life.
3718
Willy Tarreau844a7e72010-01-31 21:46:18 +01003719 Also, when this option is set, a request which requires authentication will
3720 automatically switch to use proxy authentication headers if it is itself a
3721 proxied request. That makes it possible to check or enforce authentication in
3722 front of an existing proxy.
3723
Willy Tarreau88d349d2010-01-25 12:15:43 +01003724 This option should normally never be used, except in front of a proxy.
3725
3726 See also : "option httpclose", "option forceclose" and "option
3727 http-server-close".
3728
3729
Willy Tarreaud63335a2010-02-26 12:56:52 +01003730option httpchk
3731option httpchk <uri>
3732option httpchk <method> <uri>
3733option httpchk <method> <uri> <version>
3734 Enable HTTP protocol to check on the servers health
3735 May be used in sections : defaults | frontend | listen | backend
3736 yes | no | yes | yes
3737 Arguments :
3738 <method> is the optional HTTP method used with the requests. When not set,
3739 the "OPTIONS" method is used, as it generally requires low server
3740 processing and is easy to filter out from the logs. Any method
3741 may be used, though it is not recommended to invent non-standard
3742 ones.
3743
3744 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
3745 which is accessible by default on almost any server, but may be
3746 changed to any other URI. Query strings are permitted.
3747
3748 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
3749 but some servers might behave incorrectly in HTTP 1.0, so turning
3750 it to HTTP/1.1 may sometimes help. Note that the Host field is
3751 mandatory in HTTP/1.1, and as a trick, it is possible to pass it
3752 after "\r\n" following the version string.
3753
3754 By default, server health checks only consist in trying to establish a TCP
3755 connection. When "option httpchk" is specified, a complete HTTP request is
3756 sent once the TCP connection is established, and responses 2xx and 3xx are
3757 considered valid, while all other ones indicate a server failure, including
3758 the lack of any response.
3759
3760 The port and interval are specified in the server configuration.
3761
3762 This option does not necessarily require an HTTP backend, it also works with
3763 plain TCP backends. This is particularly useful to check simple scripts bound
3764 to some dedicated ports using the inetd daemon.
3765
3766 Examples :
3767 # Relay HTTPS traffic to Apache instance and check service availability
3768 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
3769 backend https_relay
3770 mode tcp
3771 option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www
3772 server apache1 192.168.1.1:443 check port 80
3773
Simon Hormana2b9dad2013-02-12 10:45:54 +09003774 See also : "option lb-agent-chk", "option ssl-hello-chk", "option smtpchk",
3775 "option mysql-check", "option pgsql-check", "http-check" and
3776 the "check", "port" and "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01003777
3778
Willy Tarreauc27debf2008-01-06 08:57:02 +01003779option httpclose
3780no option httpclose
3781 Enable or disable passive HTTP connection closing
3782 May be used in sections : defaults | frontend | listen | backend
3783 yes | yes | yes | yes
3784 Arguments : none
3785
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02003786 By default, when a client communicates with a server, HAProxy will only
3787 analyze, log, and process the first request of each connection. If "option
3788 httpclose" is set, it will check if a "Connection: close" header is already
3789 set in each direction, and will add one if missing. Each end should react to
3790 this by actively closing the TCP connection after each transfer, thus
3791 resulting in a switch to the HTTP close mode. Any "Connection" header
3792 different from "close" will also be removed.
Willy Tarreauc27debf2008-01-06 08:57:02 +01003793
3794 It seldom happens that some servers incorrectly ignore this header and do not
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003795 close the connection even though they reply "Connection: close". For this
Willy Tarreau0dfdf192010-01-05 11:33:11 +01003796 reason, they are not compatible with older HTTP 1.0 browsers. If this happens
3797 it is possible to use the "option forceclose" which actively closes the
3798 request connection once the server responds. Option "forceclose" also
3799 releases the server connection earlier because it does not have to wait for
3800 the client to acknowledge it.
Willy Tarreauc27debf2008-01-06 08:57:02 +01003801
3802 This option may be set both in a frontend and in a backend. It is enabled if
3803 at least one of the frontend or backend holding a connection has it enabled.
3804 If "option forceclose" is specified too, it has precedence over "httpclose".
Willy Tarreau0dfdf192010-01-05 11:33:11 +01003805 If "option http-server-close" is enabled at the same time as "httpclose", it
3806 basically achieves the same result as "option forceclose".
Willy Tarreauc27debf2008-01-06 08:57:02 +01003807
3808 If this option has been enabled in a "defaults" section, it can be disabled
3809 in a specific instance by prepending the "no" keyword before it.
3810
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02003811 See also : "option forceclose", "option http-server-close" and
3812 "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01003813
3814
Emeric Brun3a058f32009-06-30 18:26:00 +02003815option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01003816 Enable logging of HTTP request, session state and timers
3817 May be used in sections : defaults | frontend | listen | backend
3818 yes | yes | yes | yes
Emeric Brun3a058f32009-06-30 18:26:00 +02003819 Arguments :
3820 clf if the "clf" argument is added, then the output format will be
3821 the CLF format instead of HAProxy's default HTTP format. You can
3822 use this when you need to feed HAProxy's logs through a specific
3823 log analyser which only support the CLF format and which is not
3824 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01003825
3826 By default, the log output format is very poor, as it only contains the
3827 source and destination addresses, and the instance name. By specifying
3828 "option httplog", each log line turns into a much richer format including,
3829 but not limited to, the HTTP request, the connection timers, the session
3830 status, the connections numbers, the captured headers and cookies, the
3831 frontend, backend and server name, and of course the source address and
3832 ports.
3833
3834 This option may be set either in the frontend or the backend.
3835
Emeric Brun3a058f32009-06-30 18:26:00 +02003836 If this option has been enabled in a "defaults" section, it can be disabled
3837 in a specific instance by prepending the "no" keyword before it. Specifying
3838 only "option httplog" will automatically clear the 'clf' mode if it was set
3839 by default.
3840
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003841 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01003842
Willy Tarreau55165fe2009-05-10 12:02:55 +02003843
3844option http_proxy
3845no option http_proxy
3846 Enable or disable plain HTTP proxy mode
3847 May be used in sections : defaults | frontend | listen | backend
3848 yes | yes | yes | yes
3849 Arguments : none
3850
3851 It sometimes happens that people need a pure HTTP proxy which understands
3852 basic proxy requests without caching nor any fancy feature. In this case,
3853 it may be worth setting up an HAProxy instance with the "option http_proxy"
3854 set. In this mode, no server is declared, and the connection is forwarded to
3855 the IP address and port found in the URL after the "http://" scheme.
3856
3857 No host address resolution is performed, so this only works when pure IP
3858 addresses are passed. Since this option's usage perimeter is rather limited,
3859 it will probably be used only by experts who know they need exactly it. Last,
3860 if the clients are susceptible of sending keep-alive requests, it will be
Cyril Bonté2409e682010-12-14 22:47:51 +01003861 needed to add "option httpclose" to ensure that all requests will correctly
Willy Tarreau55165fe2009-05-10 12:02:55 +02003862 be analyzed.
3863
3864 If this option has been enabled in a "defaults" section, it can be disabled
3865 in a specific instance by prepending the "no" keyword before it.
3866
3867 Example :
3868 # this backend understands HTTP proxy requests and forwards them directly.
3869 backend direct_forward
3870 option httpclose
3871 option http_proxy
3872
3873 See also : "option httpclose"
3874
Willy Tarreau211ad242009-10-03 21:45:07 +02003875
Jamie Gloudon801a0a32012-08-25 00:18:33 -04003876option independent-streams
3877no option independent-streams
3878 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02003879 May be used in sections : defaults | frontend | listen | backend
3880 yes | yes | yes | yes
3881 Arguments : none
3882
3883 By default, when data is sent over a socket, both the write timeout and the
3884 read timeout for that socket are refreshed, because we consider that there is
3885 activity on that socket, and we have no other means of guessing if we should
3886 receive data or not.
3887
3888 While this default behaviour is desirable for almost all applications, there
3889 exists a situation where it is desirable to disable it, and only refresh the
3890 read timeout if there are incoming data. This happens on sessions with large
3891 timeouts and low amounts of exchanged data such as telnet session. If the
3892 server suddenly disappears, the output data accumulates in the system's
3893 socket buffers, both timeouts are correctly refreshed, and there is no way
3894 to know the server does not receive them, so we don't timeout. However, when
3895 the underlying protocol always echoes sent data, it would be enough by itself
3896 to detect the issue using the read timeout. Note that this problem does not
3897 happen with more verbose protocols because data won't accumulate long in the
3898 socket buffers.
3899
3900 When this option is set on the frontend, it will disable read timeout updates
3901 on data sent to the client. There probably is little use of this case. When
3902 the option is set on the backend, it will disable read timeout updates on
3903 data sent to the server. Doing so will typically break large HTTP posts from
3904 slow lines, so use it with caution.
3905
Jamie Gloudon801a0a32012-08-25 00:18:33 -04003906 Note: older versions used to call this setting "option independant-streams"
3907 with a spelling mistake. This spelling is still supported but
3908 deprecated.
3909
Willy Tarreauce887fd2012-05-12 12:50:00 +02003910 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02003911
3912
Simon Hormana2b9dad2013-02-12 10:45:54 +09003913option lb-agent-chk
3914 Use a TCP connection to obtain a metric of server health
3915 May be used in sections : defaults | frontend | listen | backend
3916 yes | no | yes | yes
3917 Arguments : none
3918
3919 This alters health checking behaviour by connecting making a TCP
3920 connection and reading an ASCII string. The string should have one of
3921 the following forms:
3922
3923 * An ASCII representation of an positive integer percentage.
3924 e.g. "75%"
3925
3926 Values in this format will set the weight proportional to the initial
3927 weight of a server as configured when haproxy starts.
3928
3929 * The string "drain".
3930
3931 This will cause the weight of a server to be set to 0, and thus it will
3932 not accept any new connections other than those that are accepted via
3933 persistence.
3934
3935 * The string "down", optionally followed by a description string.
3936
3937 Mark the server as down and log the description string as the reason.
3938
3939 * The string "stopped", optionally followed by a description string.
3940
3941 This currently has the same behaviour as down (iii).
3942
3943 * The string "fail", optionally followed by a description string.
3944
3945 This currently has the same behaviour as down (iii).
3946
3947 The use of an alternate check-port, used to obtain agent heath check
3948 information described above as opposed to the port of the service, may be
3949 useful in conjunction with this option.
3950
3951
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02003952option ldap-check
3953 Use LDAPv3 health checks for server testing
3954 May be used in sections : defaults | frontend | listen | backend
3955 yes | no | yes | yes
3956 Arguments : none
3957
3958 It is possible to test that the server correctly talks LDAPv3 instead of just
3959 testing that it accepts the TCP connection. When this option is set, an
3960 LDAPv3 anonymous simple bind message is sent to the server, and the response
3961 is analyzed to find an LDAPv3 bind response message.
3962
3963 The server is considered valid only when the LDAP response contains success
3964 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
3965
3966 Logging of bind requests is server dependent see your documentation how to
3967 configure it.
3968
3969 Example :
3970 option ldap-check
3971
3972 See also : "option httpchk"
3973
3974
Willy Tarreau211ad242009-10-03 21:45:07 +02003975option log-health-checks
3976no option log-health-checks
3977 Enable or disable logging of health checks
3978 May be used in sections : defaults | frontend | listen | backend
3979 yes | no | yes | yes
3980 Arguments : none
3981
3982 Enable health checks logging so it possible to check for example what
3983 was happening before a server crash. Failed health check are logged if
3984 server is UP and succeeded health checks if server is DOWN, so the amount
3985 of additional information is limited.
3986
3987 If health check logging is enabled no health check status is printed
3988 when servers is set up UP/DOWN/ENABLED/DISABLED.
3989
3990 See also: "log" and section 8 about logging.
3991
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02003992
3993option log-separate-errors
3994no option log-separate-errors
3995 Change log level for non-completely successful connections
3996 May be used in sections : defaults | frontend | listen | backend
3997 yes | yes | yes | no
3998 Arguments : none
3999
4000 Sometimes looking for errors in logs is not easy. This option makes haproxy
4001 raise the level of logs containing potentially interesting information such
4002 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
4003 level changes from "info" to "err". This makes it possible to log them
4004 separately to a different file with most syslog daemons. Be careful not to
4005 remove them from the original file, otherwise you would lose ordering which
4006 provides very important information.
4007
4008 Using this option, large sites dealing with several thousand connections per
4009 second may log normal traffic to a rotating buffer and only archive smaller
4010 error logs.
4011
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004012 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02004013 logging.
4014
Willy Tarreauc27debf2008-01-06 08:57:02 +01004015
4016option logasap
4017no option logasap
4018 Enable or disable early logging of HTTP requests
4019 May be used in sections : defaults | frontend | listen | backend
4020 yes | yes | yes | no
4021 Arguments : none
4022
4023 By default, HTTP requests are logged upon termination so that the total
4024 transfer time and the number of bytes appear in the logs. When large objects
4025 are being transferred, it may take a while before the request appears in the
4026 logs. Using "option logasap", the request gets logged as soon as the server
4027 sends the complete headers. The only missing information in the logs will be
4028 the total number of bytes which will indicate everything except the amount
4029 of data transferred, and the total time which will not take the transfer
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01004030 time into account. In such a situation, it's a good practice to capture the
Willy Tarreauc27debf2008-01-06 08:57:02 +01004031 "Content-Length" response header so that the logs at least indicate how many
4032 bytes are expected to be transferred.
4033
Willy Tarreaucc6c8912009-02-22 10:53:55 +01004034 Examples :
4035 listen http_proxy 0.0.0.0:80
4036 mode http
4037 option httplog
4038 option logasap
4039 log 192.168.2.200 local3
4040
4041 >>> Feb 6 12:14:14 localhost \
4042 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
4043 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
4044 "GET /image.iso HTTP/1.0"
4045
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004046 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01004047 logging.
4048
4049
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02004050option mysql-check [ user <username> ]
4051 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01004052 May be used in sections : defaults | frontend | listen | backend
4053 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02004054 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004055 <username> This is the username which will be used when connecting to MySQL
4056 server.
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02004057
4058 If you specify a username, the check consists of sending two MySQL packet,
4059 one Client Authentication packet, and one QUIT packet, to correctly close
4060 MySQL session. We then parse the MySQL Handshake Initialisation packet and/or
4061 Error packet. It is a basic but useful test which does not produce error nor
4062 aborted connect on the server. However, it requires adding an authorization
4063 in the MySQL table, like this :
4064
4065 USE mysql;
4066 INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');
4067 FLUSH PRIVILEGES;
4068
4069 If you don't specify a username (it is deprecated and not recommended), the
4070 check only consists in parsing the Mysql Handshake Initialisation packet or
4071 Error packet, we don't send anything in this mode. It was reported that it
4072 can generate lockout if check is too frequent and/or if there is not enough
4073 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
4074 value as if a connection is established successfully within fewer than MySQL
4075 "max_connect_errors" attempts after a previous connection was interrupted,
4076 the error count for the host is cleared to zero. If HAProxy's server get
4077 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
4078
4079 Remember that this does not check database presence nor database consistency.
4080 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01004081
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02004082 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01004083
4084 Most often, an incoming MySQL server needs to see the client's IP address for
4085 various purposes, including IP privilege matching and connection logging.
4086 When possible, it is often wise to masquerade the client's IP address when
4087 connecting to the server using the "usesrc" argument of the "source" keyword,
4088 which requires the cttproxy feature to be compiled in, and the MySQL server
4089 to route the client via the machine hosting haproxy.
4090
4091 See also: "option httpchk"
4092
Rauf Kuliyev38b41562011-01-04 15:14:13 +01004093option pgsql-check [ user <username> ]
4094 Use PostgreSQL health checks for server testing
4095 May be used in sections : defaults | frontend | listen | backend
4096 yes | no | yes | yes
4097 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004098 <username> This is the username which will be used when connecting to
4099 PostgreSQL server.
Rauf Kuliyev38b41562011-01-04 15:14:13 +01004100
4101 The check sends a PostgreSQL StartupMessage and waits for either
4102 Authentication request or ErrorResponse message. It is a basic but useful
4103 test which does not produce error nor aborted connect on the server.
4104 This check is identical with the "mysql-check".
4105
4106 See also: "option httpchk"
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01004107
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004108option nolinger
4109no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004110 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004111 May be used in sections: defaults | frontend | listen | backend
4112 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01004113 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004114
4115 When clients or servers abort connections in a dirty way (eg: they are
4116 physically disconnected), the session timeouts triggers and the session is
4117 closed. But it will remain in FIN_WAIT1 state for some time in the system,
4118 using some resources and possibly limiting the ability to establish newer
4119 connections.
4120
4121 When this happens, it is possible to activate "option nolinger" which forces
4122 the system to immediately remove any socket's pending data on close. Thus,
4123 the session is instantly purged from the system's tables. This usually has
4124 side effects such as increased number of TCP resets due to old retransmits
4125 getting immediately rejected. Some firewalls may sometimes complain about
4126 this too.
4127
4128 For this reason, it is not recommended to use this option when not absolutely
4129 needed. You know that you need it when you have thousands of FIN_WAIT1
4130 sessions on your system (TIME_WAIT ones do not count).
4131
4132 This option may be used both on frontends and backends, depending on the side
4133 where it is required. Use it on the frontend for clients, and on the backend
4134 for servers.
4135
4136 If this option has been enabled in a "defaults" section, it can be disabled
4137 in a specific instance by prepending the "no" keyword before it.
4138
4139
Willy Tarreau55165fe2009-05-10 12:02:55 +02004140option originalto [ except <network> ] [ header <name> ]
4141 Enable insertion of the X-Original-To header to requests sent to servers
4142 May be used in sections : defaults | frontend | listen | backend
4143 yes | yes | yes | yes
4144 Arguments :
4145 <network> is an optional argument used to disable this option for sources
4146 matching <network>
4147 <name> an optional argument to specify a different "X-Original-To"
4148 header name.
4149
4150 Since HAProxy can work in transparent mode, every request from a client can
4151 be redirected to the proxy and HAProxy itself can proxy every request to a
4152 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
4153 be lost. This is annoying when you want access rules based on destination ip
4154 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
4155 added by HAProxy to all requests sent to the server. This header contains a
4156 value representing the original destination IP address. Since this must be
4157 configured to always use the last occurrence of this header only. Note that
4158 only the last occurrence of the header must be used, since it is really
4159 possible that the client has already brought one.
4160
4161 The keyword "header" may be used to supply a different header name to replace
4162 the default "X-Original-To". This can be useful where you might already
4163 have a "X-Original-To" header from a different application, and you need
4164 preserve it. Also if your backend server doesn't use the "X-Original-To"
4165 header and requires different one.
4166
4167 Sometimes, a same HAProxy instance may be shared between a direct client
4168 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
4169 used to decrypt HTTPS traffic). It is possible to disable the addition of the
4170 header for a known source address or network by adding the "except" keyword
4171 followed by the network address. In this case, any source IP matching the
4172 network will not cause an addition of this header. Most common uses are with
4173 private networks or 127.0.0.1.
4174
4175 This option may be specified either in the frontend or in the backend. If at
4176 least one of them uses it, the header will be added. Note that the backend's
4177 setting of the header subargument takes precedence over the frontend's if
4178 both are defined.
4179
Willy Tarreau87cf5142011-08-19 22:57:24 +02004180 It is important to note that by default, HAProxy works in tunnel mode and
4181 only inspects the first request of a connection, meaning that only the first
4182 request will have the header appended, which is certainly not what you want.
4183 In order to fix this, ensure that any of the "httpclose", "forceclose" or
4184 "http-server-close" options is set when using this option.
Willy Tarreau55165fe2009-05-10 12:02:55 +02004185
4186 Examples :
4187 # Original Destination address
4188 frontend www
4189 mode http
4190 option originalto except 127.0.0.1
4191
4192 # Those servers want the IP Address in X-Client-Dst
4193 backend www
4194 mode http
4195 option originalto header X-Client-Dst
4196
Willy Tarreau87cf5142011-08-19 22:57:24 +02004197 See also : "option httpclose", "option http-server-close",
4198 "option forceclose"
Willy Tarreau55165fe2009-05-10 12:02:55 +02004199
4200
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004201option persist
4202no option persist
4203 Enable or disable forced persistence on down servers
4204 May be used in sections: defaults | frontend | listen | backend
4205 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01004206 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004207
4208 When an HTTP request reaches a backend with a cookie which references a dead
4209 server, by default it is redispatched to another server. It is possible to
4210 force the request to be sent to the dead server first using "option persist"
4211 if absolutely needed. A common use case is when servers are under extreme
4212 load and spend their time flapping. In this case, the users would still be
4213 directed to the server they opened the session on, in the hope they would be
4214 correctly served. It is recommended to use "option redispatch" in conjunction
4215 with this option so that in the event it would not be possible to connect to
4216 the server at all (server definitely dead), the client would finally be
4217 redirected to another valid server.
4218
4219 If this option has been enabled in a "defaults" section, it can be disabled
4220 in a specific instance by prepending the "no" keyword before it.
4221
Willy Tarreau4de91492010-01-22 19:10:05 +01004222 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004223
4224
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004225option redispatch
4226no option redispatch
4227 Enable or disable session redistribution in case of connection failure
4228 May be used in sections: defaults | frontend | listen | backend
4229 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01004230 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004231
4232 In HTTP mode, if a server designated by a cookie is down, clients may
4233 definitely stick to it because they cannot flush the cookie, so they will not
4234 be able to access the service anymore.
4235
4236 Specifying "option redispatch" will allow the proxy to break their
4237 persistence and redistribute them to a working server.
4238
4239 It also allows to retry last connection to another server in case of multiple
4240 connection failures. Of course, it requires having "retries" set to a nonzero
4241 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01004242
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004243 This form is the preferred form, which replaces both the "redispatch" and
4244 "redisp" keywords.
4245
4246 If this option has been enabled in a "defaults" section, it can be disabled
4247 in a specific instance by prepending the "no" keyword before it.
4248
Willy Tarreau4de91492010-01-22 19:10:05 +01004249 See also : "redispatch", "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004250
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004251
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02004252option redis-check
4253 Use redis health checks for server testing
4254 May be used in sections : defaults | frontend | listen | backend
4255 yes | no | yes | yes
4256 Arguments : none
4257
4258 It is possible to test that the server correctly talks REDIS protocol instead
4259 of just testing that it accepts the TCP connection. When this option is set,
4260 a PING redis command is sent to the server, and the response is analyzed to
4261 find the "+PONG" response message.
4262
4263 Example :
4264 option redis-check
4265
4266 See also : "option httpchk"
4267
4268
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004269option smtpchk
4270option smtpchk <hello> <domain>
4271 Use SMTP health checks for server testing
4272 May be used in sections : defaults | frontend | listen | backend
4273 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01004274 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004275 <hello> is an optional argument. It is the "hello" command to use. It can
4276 be either "HELO" (for SMTP) or "EHLO" (for ESTMP). All other
4277 values will be turned into the default command ("HELO").
4278
4279 <domain> is the domain name to present to the server. It may only be
4280 specified (and is mandatory) if the hello command has been
4281 specified. By default, "localhost" is used.
4282
4283 When "option smtpchk" is set, the health checks will consist in TCP
4284 connections followed by an SMTP command. By default, this command is
4285 "HELO localhost". The server's return code is analyzed and only return codes
4286 starting with a "2" will be considered as valid. All other responses,
4287 including a lack of response will constitute an error and will indicate a
4288 dead server.
4289
4290 This test is meant to be used with SMTP servers or relays. Depending on the
4291 request, it is possible that some servers do not log each connection attempt,
4292 so you may want to experiment to improve the behaviour. Using telnet on port
4293 25 is often easier than adjusting the configuration.
4294
4295 Most often, an incoming SMTP server needs to see the client's IP address for
4296 various purposes, including spam filtering, anti-spoofing and logging. When
4297 possible, it is often wise to masquerade the client's IP address when
4298 connecting to the server using the "usesrc" argument of the "source" keyword,
4299 which requires the cttproxy feature to be compiled in.
4300
4301 Example :
4302 option smtpchk HELO mydomain.org
4303
4304 See also : "option httpchk", "source"
4305
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004306
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02004307option socket-stats
4308no option socket-stats
4309
4310 Enable or disable collecting & providing separate statistics for each socket.
4311 May be used in sections : defaults | frontend | listen | backend
4312 yes | yes | yes | no
4313
4314 Arguments : none
4315
4316
Willy Tarreauff4f82d2009-02-06 11:28:13 +01004317option splice-auto
4318no option splice-auto
4319 Enable or disable automatic kernel acceleration on sockets in both directions
4320 May be used in sections : defaults | frontend | listen | backend
4321 yes | yes | yes | yes
4322 Arguments : none
4323
4324 When this option is enabled either on a frontend or on a backend, haproxy
4325 will automatically evaluate the opportunity to use kernel tcp splicing to
4326 forward data between the client and the server, in either direction. Haproxy
4327 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004328 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01004329 are not much aggressive in order to limit excessive use of splicing. This
4330 option requires splicing to be enabled at compile time, and may be globally
4331 disabled with the global option "nosplice". Since splice uses pipes, using it
4332 requires that there are enough spare pipes.
4333
4334 Important note: kernel-based TCP splicing is a Linux-specific feature which
4335 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
4336 transfer data between sockets without copying these data to user-space, thus
4337 providing noticeable performance gains and CPU cycles savings. Since many
4338 early implementations are buggy, corrupt data and/or are inefficient, this
4339 feature is not enabled by default, and it should be used with extreme care.
4340 While it is not possible to detect the correctness of an implementation,
4341 2.6.29 is the first version offering a properly working implementation. In
4342 case of doubt, splicing may be globally disabled using the global "nosplice"
4343 keyword.
4344
4345 Example :
4346 option splice-auto
4347
4348 If this option has been enabled in a "defaults" section, it can be disabled
4349 in a specific instance by prepending the "no" keyword before it.
4350
4351 See also : "option splice-request", "option splice-response", and global
4352 options "nosplice" and "maxpipes"
4353
4354
4355option splice-request
4356no option splice-request
4357 Enable or disable automatic kernel acceleration on sockets for requests
4358 May be used in sections : defaults | frontend | listen | backend
4359 yes | yes | yes | yes
4360 Arguments : none
4361
4362 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004363 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01004364 the client to the server. It might still use the recv/send scheme if there
4365 are no spare pipes left. This option requires splicing to be enabled at
4366 compile time, and may be globally disabled with the global option "nosplice".
4367 Since splice uses pipes, using it requires that there are enough spare pipes.
4368
4369 Important note: see "option splice-auto" for usage limitations.
4370
4371 Example :
4372 option splice-request
4373
4374 If this option has been enabled in a "defaults" section, it can be disabled
4375 in a specific instance by prepending the "no" keyword before it.
4376
4377 See also : "option splice-auto", "option splice-response", and global options
4378 "nosplice" and "maxpipes"
4379
4380
4381option splice-response
4382no option splice-response
4383 Enable or disable automatic kernel acceleration on sockets for responses
4384 May be used in sections : defaults | frontend | listen | backend
4385 yes | yes | yes | yes
4386 Arguments : none
4387
4388 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004389 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01004390 the server to the client. It might still use the recv/send scheme if there
4391 are no spare pipes left. This option requires splicing to be enabled at
4392 compile time, and may be globally disabled with the global option "nosplice".
4393 Since splice uses pipes, using it requires that there are enough spare pipes.
4394
4395 Important note: see "option splice-auto" for usage limitations.
4396
4397 Example :
4398 option splice-response
4399
4400 If this option has been enabled in a "defaults" section, it can be disabled
4401 in a specific instance by prepending the "no" keyword before it.
4402
4403 See also : "option splice-auto", "option splice-request", and global options
4404 "nosplice" and "maxpipes"
4405
4406
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004407option srvtcpka
4408no option srvtcpka
4409 Enable or disable the sending of TCP keepalive packets on the server side
4410 May be used in sections : defaults | frontend | listen | backend
4411 yes | no | yes | yes
4412 Arguments : none
4413
4414 When there is a firewall or any session-aware component between a client and
4415 a server, and when the protocol involves very long sessions with long idle
4416 periods (eg: remote desktops), there is a risk that one of the intermediate
4417 components decides to expire a session which has remained idle for too long.
4418
4419 Enabling socket-level TCP keep-alives makes the system regularly send packets
4420 to the other end of the connection, leaving it active. The delay between
4421 keep-alive probes is controlled by the system only and depends both on the
4422 operating system and its tuning parameters.
4423
4424 It is important to understand that keep-alive packets are neither emitted nor
4425 received at the application level. It is only the network stacks which sees
4426 them. For this reason, even if one side of the proxy already uses keep-alives
4427 to maintain its connection alive, those keep-alive packets will not be
4428 forwarded to the other side of the proxy.
4429
4430 Please note that this has nothing to do with HTTP keep-alive.
4431
4432 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
4433 server side of a connection, which should help when session expirations are
4434 noticed between HAProxy and a server.
4435
4436 If this option has been enabled in a "defaults" section, it can be disabled
4437 in a specific instance by prepending the "no" keyword before it.
4438
4439 See also : "option clitcpka", "option tcpka"
4440
4441
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004442option ssl-hello-chk
4443 Use SSLv3 client hello health checks for server testing
4444 May be used in sections : defaults | frontend | listen | backend
4445 yes | no | yes | yes
4446 Arguments : none
4447
4448 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
4449 possible to test that the server correctly talks SSL instead of just testing
4450 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
4451 SSLv3 client hello messages are sent once the connection is established to
4452 the server, and the response is analyzed to find an SSL server hello message.
4453 The server is considered valid only when the response contains this server
4454 hello message.
4455
4456 All servers tested till there correctly reply to SSLv3 client hello messages,
4457 and most servers tested do not even log the requests containing only hello
4458 messages, which is appreciable.
4459
Willy Tarreau763a95b2012-10-04 23:15:39 +02004460 Note that this check works even when SSL support was not built into haproxy
4461 because it forges the SSL message. When SSL support is available, it is best
4462 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004463
Willy Tarreau763a95b2012-10-04 23:15:39 +02004464 See also: "option httpchk", "check-ssl"
4465
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004466
Willy Tarreau9ea05a72009-06-14 12:07:01 +02004467option tcp-smart-accept
4468no option tcp-smart-accept
4469 Enable or disable the saving of one ACK packet during the accept sequence
4470 May be used in sections : defaults | frontend | listen | backend
4471 yes | yes | yes | no
4472 Arguments : none
4473
4474 When an HTTP connection request comes in, the system acknowledges it on
4475 behalf of HAProxy, then the client immediately sends its request, and the
4476 system acknowledges it too while it is notifying HAProxy about the new
4477 connection. HAProxy then reads the request and responds. This means that we
4478 have one TCP ACK sent by the system for nothing, because the request could
4479 very well be acknowledged by HAProxy when it sends its response.
4480
4481 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
4482 sending this useless ACK on platforms which support it (currently at least
4483 Linux). It must not cause any problem, because the system will send it anyway
4484 after 40 ms if the response takes more time than expected to come.
4485
4486 During complex network debugging sessions, it may be desirable to disable
4487 this optimization because delayed ACKs can make troubleshooting more complex
4488 when trying to identify where packets are delayed. It is then possible to
4489 fall back to normal behaviour by specifying "no option tcp-smart-accept".
4490
4491 It is also possible to force it for non-HTTP proxies by simply specifying
4492 "option tcp-smart-accept". For instance, it can make sense with some services
4493 such as SMTP where the server speaks first.
4494
4495 It is recommended to avoid forcing this option in a defaults section. In case
4496 of doubt, consider setting it back to automatic values by prepending the
4497 "default" keyword before it, or disabling it using the "no" keyword.
4498
Willy Tarreaud88edf22009-06-14 15:48:17 +02004499 See also : "option tcp-smart-connect"
4500
4501
4502option tcp-smart-connect
4503no option tcp-smart-connect
4504 Enable or disable the saving of one ACK packet during the connect sequence
4505 May be used in sections : defaults | frontend | listen | backend
4506 yes | no | yes | yes
4507 Arguments : none
4508
4509 On certain systems (at least Linux), HAProxy can ask the kernel not to
4510 immediately send an empty ACK upon a connection request, but to directly
4511 send the buffer request instead. This saves one packet on the network and
4512 thus boosts performance. It can also be useful for some servers, because they
4513 immediately get the request along with the incoming connection.
4514
4515 This feature is enabled when "option tcp-smart-connect" is set in a backend.
4516 It is not enabled by default because it makes network troubleshooting more
4517 complex.
4518
4519 It only makes sense to enable it with protocols where the client speaks first
4520 such as HTTP. In other situations, if there is no data to send in place of
4521 the ACK, a normal ACK is sent.
4522
4523 If this option has been enabled in a "defaults" section, it can be disabled
4524 in a specific instance by prepending the "no" keyword before it.
4525
4526 See also : "option tcp-smart-accept"
4527
Willy Tarreau9ea05a72009-06-14 12:07:01 +02004528
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004529option tcpka
4530 Enable or disable the sending of TCP keepalive packets on both sides
4531 May be used in sections : defaults | frontend | listen | backend
4532 yes | yes | yes | yes
4533 Arguments : none
4534
4535 When there is a firewall or any session-aware component between a client and
4536 a server, and when the protocol involves very long sessions with long idle
4537 periods (eg: remote desktops), there is a risk that one of the intermediate
4538 components decides to expire a session which has remained idle for too long.
4539
4540 Enabling socket-level TCP keep-alives makes the system regularly send packets
4541 to the other end of the connection, leaving it active. The delay between
4542 keep-alive probes is controlled by the system only and depends both on the
4543 operating system and its tuning parameters.
4544
4545 It is important to understand that keep-alive packets are neither emitted nor
4546 received at the application level. It is only the network stacks which sees
4547 them. For this reason, even if one side of the proxy already uses keep-alives
4548 to maintain its connection alive, those keep-alive packets will not be
4549 forwarded to the other side of the proxy.
4550
4551 Please note that this has nothing to do with HTTP keep-alive.
4552
4553 Using option "tcpka" enables the emission of TCP keep-alive probes on both
4554 the client and server sides of a connection. Note that this is meaningful
4555 only in "defaults" or "listen" sections. If this option is used in a
4556 frontend, only the client side will get keep-alives, and if this option is
4557 used in a backend, only the server side will get keep-alives. For this
4558 reason, it is strongly recommended to explicitly use "option clitcpka" and
4559 "option srvtcpka" when the configuration is split between frontends and
4560 backends.
4561
4562 See also : "option clitcpka", "option srvtcpka"
4563
Willy Tarreau844e3c52008-01-11 16:28:18 +01004564
4565option tcplog
4566 Enable advanced logging of TCP connections with session state and timers
4567 May be used in sections : defaults | frontend | listen | backend
4568 yes | yes | yes | yes
4569 Arguments : none
4570
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 tcplog", each log line turns into a much richer format including, but
4574 not limited to, the connection timers, the session status, the connections
4575 numbers, the frontend, backend and server name, and of course the source
4576 address and ports. This option is useful for pure TCP proxies in order to
4577 find which of the client or server disconnects or times out. For normal HTTP
4578 proxies, it's better to use "option httplog" which is even more complete.
4579
4580 This option may be set either in the frontend or the backend.
4581
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004582 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01004583
4584
Willy Tarreau844e3c52008-01-11 16:28:18 +01004585option transparent
4586no option transparent
4587 Enable client-side transparent proxying
4588 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01004589 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01004590 Arguments : none
4591
4592 This option was introduced in order to provide layer 7 persistence to layer 3
4593 load balancers. The idea is to use the OS's ability to redirect an incoming
4594 connection for a remote address to a local process (here HAProxy), and let
4595 this process know what address was initially requested. When this option is
4596 used, sessions without cookies will be forwarded to the original destination
4597 IP address of the incoming request (which should match that of another
4598 equipment), while requests with cookies will still be forwarded to the
4599 appropriate server.
4600
4601 Note that contrary to a common belief, this option does NOT make HAProxy
4602 present the client's IP to the server when establishing the connection.
4603
Willy Tarreaua1146052011-03-01 09:51:54 +01004604 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01004605 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01004606
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004607
Emeric Brun647caf12009-06-30 17:57:00 +02004608persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02004609persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02004610 Enable RDP cookie-based persistence
4611 May be used in sections : defaults | frontend | listen | backend
4612 yes | no | yes | yes
4613 Arguments :
4614 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02004615 default cookie name "msts" will be used. There currently is no
4616 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02004617
4618 This statement enables persistence based on an RDP cookie. The RDP cookie
4619 contains all information required to find the server in the list of known
4620 servers. So when this option is set in the backend, the request is analysed
4621 and if an RDP cookie is found, it is decoded. If it matches a known server
4622 which is still UP (or if "option persist" is set), then the connection is
4623 forwarded to this server.
4624
4625 Note that this only makes sense in a TCP backend, but for this to work, the
4626 frontend must have waited long enough to ensure that an RDP cookie is present
4627 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004628 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02004629 a single "listen" section.
4630
Willy Tarreau61e28f22010-05-16 22:31:05 +02004631 Also, it is important to understand that the terminal server will emit this
4632 RDP cookie only if it is configured for "token redirection mode", which means
4633 that the "IP address redirection" option is disabled.
4634
Emeric Brun647caf12009-06-30 17:57:00 +02004635 Example :
4636 listen tse-farm
4637 bind :3389
4638 # wait up to 5s for an RDP cookie in the request
4639 tcp-request inspect-delay 5s
4640 tcp-request content accept if RDP_COOKIE
4641 # apply RDP cookie persistence
4642 persist rdp-cookie
4643 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02004644 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02004645 balance rdp-cookie
4646 server srv1 1.1.1.1:3389
4647 server srv2 1.1.1.2:3389
4648
Simon Hormanab814e02011-06-24 14:50:20 +09004649 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
4650 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02004651
4652
Willy Tarreau3a7d2072009-03-05 23:48:25 +01004653rate-limit sessions <rate>
4654 Set a limit on the number of new sessions accepted per second on a frontend
4655 May be used in sections : defaults | frontend | listen | backend
4656 yes | yes | yes | no
4657 Arguments :
4658 <rate> The <rate> parameter is an integer designating the maximum number
4659 of new sessions per second to accept on the frontend.
4660
4661 When the frontend reaches the specified number of new sessions per second, it
4662 stops accepting new connections until the rate drops below the limit again.
4663 During this time, the pending sessions will be kept in the socket's backlog
4664 (in system buffers) and haproxy will not even be aware that sessions are
4665 pending. When applying very low limit on a highly loaded service, it may make
4666 sense to increase the socket's backlog using the "backlog" keyword.
4667
4668 This feature is particularly efficient at blocking connection-based attacks
4669 or service abuse on fragile servers. Since the session rate is measured every
4670 millisecond, it is extremely accurate. Also, the limit applies immediately,
4671 no delay is needed at all to detect the threshold.
4672
4673 Example : limit the connection rate on SMTP to 10 per second max
4674 listen smtp
4675 mode tcp
4676 bind :25
4677 rate-limit sessions 10
4678 server 127.0.0.1:1025
4679
Willy Tarreaua17c2d92011-07-25 08:16:20 +02004680 Note : when the maximum rate is reached, the frontend's status is not changed
4681 but its sockets appear as "WAITING" in the statistics if the
4682 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +01004683
4684 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
4685
4686
Willy Tarreau2e1dca82012-09-12 08:43:15 +02004687redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
4688redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
4689redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +02004690 Return an HTTP redirection if/unless a condition is matched
4691 May be used in sections : defaults | frontend | listen | backend
4692 no | yes | yes | yes
4693
4694 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +01004695 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02004696
Willy Tarreau0140f252008-11-19 21:07:09 +01004697 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +02004698 <loc> With "redirect location", the exact value in <loc> is placed into
4699 the HTTP "Location" header.
4700
4701 <pfx> With "redirect prefix", the "Location" header is built from the
4702 concatenation of <pfx> and the complete URI path, including the
4703 query string, unless the "drop-query" option is specified (see
4704 below). As a special case, if <pfx> equals exactly "/", then
4705 nothing is inserted before the original URI. It allows one to
4706 redirect to the same URL (for instance, to insert a cookie).
4707
4708 <sch> With "redirect scheme", then the "Location" header is built by
4709 concatenating <sch> with "://" then the first occurrence of the
4710 "Host" header, and then the URI path, including the query string
4711 unless the "drop-query" option is specified (see below). If no
4712 path is found or if the path is "*", then "/" is used instead. If
4713 no "Host" header is found, then an empty host component will be
4714 returned, which most recent browsers interprete as redirecting to
4715 the same host. This directive is mostly used to redirect HTTP to
4716 HTTPS.
Willy Tarreau0140f252008-11-19 21:07:09 +01004717
4718 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +01004719 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
4720 with 302 used by default if no code is specified. 301 means
4721 "Moved permanently", and a browser may cache the Location. 302
4722 means "Moved permanently" and means that the browser should not
4723 cache the redirection. 303 is equivalent to 302 except that the
4724 browser will fetch the location with a GET method. 307 is just
4725 like 302 but makes it clear that the same method must be reused.
4726 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +01004727
4728 <option> There are several options which can be specified to adjust the
4729 expected behaviour of a redirection :
4730
4731 - "drop-query"
4732 When this keyword is used in a prefix-based redirection, then the
4733 location will be set without any possible query-string, which is useful
4734 for directing users to a non-secure page for instance. It has no effect
4735 with a location-type redirect.
4736
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01004737 - "append-slash"
4738 This keyword may be used in conjunction with "drop-query" to redirect
4739 users who use a URL not ending with a '/' to the same one with the '/'.
4740 It can be useful to ensure that search engines will only see one URL.
4741 For this, a return code 301 is preferred.
4742
Willy Tarreau0140f252008-11-19 21:07:09 +01004743 - "set-cookie NAME[=value]"
4744 A "Set-Cookie" header will be added with NAME (and optionally "=value")
4745 to the response. This is sometimes used to indicate that a user has
4746 been seen, for instance to protect against some types of DoS. No other
4747 cookie option is added, so the cookie will be a session cookie. Note
4748 that for a browser, a sole cookie name without an equal sign is
4749 different from a cookie with an equal sign.
4750
4751 - "clear-cookie NAME[=]"
4752 A "Set-Cookie" header will be added with NAME (and optionally "="), but
4753 with the "Max-Age" attribute set to zero. This will tell the browser to
4754 delete this cookie. It is useful for instance on logout pages. It is
4755 important to note that clearing the cookie "NAME" will not remove a
4756 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
4757 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02004758
4759 Example: move the login URL only to HTTPS.
4760 acl clear dst_port 80
4761 acl secure dst_port 8080
4762 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +01004763 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +01004764 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +01004765 acl cookie_set hdr_sub(cookie) SEEN=1
4766
4767 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +01004768 redirect prefix https://mysite.com if login_page !secure
4769 redirect prefix http://mysite.com drop-query if login_page !uid_given
4770 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +01004771 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +02004772
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01004773 Example: send redirects for request for articles without a '/'.
4774 acl missing_slash path_reg ^/article/[^/]*$
4775 redirect code 301 prefix / drop-query append-slash if missing_slash
4776
Willy Tarreau2e1dca82012-09-12 08:43:15 +02004777 Example: redirect all HTTP traffic to HTTPS when SSL is handled by haproxy.
David BERARDe7153042012-11-03 00:11:31 +01004778 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +02004779
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004780 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02004781
4782
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004783redisp (deprecated)
4784redispatch (deprecated)
4785 Enable or disable session redistribution in case of connection failure
4786 May be used in sections: defaults | frontend | listen | backend
4787 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01004788 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004789
4790 In HTTP mode, if a server designated by a cookie is down, clients may
4791 definitely stick to it because they cannot flush the cookie, so they will not
4792 be able to access the service anymore.
4793
4794 Specifying "redispatch" will allow the proxy to break their persistence and
4795 redistribute them to a working server.
4796
4797 It also allows to retry last connection to another server in case of multiple
4798 connection failures. Of course, it requires having "retries" set to a nonzero
4799 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01004800
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004801 This form is deprecated, do not use it in any new configuration, use the new
4802 "option redispatch" instead.
4803
4804 See also : "option redispatch"
4805
Willy Tarreaueabeafa2008-01-16 16:17:06 +01004806
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01004807reqadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01004808 Add a header at the end of the HTTP request
4809 May be used in sections : defaults | frontend | listen | backend
4810 no | yes | yes | yes
4811 Arguments :
4812 <string> is the complete line to be added. Any space or known delimiter
4813 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004814 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01004815
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01004816 <cond> is an optional matching condition built from ACLs. It makes it
4817 possible to ignore this rule when other conditions are not met.
4818
Willy Tarreau303c0352008-01-17 19:01:39 +01004819 A new line consisting in <string> followed by a line feed will be added after
4820 the last header of an HTTP request.
4821
4822 Header transformations only apply to traffic which passes through HAProxy,
4823 and not to traffic generated by HAProxy, such as health-checks or error
4824 responses.
4825
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01004826 Example : add "X-Proto: SSL" to requests coming via port 81
4827 acl is-ssl dst_port 81
4828 reqadd X-Proto:\ SSL if is-ssl
4829
4830 See also: "rspadd", section 6 about HTTP header manipulation, and section 7
4831 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01004832
4833
Willy Tarreau5321c422010-01-28 20:35:13 +01004834reqallow <search> [{if | unless} <cond>]
4835reqiallow <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01004836 Definitely allow an HTTP request if a line matches a regular expression
4837 May be used in sections : defaults | frontend | listen | backend
4838 no | yes | yes | yes
4839 Arguments :
4840 <search> is the regular expression applied to HTTP headers and to the
4841 request line. This is an extended regular expression. Parenthesis
4842 grouping is supported and no preliminary backslash is required.
4843 Any space or known delimiter must be escaped using a backslash
4844 ('\'). The pattern applies to a full line at a time. The
4845 "reqallow" keyword strictly matches case while "reqiallow"
4846 ignores case.
4847
Willy Tarreau5321c422010-01-28 20:35:13 +01004848 <cond> is an optional matching condition built from ACLs. It makes it
4849 possible to ignore this rule when other conditions are not met.
4850
Willy Tarreau303c0352008-01-17 19:01:39 +01004851 A request containing any line which matches extended regular expression
4852 <search> will mark the request as allowed, even if any later test would
4853 result in a deny. The test applies both to the request line and to request
4854 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01004855 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01004856
4857 It is easier, faster and more powerful to use ACLs to write access policies.
4858 Reqdeny, reqallow and reqpass should be avoided in new designs.
4859
4860 Example :
4861 # allow www.* but refuse *.local
4862 reqiallow ^Host:\ www\.
4863 reqideny ^Host:\ .*\.local
4864
Willy Tarreau5321c422010-01-28 20:35:13 +01004865 See also: "reqdeny", "block", section 6 about HTTP header manipulation, and
4866 section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01004867
4868
Willy Tarreau5321c422010-01-28 20:35:13 +01004869reqdel <search> [{if | unless} <cond>]
4870reqidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01004871 Delete all headers matching a regular expression in an HTTP request
4872 May be used in sections : defaults | frontend | listen | backend
4873 no | yes | yes | yes
4874 Arguments :
4875 <search> is the regular expression applied to HTTP headers and to the
4876 request line. This is an extended regular expression. Parenthesis
4877 grouping is supported and no preliminary backslash is required.
4878 Any space or known delimiter must be escaped using a backslash
4879 ('\'). The pattern applies to a full line at a time. The "reqdel"
4880 keyword strictly matches case while "reqidel" ignores case.
4881
Willy Tarreau5321c422010-01-28 20:35:13 +01004882 <cond> is an optional matching condition built from ACLs. It makes it
4883 possible to ignore this rule when other conditions are not met.
4884
Willy Tarreau303c0352008-01-17 19:01:39 +01004885 Any header line matching extended regular expression <search> in the request
4886 will be completely deleted. Most common use of this is to remove unwanted
4887 and/or dangerous headers or cookies from a request before passing it to the
4888 next servers.
4889
4890 Header transformations only apply to traffic which passes through HAProxy,
4891 and not to traffic generated by HAProxy, such as health-checks or error
4892 responses. Keep in mind that header names are not case-sensitive.
4893
4894 Example :
4895 # remove X-Forwarded-For header and SERVER cookie
4896 reqidel ^X-Forwarded-For:.*
4897 reqidel ^Cookie:.*SERVER=
4898
Willy Tarreau5321c422010-01-28 20:35:13 +01004899 See also: "reqadd", "reqrep", "rspdel", section 6 about HTTP header
4900 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01004901
4902
Willy Tarreau5321c422010-01-28 20:35:13 +01004903reqdeny <search> [{if | unless} <cond>]
4904reqideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01004905 Deny an HTTP request if a line matches a regular expression
4906 May be used in sections : defaults | frontend | listen | backend
4907 no | yes | yes | yes
4908 Arguments :
4909 <search> is the regular expression applied to HTTP headers and to the
4910 request line. This is an extended regular expression. Parenthesis
4911 grouping is supported and no preliminary backslash is required.
4912 Any space or known delimiter must be escaped using a backslash
4913 ('\'). The pattern applies to a full line at a time. The
4914 "reqdeny" keyword strictly matches case while "reqideny" ignores
4915 case.
4916
Willy Tarreau5321c422010-01-28 20:35:13 +01004917 <cond> is an optional matching condition built from ACLs. It makes it
4918 possible to ignore this rule when other conditions are not met.
4919
Willy Tarreau303c0352008-01-17 19:01:39 +01004920 A request containing any line which matches extended regular expression
4921 <search> will mark the request as denied, even if any later test would
4922 result in an allow. The test applies both to the request line and to request
4923 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01004924 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01004925
Willy Tarreauced27012008-01-17 20:35:34 +01004926 A denied request will generate an "HTTP 403 forbidden" response once the
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01004927 complete request has been parsed. This is consistent with what is practiced
Willy Tarreaud72758d2010-01-12 10:42:19 +01004928 using ACLs.
Willy Tarreauced27012008-01-17 20:35:34 +01004929
Willy Tarreau303c0352008-01-17 19:01:39 +01004930 It is easier, faster and more powerful to use ACLs to write access policies.
4931 Reqdeny, reqallow and reqpass should be avoided in new designs.
4932
4933 Example :
4934 # refuse *.local, then allow www.*
4935 reqideny ^Host:\ .*\.local
4936 reqiallow ^Host:\ www\.
4937
Willy Tarreau5321c422010-01-28 20:35:13 +01004938 See also: "reqallow", "rspdeny", "block", section 6 about HTTP header
4939 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01004940
4941
Willy Tarreau5321c422010-01-28 20:35:13 +01004942reqpass <search> [{if | unless} <cond>]
4943reqipass <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01004944 Ignore any HTTP request line matching a regular expression in next rules
4945 May be used in sections : defaults | frontend | listen | backend
4946 no | yes | yes | yes
4947 Arguments :
4948 <search> is the regular expression applied to HTTP headers and to the
4949 request line. This is an extended regular expression. Parenthesis
4950 grouping is supported and no preliminary backslash is required.
4951 Any space or known delimiter must be escaped using a backslash
4952 ('\'). The pattern applies to a full line at a time. The
4953 "reqpass" keyword strictly matches case while "reqipass" ignores
4954 case.
4955
Willy Tarreau5321c422010-01-28 20:35:13 +01004956 <cond> is an optional matching condition built from ACLs. It makes it
4957 possible to ignore this rule when other conditions are not met.
4958
Willy Tarreau303c0352008-01-17 19:01:39 +01004959 A request containing any line which matches extended regular expression
4960 <search> will skip next rules, without assigning any deny or allow verdict.
4961 The test applies both to the request line and to request headers. Keep in
4962 mind that URLs in request line are case-sensitive while header names are not.
4963
4964 It is easier, faster and more powerful to use ACLs to write access policies.
4965 Reqdeny, reqallow and reqpass should be avoided in new designs.
4966
4967 Example :
4968 # refuse *.local, then allow www.*, but ignore "www.private.local"
4969 reqipass ^Host:\ www.private\.local
4970 reqideny ^Host:\ .*\.local
4971 reqiallow ^Host:\ www\.
4972
Willy Tarreau5321c422010-01-28 20:35:13 +01004973 See also: "reqallow", "reqdeny", "block", section 6 about HTTP header
4974 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01004975
4976
Willy Tarreau5321c422010-01-28 20:35:13 +01004977reqrep <search> <string> [{if | unless} <cond>]
4978reqirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01004979 Replace a regular expression with a string in an HTTP request line
4980 May be used in sections : defaults | frontend | listen | backend
4981 no | yes | yes | yes
4982 Arguments :
4983 <search> is the regular expression applied to HTTP headers and to the
4984 request line. This is an extended regular expression. Parenthesis
4985 grouping is supported and no preliminary backslash is required.
4986 Any space or known delimiter must be escaped using a backslash
4987 ('\'). The pattern applies to a full line at a time. The "reqrep"
4988 keyword strictly matches case while "reqirep" ignores case.
4989
4990 <string> is the complete line to be added. Any space or known delimiter
4991 must be escaped using a backslash ('\'). References to matched
4992 pattern groups are possible using the common \N form, with N
4993 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004994 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01004995
Willy Tarreau5321c422010-01-28 20:35:13 +01004996 <cond> is an optional matching condition built from ACLs. It makes it
4997 possible to ignore this rule when other conditions are not met.
4998
Willy Tarreau303c0352008-01-17 19:01:39 +01004999 Any line matching extended regular expression <search> in the request (both
5000 the request line and header lines) will be completely replaced with <string>.
5001 Most common use of this is to rewrite URLs or domain names in "Host" headers.
5002
5003 Header transformations only apply to traffic which passes through HAProxy,
5004 and not to traffic generated by HAProxy, such as health-checks or error
5005 responses. Note that for increased readability, it is suggested to add enough
5006 spaces between the request and the response. Keep in mind that URLs in
5007 request line are case-sensitive while header names are not.
5008
5009 Example :
5010 # replace "/static/" with "/" at the beginning of any request path.
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04005011 reqrep ^([^\ :]*)\ /static/(.*) \1\ /\2
Willy Tarreau303c0352008-01-17 19:01:39 +01005012 # replace "www.mydomain.com" with "www" in the host name.
5013 reqirep ^Host:\ www.mydomain.com Host:\ www
5014
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04005015 See also: "reqadd", "reqdel", "rsprep", "tune.bufsize", section 6 about
5016 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005017
5018
Willy Tarreau5321c422010-01-28 20:35:13 +01005019reqtarpit <search> [{if | unless} <cond>]
5020reqitarpit <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005021 Tarpit an HTTP request containing a line matching a regular expression
5022 May be used in sections : defaults | frontend | listen | backend
5023 no | yes | yes | yes
5024 Arguments :
5025 <search> is the regular expression applied to HTTP headers and to the
5026 request line. This is an extended regular expression. Parenthesis
5027 grouping is supported and no preliminary backslash is required.
5028 Any space or known delimiter must be escaped using a backslash
5029 ('\'). The pattern applies to a full line at a time. The
5030 "reqtarpit" keyword strictly matches case while "reqitarpit"
5031 ignores case.
5032
Willy Tarreau5321c422010-01-28 20:35:13 +01005033 <cond> is an optional matching condition built from ACLs. It makes it
5034 possible to ignore this rule when other conditions are not met.
5035
Willy Tarreau303c0352008-01-17 19:01:39 +01005036 A request containing any line which matches extended regular expression
5037 <search> will be tarpitted, which means that it will connect to nowhere, will
Willy Tarreauced27012008-01-17 20:35:34 +01005038 be kept open for a pre-defined time, then will return an HTTP error 500 so
5039 that the attacker does not suspect it has been tarpitted. The status 500 will
5040 be reported in the logs, but the completion flags will indicate "PT". The
Willy Tarreau303c0352008-01-17 19:01:39 +01005041 delay is defined by "timeout tarpit", or "timeout connect" if the former is
5042 not set.
5043
5044 The goal of the tarpit is to slow down robots attacking servers with
5045 identifiable requests. Many robots limit their outgoing number of connections
5046 and stay connected waiting for a reply which can take several minutes to
5047 come. Depending on the environment and attack, it may be particularly
5048 efficient at reducing the load on the network and firewalls.
5049
Willy Tarreau5321c422010-01-28 20:35:13 +01005050 Examples :
Willy Tarreau303c0352008-01-17 19:01:39 +01005051 # ignore user-agents reporting any flavour of "Mozilla" or "MSIE", but
5052 # block all others.
5053 reqipass ^User-Agent:\.*(Mozilla|MSIE)
5054 reqitarpit ^User-Agent:
5055
Willy Tarreau5321c422010-01-28 20:35:13 +01005056 # block bad guys
5057 acl badguys src 10.1.0.3 172.16.13.20/28
5058 reqitarpit . if badguys
5059
5060 See also: "reqallow", "reqdeny", "reqpass", section 6 about HTTP header
5061 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005062
5063
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02005064retries <value>
5065 Set the number of retries to perform on a server after a connection failure
5066 May be used in sections: defaults | frontend | listen | backend
5067 yes | no | yes | yes
5068 Arguments :
5069 <value> is the number of times a connection attempt should be retried on
5070 a server when a connection either is refused or times out. The
5071 default value is 3.
5072
5073 It is important to understand that this value applies to the number of
5074 connection attempts, not full requests. When a connection has effectively
5075 been established to a server, there will be no more retry.
5076
5077 In order to avoid immediate reconnections to a server which is restarting,
5078 a turn-around timer of 1 second is applied before a retry occurs.
5079
5080 When "option redispatch" is set, the last retry may be performed on another
5081 server even if a cookie references a different server.
5082
5083 See also : "option redispatch"
5084
5085
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005086rspadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01005087 Add a header at the end of the HTTP response
5088 May be used in sections : defaults | frontend | listen | backend
5089 no | yes | yes | yes
5090 Arguments :
5091 <string> is the complete line to be added. Any space or known delimiter
5092 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005093 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01005094
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005095 <cond> is an optional matching condition built from ACLs. It makes it
5096 possible to ignore this rule when other conditions are not met.
5097
Willy Tarreau303c0352008-01-17 19:01:39 +01005098 A new line consisting in <string> followed by a line feed will be added after
5099 the last header of an HTTP response.
5100
5101 Header transformations only apply to traffic which passes through HAProxy,
5102 and not to traffic generated by HAProxy, such as health-checks or error
5103 responses.
5104
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005105 See also: "reqadd", section 6 about HTTP header manipulation, and section 7
5106 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005107
5108
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005109rspdel <search> [{if | unless} <cond>]
5110rspidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005111 Delete all headers matching a regular expression in an HTTP response
5112 May be used in sections : defaults | frontend | listen | backend
5113 no | yes | yes | yes
5114 Arguments :
5115 <search> is the regular expression applied to HTTP headers and to the
5116 response line. This is an extended regular expression, so
5117 parenthesis grouping is supported and no preliminary backslash
5118 is required. Any space or known delimiter must be escaped using
5119 a backslash ('\'). The pattern applies to a full line at a time.
5120 The "rspdel" keyword strictly matches case while "rspidel"
5121 ignores case.
5122
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005123 <cond> is an optional matching condition built from ACLs. It makes it
5124 possible to ignore this rule when other conditions are not met.
5125
Willy Tarreau303c0352008-01-17 19:01:39 +01005126 Any header line matching extended regular expression <search> in the response
5127 will be completely deleted. Most common use of this is to remove unwanted
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02005128 and/or sensitive headers or cookies from a response before passing it to the
Willy Tarreau303c0352008-01-17 19:01:39 +01005129 client.
5130
5131 Header transformations only apply to traffic which passes through HAProxy,
5132 and not to traffic generated by HAProxy, such as health-checks or error
5133 responses. Keep in mind that header names are not case-sensitive.
5134
5135 Example :
5136 # remove the Server header from responses
Willy Tarreau5e80e022013-05-25 08:31:25 +02005137 rspidel ^Server:.*
Willy Tarreau303c0352008-01-17 19:01:39 +01005138
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005139 See also: "rspadd", "rsprep", "reqdel", section 6 about HTTP header
5140 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005141
5142
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005143rspdeny <search> [{if | unless} <cond>]
5144rspideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005145 Block an HTTP response if a line matches a regular expression
5146 May be used in sections : defaults | frontend | listen | backend
5147 no | yes | yes | yes
5148 Arguments :
5149 <search> is the regular expression applied to HTTP headers and to the
5150 response line. This is an extended regular expression, so
5151 parenthesis grouping is supported and no preliminary backslash
5152 is required. Any space or known delimiter must be escaped using
5153 a backslash ('\'). The pattern applies to a full line at a time.
5154 The "rspdeny" keyword strictly matches case while "rspideny"
5155 ignores case.
5156
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005157 <cond> is an optional matching condition built from ACLs. It makes it
5158 possible to ignore this rule when other conditions are not met.
5159
Willy Tarreau303c0352008-01-17 19:01:39 +01005160 A response containing any line which matches extended regular expression
5161 <search> will mark the request as denied. The test applies both to the
5162 response line and to response headers. Keep in mind that header names are not
5163 case-sensitive.
5164
5165 Main use of this keyword is to prevent sensitive information leak and to
Willy Tarreauced27012008-01-17 20:35:34 +01005166 block the response before it reaches the client. If a response is denied, it
5167 will be replaced with an HTTP 502 error so that the client never retrieves
5168 any sensitive data.
Willy Tarreau303c0352008-01-17 19:01:39 +01005169
5170 It is easier, faster and more powerful to use ACLs to write access policies.
5171 Rspdeny should be avoided in new designs.
5172
5173 Example :
5174 # Ensure that no content type matching ms-word will leak
5175 rspideny ^Content-type:\.*/ms-word
5176
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005177 See also: "reqdeny", "acl", "block", section 6 about HTTP header manipulation
5178 and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005179
5180
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005181rsprep <search> <string> [{if | unless} <cond>]
5182rspirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005183 Replace a regular expression with a string in an HTTP response line
5184 May be used in sections : defaults | frontend | listen | backend
5185 no | yes | yes | yes
5186 Arguments :
5187 <search> is the regular expression applied to HTTP headers and to the
5188 response line. This is an extended regular expression, so
5189 parenthesis grouping is supported and no preliminary backslash
5190 is required. Any space or known delimiter must be escaped using
5191 a backslash ('\'). The pattern applies to a full line at a time.
5192 The "rsprep" keyword strictly matches case while "rspirep"
5193 ignores case.
5194
5195 <string> is the complete line to be added. Any space or known delimiter
5196 must be escaped using a backslash ('\'). References to matched
5197 pattern groups are possible using the common \N form, with N
5198 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005199 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01005200
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005201 <cond> is an optional matching condition built from ACLs. It makes it
5202 possible to ignore this rule when other conditions are not met.
5203
Willy Tarreau303c0352008-01-17 19:01:39 +01005204 Any line matching extended regular expression <search> in the response (both
5205 the response line and header lines) will be completely replaced with
5206 <string>. Most common use of this is to rewrite Location headers.
5207
5208 Header transformations only apply to traffic which passes through HAProxy,
5209 and not to traffic generated by HAProxy, such as health-checks or error
5210 responses. Note that for increased readability, it is suggested to add enough
5211 spaces between the request and the response. Keep in mind that header names
5212 are not case-sensitive.
5213
5214 Example :
5215 # replace "Location: 127.0.0.1:8080" with "Location: www.mydomain.com"
5216 rspirep ^Location:\ 127.0.0.1:8080 Location:\ www.mydomain.com
5217
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005218 See also: "rspadd", "rspdel", "reqrep", section 6 about HTTP header
5219 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005220
5221
David du Colombier486df472011-03-17 10:40:26 +01005222server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005223 Declare a server in a backend
5224 May be used in sections : defaults | frontend | listen | backend
5225 no | no | yes | yes
5226 Arguments :
5227 <name> is the internal name assigned to this server. This name will
Cyril Bonté941a0c62012-10-15 19:44:24 +02005228 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -05005229 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005230
David du Colombier486df472011-03-17 10:40:26 +01005231 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
5232 resolvable hostname is supported, but this name will be resolved
5233 during start-up. Address "0.0.0.0" or "*" has a special meaning.
5234 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +02005235 address as the one from the client connection. This is useful in
5236 transparent proxy architectures where the client's connection is
5237 intercepted and haproxy must forward to the original destination
5238 address. This is more or less what the "transparent" keyword does
5239 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +01005240 to report statistics. Optionally, an address family prefix may be
5241 used before the address to force the family regardless of the
5242 address format, which can be useful to specify a path to a unix
5243 socket with no slash ('/'). Currently supported prefixes are :
5244 - 'ipv4@' -> address is always IPv4
5245 - 'ipv6@' -> address is always IPv6
5246 - 'unix@' -> address is a path to a local unix socket
Willy Tarreaudad36a32013-03-11 01:20:04 +01005247 Any part of the address string may reference any number of
5248 environment variables by preceding their name with a dollar
5249 sign ('$') and optionally enclosing them with braces ('{}'),
5250 similarly to what is done in Bourne shell.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005251
Willy Tarreaub6205fd2012-09-24 12:27:33 +02005252 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005253 be sent to this port. If unset, the same port the client
5254 connected to will be used. The port may also be prefixed by a "+"
5255 or a "-". In this case, the server's port will be determined by
5256 adding this value to the client's port.
5257
5258 <param*> is a list of parameters for this server. The "server" keywords
5259 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005260 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005261
5262 Examples :
5263 server first 10.1.1.1:1080 cookie first check inter 1000
5264 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +01005265 server transp ipv4@
Willy Tarreaudad36a32013-03-11 01:20:04 +01005266 server backup ${SRV_BACKUP}:1080 backup
5267 server www1_dc1 ${LAN_DC1}.101:80
5268 server www1_dc2 ${LAN_DC2}.101:80
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005269
Mark Lamourinec2247f02012-01-04 13:02:01 -05005270 See also: "default-server", "http-send-name-header" and section 5 about
5271 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005272
5273
5274source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02005275source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +01005276source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005277 Set the source address for outgoing connections
5278 May be used in sections : defaults | frontend | listen | backend
5279 yes | no | yes | yes
5280 Arguments :
5281 <addr> is the IPv4 address HAProxy will bind to before connecting to a
5282 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +01005283
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005284 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +01005285 the most appropriate address to reach its destination. Optionally
5286 an address family prefix may be used before the address to force
5287 the family regardless of the address format, which can be useful
5288 to specify a path to a unix socket with no slash ('/'). Currently
5289 supported prefixes are :
5290 - 'ipv4@' -> address is always IPv4
5291 - 'ipv6@' -> address is always IPv6
5292 - 'unix@' -> address is a path to a local unix socket
Willy Tarreaudad36a32013-03-11 01:20:04 +01005293 Any part of the address string may reference any number of
5294 environment variables by preceding their name with a dollar
5295 sign ('$') and optionally enclosing them with braces ('{}'),
5296 similarly to what is done in Bourne shell.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005297
5298 <port> is an optional port. It is normally not needed but may be useful
5299 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02005300 the system will select a free port. Note that port ranges are not
5301 supported in the backend. If you want to force port ranges, you
5302 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005303
5304 <addr2> is the IP address to present to the server when connections are
5305 forwarded in full transparent proxy mode. This is currently only
5306 supported on some patched Linux kernels. When this address is
5307 specified, clients connecting to the server will be presented
5308 with this address, while health checks will still use the address
5309 <addr>.
5310
5311 <port2> is the optional port to present to the server when connections
5312 are forwarded in full transparent proxy mode (see <addr2> above).
5313 The default value of zero means the system will select a free
5314 port.
5315
Willy Tarreaubce70882009-09-07 11:51:47 +02005316 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
5317 This is the name of a comma-separated header list which can
5318 contain multiple IP addresses. By default, the last occurrence is
5319 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +01005320 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +02005321 by previous proxy, typically Stunnel. In order to use another
5322 occurrence from the last one, please see the <occ> parameter
5323 below. When the header (or occurrence) is not found, no binding
5324 is performed so that the proxy's default IP address is used. Also
5325 keep in mind that the header name is case insensitive, as for any
5326 HTTP header.
5327
5328 <occ> is the occurrence number of a value to be used in a multi-value
5329 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005330 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +02005331 address. Positive values indicate a position from the first
5332 occurrence, 1 being the first one. Negative values indicate
5333 positions relative to the last one, -1 being the last one. This
5334 is helpful for situations where an X-Forwarded-For header is set
5335 at the entry point of an infrastructure and must be used several
5336 proxy layers away. When this value is not specified, -1 is
5337 assumed. Passing a zero here disables the feature.
5338
Willy Tarreaud53f96b2009-02-04 18:46:54 +01005339 <name> is an optional interface name to which to bind to for outgoing
5340 traffic. On systems supporting this features (currently, only
5341 Linux), this allows one to bind all traffic to the server to
5342 this interface even if it is not the one the system would select
5343 based on routing tables. This should be used with extreme care.
5344 Note that using this option requires root privileges.
5345
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005346 The "source" keyword is useful in complex environments where a specific
5347 address only is allowed to connect to the servers. It may be needed when a
5348 private address must be used through a public gateway for instance, and it is
5349 known that the system cannot determine the adequate source address by itself.
5350
5351 An extension which is available on certain patched Linux kernels may be used
5352 through the "usesrc" optional keyword. It makes it possible to connect to the
5353 servers with an IP address which does not belong to the system itself. This
5354 is called "full transparent proxy mode". For this to work, the destination
5355 servers have to route their traffic back to this address through the machine
5356 running HAProxy, and IP forwarding must generally be enabled on this machine.
5357
5358 In this "full transparent proxy" mode, it is possible to force a specific IP
5359 address to be presented to the servers. This is not much used in fact. A more
5360 common use is to tell HAProxy to present the client's IP address. For this,
5361 there are two methods :
5362
5363 - present the client's IP and port addresses. This is the most transparent
5364 mode, but it can cause problems when IP connection tracking is enabled on
5365 the machine, because a same connection may be seen twice with different
5366 states. However, this solution presents the huge advantage of not
5367 limiting the system to the 64k outgoing address+port couples, because all
5368 of the client ranges may be used.
5369
5370 - present only the client's IP address and select a spare port. This
5371 solution is still quite elegant but slightly less transparent (downstream
5372 firewalls logs will not match upstream's). It also presents the downside
5373 of limiting the number of concurrent connections to the usual 64k ports.
5374 However, since the upstream and downstream ports are different, local IP
5375 connection tracking on the machine will not be upset by the reuse of the
5376 same session.
5377
5378 Note that depending on the transparent proxy technology used, it may be
5379 required to force the source address. In fact, cttproxy version 2 requires an
5380 IP address in <addr> above, and does not support setting of "0.0.0.0" as the
5381 IP address because it creates NAT entries which much match the exact outgoing
5382 address. Tproxy version 4 and some other kernel patches which work in pure
5383 forwarding mode generally will not have this limitation.
5384
5385 This option sets the default source for all servers in the backend. It may
5386 also be specified in a "defaults" section. Finer source address specification
5387 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005388 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005389
5390 Examples :
5391 backend private
5392 # Connect to the servers using our 192.168.1.200 source address
5393 source 192.168.1.200
5394
5395 backend transparent_ssl1
5396 # Connect to the SSL farm from the client's source address
5397 source 192.168.1.200 usesrc clientip
5398
5399 backend transparent_ssl2
5400 # Connect to the SSL farm from the client's source address and port
5401 # not recommended if IP conntrack is present on the local machine.
5402 source 192.168.1.200 usesrc client
5403
5404 backend transparent_ssl3
5405 # Connect to the SSL farm from the client's source address. It
5406 # is more conntrack-friendly.
5407 source 192.168.1.200 usesrc clientip
5408
5409 backend transparent_smtp
5410 # Connect to the SMTP farm from the client's source address/port
5411 # with Tproxy version 4.
5412 source 0.0.0.0 usesrc clientip
5413
Willy Tarreaubce70882009-09-07 11:51:47 +02005414 backend transparent_http
5415 # Connect to the servers using the client's IP as seen by previous
5416 # proxy.
5417 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
5418
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005419 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005420 the Linux kernel on www.balabit.com, the "bind" keyword.
5421
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005422
Willy Tarreau844e3c52008-01-11 16:28:18 +01005423srvtimeout <timeout> (deprecated)
5424 Set the maximum inactivity time on the server side.
5425 May be used in sections : defaults | frontend | listen | backend
5426 yes | no | yes | yes
5427 Arguments :
5428 <timeout> is the timeout value specified in milliseconds by default, but
5429 can be in any other unit if the number is suffixed by the unit,
5430 as explained at the top of this document.
5431
5432 The inactivity timeout applies when the server is expected to acknowledge or
5433 send data. In HTTP mode, this timeout is particularly important to consider
5434 during the first phase of the server's response, when it has to send the
5435 headers, as it directly represents the server's processing time for the
5436 request. To find out what value to put there, it's often good to start with
5437 what would be considered as unacceptable response times, then check the logs
5438 to observe the response time distribution, and adjust the value accordingly.
5439
5440 The value is specified in milliseconds by default, but can be in any other
5441 unit if the number is suffixed by the unit, as specified at the top of this
5442 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
5443 recommended that the client timeout remains equal to the server timeout in
5444 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01005445 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01005446 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreaud72758d2010-01-12 10:42:19 +01005447 seconds (eg: 4 or 5 seconds minimum).
Willy Tarreau844e3c52008-01-11 16:28:18 +01005448
5449 This parameter is specific to backends, but can be specified once for all in
5450 "defaults" sections. This is in fact one of the easiest solutions not to
5451 forget about it. An unspecified timeout results in an infinite timeout, which
5452 is not recommended. Such a usage is accepted and works but reports a warning
5453 during startup because it may results in accumulation of expired sessions in
5454 the system if the system's timeouts are not configured either.
5455
5456 This parameter is provided for compatibility but is currently deprecated.
5457 Please use "timeout server" instead.
5458
Willy Tarreauce887fd2012-05-12 12:50:00 +02005459 See also : "timeout server", "timeout tunnel", "timeout client" and
5460 "clitimeout".
Willy Tarreau844e3c52008-01-11 16:28:18 +01005461
5462
Cyril Bonté66c327d2010-10-12 00:14:37 +02005463stats admin { if | unless } <cond>
5464 Enable statistics admin level if/unless a condition is matched
5465 May be used in sections : defaults | frontend | listen | backend
5466 no | no | yes | yes
5467
5468 This statement enables the statistics admin level if/unless a condition is
5469 matched.
5470
5471 The admin level allows to enable/disable servers from the web interface. By
5472 default, statistics page is read-only for security reasons.
5473
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01005474 Note : Consider not using this feature in multi-process mode (nbproc > 1)
5475 unless you know what you do : memory is not shared between the
5476 processes, which can result in random behaviours.
5477
Cyril Bonté23b39d92011-02-10 22:54:44 +01005478 Currently, the POST request is limited to the buffer size minus the reserved
5479 buffer space, which means that if the list of servers is too long, the
5480 request won't be processed. It is recommended to alter few servers at a
5481 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +02005482
5483 Example :
5484 # statistics admin level only for localhost
5485 backend stats_localhost
5486 stats enable
5487 stats admin if LOCALHOST
5488
5489 Example :
5490 # statistics admin level always enabled because of the authentication
5491 backend stats_auth
5492 stats enable
5493 stats auth admin:AdMiN123
5494 stats admin if TRUE
5495
5496 Example :
5497 # statistics admin level depends on the authenticated user
5498 userlist stats-auth
5499 group admin users admin
5500 user admin insecure-password AdMiN123
5501 group readonly users haproxy
5502 user haproxy insecure-password haproxy
5503
5504 backend stats_auth
5505 stats enable
5506 acl AUTH http_auth(stats-auth)
5507 acl AUTH_ADMIN http_auth_group(stats-auth) admin
5508 stats http-request auth unless AUTH
5509 stats admin if AUTH_ADMIN
5510
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01005511 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
5512 "bind-process", section 3.4 about userlists and section 7 about
5513 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +02005514
5515
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005516stats auth <user>:<passwd>
5517 Enable statistics with authentication and grant access to an account
5518 May be used in sections : defaults | frontend | listen | backend
5519 yes | no | yes | yes
5520 Arguments :
5521 <user> is a user name to grant access to
5522
5523 <passwd> is the cleartext password associated to this user
5524
5525 This statement enables statistics with default settings, and restricts access
5526 to declared users only. It may be repeated as many times as necessary to
5527 allow as many users as desired. When a user tries to access the statistics
5528 without a valid account, a "401 Forbidden" response will be returned so that
5529 the browser asks the user to provide a valid user and password. The real
5530 which will be returned to the browser is configurable using "stats realm".
5531
5532 Since the authentication method is HTTP Basic Authentication, the passwords
5533 circulate in cleartext on the network. Thus, it was decided that the
5534 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02005535 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005536
5537 It is also possible to reduce the scope of the proxies which appear in the
5538 report using "stats scope".
5539
5540 Though this statement alone is enough to enable statistics reporting, it is
5541 recommended to set all other settings in order to avoid relying on default
5542 unobvious parameters.
5543
5544 Example :
5545 # public access (limited to this backend only)
5546 backend public_www
5547 server srv1 192.168.0.1:80
5548 stats enable
5549 stats hide-version
5550 stats scope .
5551 stats uri /admin?stats
5552 stats realm Haproxy\ Statistics
5553 stats auth admin1:AdMiN123
5554 stats auth admin2:AdMiN321
5555
5556 # internal monitoring access (unlimited)
5557 backend private_monitoring
5558 stats enable
5559 stats uri /admin?stats
5560 stats refresh 5s
5561
5562 See also : "stats enable", "stats realm", "stats scope", "stats uri"
5563
5564
5565stats enable
5566 Enable statistics reporting with default settings
5567 May be used in sections : defaults | frontend | listen | backend
5568 yes | no | yes | yes
5569 Arguments : none
5570
5571 This statement enables statistics reporting with default settings defined
5572 at build time. Unless stated otherwise, these settings are used :
5573 - stats uri : /haproxy?stats
5574 - stats realm : "HAProxy Statistics"
5575 - stats auth : no authentication
5576 - stats scope : no restriction
5577
5578 Though this statement alone is enough to enable statistics reporting, it is
5579 recommended to set all other settings in order to avoid relying on default
5580 unobvious parameters.
5581
5582 Example :
5583 # public access (limited to this backend only)
5584 backend public_www
5585 server srv1 192.168.0.1:80
5586 stats enable
5587 stats hide-version
5588 stats scope .
5589 stats uri /admin?stats
5590 stats realm Haproxy\ Statistics
5591 stats auth admin1:AdMiN123
5592 stats auth admin2:AdMiN321
5593
5594 # internal monitoring access (unlimited)
5595 backend private_monitoring
5596 stats enable
5597 stats uri /admin?stats
5598 stats refresh 5s
5599
5600 See also : "stats auth", "stats realm", "stats uri"
5601
5602
Willy Tarreaud63335a2010-02-26 12:56:52 +01005603stats hide-version
5604 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02005605 May be used in sections : defaults | frontend | listen | backend
5606 yes | no | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01005607 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02005608
Willy Tarreaud63335a2010-02-26 12:56:52 +01005609 By default, the stats page reports some useful status information along with
5610 the statistics. Among them is HAProxy's version. However, it is generally
5611 considered dangerous to report precise version to anyone, as it can help them
5612 target known weaknesses with specific attacks. The "stats hide-version"
5613 statement removes the version from the statistics report. This is recommended
5614 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02005615
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02005616 Though this statement alone is enough to enable statistics reporting, it is
5617 recommended to set all other settings in order to avoid relying on default
5618 unobvious parameters.
5619
Willy Tarreaud63335a2010-02-26 12:56:52 +01005620 Example :
5621 # public access (limited to this backend only)
5622 backend public_www
5623 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02005624 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01005625 stats hide-version
5626 stats scope .
5627 stats uri /admin?stats
5628 stats realm Haproxy\ Statistics
5629 stats auth admin1:AdMiN123
5630 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02005631
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02005632 # internal monitoring access (unlimited)
5633 backend private_monitoring
5634 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01005635 stats uri /admin?stats
5636 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +01005637
Willy Tarreaud63335a2010-02-26 12:56:52 +01005638 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02005639
Willy Tarreau983e01e2010-01-11 18:42:06 +01005640
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02005641stats http-request { allow | deny | auth [realm <realm>] }
5642 [ { if | unless } <condition> ]
5643 Access control for statistics
5644
5645 May be used in sections: defaults | frontend | listen | backend
5646 no | no | yes | yes
5647
5648 As "http-request", these set of options allow to fine control access to
5649 statistics. Each option may be followed by if/unless and acl.
5650 First option with matched condition (or option without condition) is final.
5651 For "deny" a 403 error will be returned, for "allow" normal processing is
5652 performed, for "auth" a 401/407 error code is returned so the client
5653 should be asked to enter a username and password.
5654
5655 There is no fixed limit to the number of http-request statements per
5656 instance.
5657
5658 See also : "http-request", section 3.4 about userlists and section 7
5659 about ACL usage.
5660
5661
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005662stats realm <realm>
5663 Enable statistics and set authentication realm
5664 May be used in sections : defaults | frontend | listen | backend
5665 yes | no | yes | yes
5666 Arguments :
5667 <realm> is the name of the HTTP Basic Authentication realm reported to
5668 the browser. The browser uses it to display it in the pop-up
5669 inviting the user to enter a valid username and password.
5670
5671 The realm is read as a single word, so any spaces in it should be escaped
5672 using a backslash ('\').
5673
5674 This statement is useful only in conjunction with "stats auth" since it is
5675 only related to authentication.
5676
5677 Though this statement alone is enough to enable statistics reporting, it is
5678 recommended to set all other settings in order to avoid relying on default
5679 unobvious parameters.
5680
5681 Example :
5682 # public access (limited to this backend only)
5683 backend public_www
5684 server srv1 192.168.0.1:80
5685 stats enable
5686 stats hide-version
5687 stats scope .
5688 stats uri /admin?stats
5689 stats realm Haproxy\ Statistics
5690 stats auth admin1:AdMiN123
5691 stats auth admin2:AdMiN321
5692
5693 # internal monitoring access (unlimited)
5694 backend private_monitoring
5695 stats enable
5696 stats uri /admin?stats
5697 stats refresh 5s
5698
5699 See also : "stats auth", "stats enable", "stats uri"
5700
5701
5702stats refresh <delay>
5703 Enable statistics with automatic refresh
5704 May be used in sections : defaults | frontend | listen | backend
5705 yes | no | yes | yes
5706 Arguments :
5707 <delay> is the suggested refresh delay, specified in seconds, which will
5708 be returned to the browser consulting the report page. While the
5709 browser is free to apply any delay, it will generally respect it
5710 and refresh the page this every seconds. The refresh interval may
5711 be specified in any other non-default time unit, by suffixing the
5712 unit after the value, as explained at the top of this document.
5713
5714 This statement is useful on monitoring displays with a permanent page
5715 reporting the load balancer's activity. When set, the HTML report page will
5716 include a link "refresh"/"stop refresh" so that the user can select whether
5717 he wants automatic refresh of the page or not.
5718
5719 Though this statement alone is enough to enable statistics reporting, it is
5720 recommended to set all other settings in order to avoid relying on default
5721 unobvious parameters.
5722
5723 Example :
5724 # public access (limited to this backend only)
5725 backend public_www
5726 server srv1 192.168.0.1:80
5727 stats enable
5728 stats hide-version
5729 stats scope .
5730 stats uri /admin?stats
5731 stats realm Haproxy\ Statistics
5732 stats auth admin1:AdMiN123
5733 stats auth admin2:AdMiN321
5734
5735 # internal monitoring access (unlimited)
5736 backend private_monitoring
5737 stats enable
5738 stats uri /admin?stats
5739 stats refresh 5s
5740
5741 See also : "stats auth", "stats enable", "stats realm", "stats uri"
5742
5743
5744stats scope { <name> | "." }
5745 Enable statistics and limit access scope
5746 May be used in sections : defaults | frontend | listen | backend
5747 yes | no | yes | yes
5748 Arguments :
5749 <name> is the name of a listen, frontend or backend section to be
5750 reported. The special name "." (a single dot) designates the
5751 section in which the statement appears.
5752
5753 When this statement is specified, only the sections enumerated with this
5754 statement will appear in the report. All other ones will be hidden. This
5755 statement may appear as many times as needed if multiple sections need to be
5756 reported. Please note that the name checking is performed as simple string
5757 comparisons, and that it is never checked that a give section name really
5758 exists.
5759
5760 Though this statement alone is enough to enable statistics reporting, it is
5761 recommended to set all other settings in order to avoid relying on default
5762 unobvious parameters.
5763
5764 Example :
5765 # public access (limited to this backend only)
5766 backend public_www
5767 server srv1 192.168.0.1:80
5768 stats enable
5769 stats hide-version
5770 stats scope .
5771 stats uri /admin?stats
5772 stats realm Haproxy\ Statistics
5773 stats auth admin1:AdMiN123
5774 stats auth admin2:AdMiN321
5775
5776 # internal monitoring access (unlimited)
5777 backend private_monitoring
5778 stats enable
5779 stats uri /admin?stats
5780 stats refresh 5s
5781
5782 See also : "stats auth", "stats enable", "stats realm", "stats uri"
5783
Willy Tarreaud63335a2010-02-26 12:56:52 +01005784
Willy Tarreauc9705a12010-07-27 20:05:50 +02005785stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +01005786 Enable reporting of a description on the statistics page.
5787 May be used in sections : defaults | frontend | listen | backend
5788 yes | no | yes | yes
5789
Willy Tarreauc9705a12010-07-27 20:05:50 +02005790 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +01005791 description from global section is automatically used instead.
5792
5793 This statement is useful for users that offer shared services to their
5794 customers, where node or description should be different for each customer.
5795
5796 Though this statement alone is enough to enable statistics reporting, it is
5797 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04005798 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +01005799
5800 Example :
5801 # internal monitoring access (unlimited)
5802 backend private_monitoring
5803 stats enable
5804 stats show-desc Master node for Europe, Asia, Africa
5805 stats uri /admin?stats
5806 stats refresh 5s
5807
5808 See also: "show-node", "stats enable", "stats uri" and "description" in
5809 global section.
5810
5811
5812stats show-legends
5813 Enable reporting additional informations on the statistics page :
5814 - cap: capabilities (proxy)
5815 - mode: one of tcp, http or health (proxy)
5816 - id: SNMP ID (proxy, socket, server)
5817 - IP (socket, server)
5818 - cookie (backend, server)
5819
5820 Though this statement alone is enough to enable statistics reporting, it is
5821 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04005822 unobvious parameters. Default behaviour is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +01005823
5824 See also: "stats enable", "stats uri".
5825
5826
5827stats show-node [ <name> ]
5828 Enable reporting of a host name on the statistics page.
5829 May be used in sections : defaults | frontend | listen | backend
5830 yes | no | yes | yes
5831 Arguments:
5832 <name> is an optional name to be reported. If unspecified, the
5833 node name from global section is automatically used instead.
5834
5835 This statement is useful for users that offer shared services to their
5836 customers, where node or description might be different on a stats page
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04005837 provided for each customer. Default behaviour is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +01005838
5839 Though this statement alone is enough to enable statistics reporting, it is
5840 recommended to set all other settings in order to avoid relying on default
5841 unobvious parameters.
5842
5843 Example:
5844 # internal monitoring access (unlimited)
5845 backend private_monitoring
5846 stats enable
5847 stats show-node Europe-1
5848 stats uri /admin?stats
5849 stats refresh 5s
5850
5851 See also: "show-desc", "stats enable", "stats uri", and "node" in global
5852 section.
5853
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005854
5855stats uri <prefix>
5856 Enable statistics and define the URI prefix to access them
5857 May be used in sections : defaults | frontend | listen | backend
5858 yes | no | yes | yes
5859 Arguments :
5860 <prefix> is the prefix of any URI which will be redirected to stats. This
5861 prefix may contain a question mark ('?') to indicate part of a
5862 query string.
5863
5864 The statistics URI is intercepted on the relayed traffic, so it appears as a
5865 page within the normal application. It is strongly advised to ensure that the
5866 selected URI will never appear in the application, otherwise it will never be
5867 possible to reach it in the application.
5868
5869 The default URI compiled in haproxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005870 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005871 It is generally a good idea to include a question mark in the URI so that
5872 intermediate proxies refrain from caching the results. Also, since any string
5873 beginning with the prefix will be accepted as a stats request, the question
5874 mark helps ensuring that no valid URI will begin with the same words.
5875
5876 It is sometimes very convenient to use "/" as the URI prefix, and put that
5877 statement in a "listen" instance of its own. That makes it easy to dedicate
5878 an address or a port to statistics only.
5879
5880 Though this statement alone is enough to enable statistics reporting, it is
5881 recommended to set all other settings in order to avoid relying on default
5882 unobvious parameters.
5883
5884 Example :
5885 # public access (limited to this backend only)
5886 backend public_www
5887 server srv1 192.168.0.1:80
5888 stats enable
5889 stats hide-version
5890 stats scope .
5891 stats uri /admin?stats
5892 stats realm Haproxy\ Statistics
5893 stats auth admin1:AdMiN123
5894 stats auth admin2:AdMiN321
5895
5896 # internal monitoring access (unlimited)
5897 backend private_monitoring
5898 stats enable
5899 stats uri /admin?stats
5900 stats refresh 5s
5901
5902 See also : "stats auth", "stats enable", "stats realm"
5903
5904
Willy Tarreaud63335a2010-02-26 12:56:52 +01005905stick match <pattern> [table <table>] [{if | unless} <cond>]
5906 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005907 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +01005908 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01005909
5910 Arguments :
Willy Tarreau74ca5042013-06-11 23:12:07 +02005911 <pattern> is a pattern extraction rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01005912 describes what elements of the incoming request or connection
5913 will be analysed in the hope to find a matching entry in a
5914 stickiness table. This rule is mandatory.
5915
5916 <table> is an optional stickiness table name. If unspecified, the same
5917 backend's table is used. A stickiness table is declared using
5918 the "stick-table" statement.
5919
5920 <cond> is an optional matching condition. It makes it possible to match
5921 on a certain criterion only when other conditions are met (or
5922 not met). For instance, it could be used to match on a source IP
5923 address except when a request passes through a known proxy, in
5924 which case we'd match on a header containing that IP address.
5925
5926 Some protocols or applications require complex stickiness rules and cannot
5927 always simply rely on cookies nor hashing. The "stick match" statement
5928 describes a rule to extract the stickiness criterion from an incoming request
5929 or connection. See section 7 for a complete list of possible patterns and
5930 transformation rules.
5931
5932 The table has to be declared using the "stick-table" statement. It must be of
5933 a type compatible with the pattern. By default it is the one which is present
5934 in the same backend. It is possible to share a table with other backends by
5935 referencing it using the "table" keyword. If another table is referenced,
5936 the server's ID inside the backends are used. By default, all server IDs
5937 start at 1 in each backend, so the server ordering is enough. But in case of
5938 doubt, it is highly recommended to force server IDs using their "id" setting.
5939
5940 It is possible to restrict the conditions where a "stick match" statement
5941 will apply, using "if" or "unless" followed by a condition. See section 7 for
5942 ACL based conditions.
5943
5944 There is no limit on the number of "stick match" statements. The first that
5945 applies and matches will cause the request to be directed to the same server
5946 as was used for the request which created the entry. That way, multiple
5947 matches can be used as fallbacks.
5948
5949 The stick rules are checked after the persistence cookies, so they will not
5950 affect stickiness if a cookie has already been used to select a server. That
5951 way, it becomes very easy to insert cookies and match on IP addresses in
5952 order to maintain stickiness between HTTP and HTTPS.
5953
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01005954 Note : Consider not using this feature in multi-process mode (nbproc > 1)
5955 unless you know what you do : memory is not shared between the
5956 processes, which can result in random behaviours.
5957
Willy Tarreaub937b7e2010-01-12 15:27:54 +01005958 Example :
5959 # forward SMTP users to the same server they just used for POP in the
5960 # last 30 minutes
5961 backend pop
5962 mode tcp
5963 balance roundrobin
5964 stick store-request src
5965 stick-table type ip size 200k expire 30m
5966 server s1 192.168.1.1:110
5967 server s2 192.168.1.1:110
5968
5969 backend smtp
5970 mode tcp
5971 balance roundrobin
5972 stick match src table pop
5973 server s1 192.168.1.1:25
5974 server s2 192.168.1.1:25
5975
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01005976 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
5977 about ACLs and pattern extraction.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01005978
5979
5980stick on <pattern> [table <table>] [{if | unless} <condition>]
5981 Define a request pattern to associate a user to a server
5982 May be used in sections : defaults | frontend | listen | backend
5983 no | no | yes | yes
5984
5985 Note : This form is exactly equivalent to "stick match" followed by
5986 "stick store-request", all with the same arguments. Please refer
5987 to both keywords for details. It is only provided as a convenience
5988 for writing more maintainable configurations.
5989
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01005990 Note : Consider not using this feature in multi-process mode (nbproc > 1)
5991 unless you know what you do : memory is not shared between the
5992 processes, which can result in random behaviours.
5993
Willy Tarreaub937b7e2010-01-12 15:27:54 +01005994 Examples :
5995 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +01005996 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +01005997
5998 # ...is strictly equivalent to this one :
5999 stick match src table pop if !localhost
6000 stick store-request src table pop if !localhost
6001
6002
6003 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
6004 # well as HTTP without cookie. Share the same table between both accesses.
6005 backend http
6006 mode http
6007 balance roundrobin
6008 stick on src table https
6009 cookie SRV insert indirect nocache
6010 server s1 192.168.1.1:80 cookie s1
6011 server s2 192.168.1.1:80 cookie s2
6012
6013 backend https
6014 mode tcp
6015 balance roundrobin
6016 stick-table type ip size 200k expire 30m
6017 stick on src
6018 server s1 192.168.1.1:443
6019 server s2 192.168.1.1:443
6020
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006021 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006022
6023
6024stick store-request <pattern> [table <table>] [{if | unless} <condition>]
6025 Define a request pattern used to create an entry in a stickiness table
6026 May be used in sections : defaults | frontend | listen | backend
6027 no | no | yes | yes
6028
6029 Arguments :
Willy Tarreau74ca5042013-06-11 23:12:07 +02006030 <pattern> is a pattern extraction rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006031 describes what elements of the incoming request or connection
6032 will be analysed, extracted and stored in the table once a
6033 server is selected.
6034
6035 <table> is an optional stickiness table name. If unspecified, the same
6036 backend's table is used. A stickiness table is declared using
6037 the "stick-table" statement.
6038
6039 <cond> is an optional storage condition. It makes it possible to store
6040 certain criteria only when some conditions are met (or not met).
6041 For instance, it could be used to store the source IP address
6042 except when the request passes through a known proxy, in which
6043 case we'd store a converted form of a header containing that IP
6044 address.
6045
6046 Some protocols or applications require complex stickiness rules and cannot
6047 always simply rely on cookies nor hashing. The "stick store-request" statement
6048 describes a rule to decide what to extract from the request and when to do
6049 it, in order to store it into a stickiness table for further requests to
6050 match it using the "stick match" statement. Obviously the extracted part must
6051 make sense and have a chance to be matched in a further request. Storing a
6052 client's IP address for instance often makes sense. Storing an ID found in a
6053 URL parameter also makes sense. Storing a source port will almost never make
6054 any sense because it will be randomly matched. See section 7 for a complete
6055 list of possible patterns and transformation rules.
6056
6057 The table has to be declared using the "stick-table" statement. It must be of
6058 a type compatible with the pattern. By default it is the one which is present
6059 in the same backend. It is possible to share a table with other backends by
6060 referencing it using the "table" keyword. If another table is referenced,
6061 the server's ID inside the backends are used. By default, all server IDs
6062 start at 1 in each backend, so the server ordering is enough. But in case of
6063 doubt, it is highly recommended to force server IDs using their "id" setting.
6064
6065 It is possible to restrict the conditions where a "stick store-request"
6066 statement will apply, using "if" or "unless" followed by a condition. This
6067 condition will be evaluated while parsing the request, so any criteria can be
6068 used. See section 7 for ACL based conditions.
6069
6070 There is no limit on the number of "stick store-request" statements, but
6071 there is a limit of 8 simultaneous stores per request or response. This
6072 makes it possible to store up to 8 criteria, all extracted from either the
6073 request or the response, regardless of the number of rules. Only the 8 first
6074 ones which match will be kept. Using this, it is possible to feed multiple
6075 tables at once in the hope to increase the chance to recognize a user on
6076 another protocol or access method.
6077
6078 The "store-request" rules are evaluated once the server connection has been
6079 established, so that the table will contain the real server that processed
6080 the request.
6081
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006082 Note : Consider not using this feature in multi-process mode (nbproc > 1)
6083 unless you know what you do : memory is not shared between the
6084 processes, which can result in random behaviours.
6085
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006086 Example :
6087 # forward SMTP users to the same server they just used for POP in the
6088 # last 30 minutes
6089 backend pop
6090 mode tcp
6091 balance roundrobin
6092 stick store-request src
6093 stick-table type ip size 200k expire 30m
6094 server s1 192.168.1.1:110
6095 server s2 192.168.1.1:110
6096
6097 backend smtp
6098 mode tcp
6099 balance roundrobin
6100 stick match src table pop
6101 server s1 192.168.1.1:25
6102 server s2 192.168.1.1:25
6103
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006104 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
6105 about ACLs and pattern extraction.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006106
6107
Emeric Brun7c6b82e2010-09-24 16:34:28 +02006108stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Emeric Brunf099e792010-09-27 12:05:28 +02006109 size <size> [expire <expire>] [nopurge] [peers <peersect>]
6110 [store <data_type>]*
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006111 Configure the stickiness table for the current backend
6112 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +02006113 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006114
6115 Arguments :
6116 ip a table declared with "type ip" will only store IPv4 addresses.
6117 This form is very compact (about 50 bytes per entry) and allows
6118 very fast entry lookup and stores with almost no overhead. This
6119 is mainly used to store client source IP addresses.
6120
David du Colombier9a6d3c92011-03-17 10:40:24 +01006121 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
6122 This form is very compact (about 60 bytes per entry) and allows
6123 very fast entry lookup and stores with almost no overhead. This
6124 is mainly used to store client source IP addresses.
6125
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006126 integer a table declared with "type integer" will store 32bit integers
6127 which can represent a client identifier found in a request for
6128 instance.
6129
6130 string a table declared with "type string" will store substrings of up
6131 to <len> characters. If the string provided by the pattern
6132 extractor is larger than <len>, it will be truncated before
6133 being stored. During matching, at most <len> characters will be
6134 compared between the string in the table and the extracted
6135 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +02006136 to 32 characters.
6137
6138 binary a table declared with "type binary" will store binary blocks
6139 of <len> bytes. If the block provided by the pattern
6140 extractor is larger than <len>, it will be truncated before
6141 being stored. If the block provided by the pattern extractor
6142 is shorter than <len>, it will be padded by 0. When not
6143 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006144
6145 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +02006146 "string" type table (See type "string" above). Or the number
6147 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006148 changing this parameter as memory usage will proportionally
6149 increase.
6150
6151 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +01006152 value directly impacts memory usage. Count approximately
6153 50 bytes per entry, plus the size of a string if any. The size
6154 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006155
6156 [nopurge] indicates that we refuse to purge older entries when the table
6157 is full. When not specified and the table is full when haproxy
6158 wants to store an entry in it, it will flush a few of the oldest
6159 entries in order to release some space for the new ones. This is
6160 most often the desired behaviour. In some specific cases, it
6161 be desirable to refuse new entries instead of purging the older
6162 ones. That may be the case when the amount of data to store is
6163 far above the hardware limits and we prefer not to offer access
6164 to new clients than to reject the ones already connected. When
6165 using this parameter, be sure to properly set the "expire"
6166 parameter (see below).
6167
Emeric Brunf099e792010-09-27 12:05:28 +02006168 <peersect> is the name of the peers section to use for replication. Entries
6169 which associate keys to server IDs are kept synchronized with
6170 the remote peers declared in this section. All entries are also
6171 automatically learned from the local peer (old process) during a
6172 soft restart.
6173
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006174 NOTE : peers can't be used in multi-process mode.
6175
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006176 <expire> defines the maximum duration of an entry in the table since it
6177 was last created, refreshed or matched. The expiration delay is
6178 defined using the standard time format, similarly as the various
6179 timeouts. The maximum duration is slightly above 24 days. See
6180 section 2.2 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +02006181 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006182 be removed once full. Be sure not to use the "nopurge" parameter
6183 if not expiration delay is specified.
6184
Willy Tarreau08d5f982010-06-06 13:34:54 +02006185 <data_type> is used to store additional information in the stick-table. This
6186 may be used by ACLs in order to control various criteria related
6187 to the activity of the client matching the stick-table. For each
6188 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +02006189 that the additional data can fit. Several data types may be
6190 stored with an entry. Multiple data types may be specified after
6191 the "store" keyword, as a comma-separated list. Alternatively,
6192 it is possible to repeat the "store" keyword followed by one or
6193 several data types. Except for the "server_id" type which is
6194 automatically detected and enabled, all data types must be
6195 explicitly declared to be stored. If an ACL references a data
6196 type which is not stored, the ACL will simply not match. Some
6197 data types require an argument which must be passed just after
6198 the type between parenthesis. See below for the supported data
6199 types and their arguments.
6200
6201 The data types that can be stored with an entry are the following :
6202 - server_id : this is an integer which holds the numeric ID of the server a
6203 request was assigned to. It is used by the "stick match", "stick store",
6204 and "stick on" rules. It is automatically enabled when referenced.
6205
6206 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
6207 integer which may be used for anything. Most of the time it will be used
6208 to put a special tag on some entries, for instance to note that a
6209 specific behaviour was detected and must be known for future matches.
6210
Willy Tarreauba2ffd12013-05-29 15:54:14 +02006211 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
6212 over a period. It is a positive 32-bit integer integer which may be used
6213 for anything. Just like <gpc0>, it counts events, but instead of keeping
6214 a cumulative count, it maintains the rate at which the counter is
6215 incremented. Most of the time it will be used to measure the frequency of
6216 occurrence of certain events (eg: requests to a specific URL).
6217
Willy Tarreauc9705a12010-07-27 20:05:50 +02006218 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
6219 the absolute number of connections received from clients which matched
6220 this entry. It does not mean the connections were accepted, just that
6221 they were received.
6222
6223 - conn_cur : Current Connections. It is a positive 32-bit integer which
6224 stores the concurrent connection counts for the entry. It is incremented
6225 once an incoming connection matches the entry, and decremented once the
6226 connection leaves. That way it is possible to know at any time the exact
6227 number of concurrent connections for an entry.
6228
6229 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
6230 integer parameter <period> which indicates in milliseconds the length
6231 of the period over which the average is measured. It reports the average
6232 incoming connection rate over that period, in connections per period. The
6233 result is an integer which can be matched using ACLs.
6234
6235 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
6236 the absolute number of sessions received from clients which matched this
6237 entry. A session is a connection that was accepted by the layer 4 rules.
6238
6239 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
6240 integer parameter <period> which indicates in milliseconds the length
6241 of the period over which the average is measured. It reports the average
6242 incoming session rate over that period, in sessions per period. The
6243 result is an integer which can be matched using ACLs.
6244
6245 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
6246 counts the absolute number of HTTP requests received from clients which
6247 matched this entry. It does not matter whether they are valid requests or
6248 not. Note that this is different from sessions when keep-alive is used on
6249 the client side.
6250
6251 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
6252 integer parameter <period> which indicates in milliseconds the length
6253 of the period over which the average is measured. It reports the average
6254 HTTP request rate over that period, in requests per period. The result is
6255 an integer which can be matched using ACLs. It does not matter whether
6256 they are valid requests or not. Note that this is different from sessions
6257 when keep-alive is used on the client side.
6258
6259 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
6260 counts the absolute number of HTTP requests errors induced by clients
6261 which matched this entry. Errors are counted on invalid and truncated
6262 requests, as well as on denied or tarpitted requests, and on failed
6263 authentications. If the server responds with 4xx, then the request is
6264 also counted as an error since it's an error triggered by the client
6265 (eg: vulnerability scan).
6266
6267 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
6268 integer parameter <period> which indicates in milliseconds the length
6269 of the period over which the average is measured. It reports the average
6270 HTTP request error rate over that period, in requests per period (see
6271 http_err_cnt above for what is accounted as an error). The result is an
6272 integer which can be matched using ACLs.
6273
6274 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
6275 integer which counts the cumulated amount of bytes received from clients
6276 which matched this entry. Headers are included in the count. This may be
6277 used to limit abuse of upload features on photo or video servers.
6278
6279 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
6280 integer parameter <period> which indicates in milliseconds the length
6281 of the period over which the average is measured. It reports the average
6282 incoming bytes rate over that period, in bytes per period. It may be used
6283 to detect users which upload too much and too fast. Warning: with large
6284 uploads, it is possible that the amount of uploaded data will be counted
6285 once upon termination, thus causing spikes in the average transfer speed
6286 instead of having a smooth one. This may partially be smoothed with
6287 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
6288 recommended for better fairness.
6289
6290 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
6291 integer which counts the cumulated amount of bytes sent to clients which
6292 matched this entry. Headers are included in the count. This may be used
6293 to limit abuse of bots sucking the whole site.
6294
6295 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
6296 an integer parameter <period> which indicates in milliseconds the length
6297 of the period over which the average is measured. It reports the average
6298 outgoing bytes rate over that period, in bytes per period. It may be used
6299 to detect users which download too much and too fast. Warning: with large
6300 transfers, it is possible that the amount of transferred data will be
6301 counted once upon termination, thus causing spikes in the average
6302 transfer speed instead of having a smooth one. This may partially be
6303 smoothed with "option contstats" though this is not perfect yet. Use of
6304 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +02006305
Willy Tarreauc00cdc22010-06-06 16:48:26 +02006306 There is only one stick-table per proxy. At the moment of writing this doc,
6307 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006308 to be required, simply create a dummy backend with a stick-table in it and
6309 reference it.
6310
6311 It is important to understand that stickiness based on learning information
6312 has some limitations, including the fact that all learned associations are
6313 lost upon restart. In general it can be good as a complement but not always
6314 as an exclusive stickiness.
6315
Willy Tarreauc9705a12010-07-27 20:05:50 +02006316 Last, memory requirements may be important when storing many data types.
6317 Indeed, storing all indicators above at once in each entry requires 116 bytes
6318 per entry, or 116 MB for a 1-million entries table. This is definitely not
6319 something that can be ignored.
6320
6321 Example:
6322 # Keep track of counters of up to 1 million IP addresses over 5 minutes
6323 # and store a general purpose counter and the average connection rate
6324 # computed over a sliding window of 30 seconds.
6325 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
6326
6327 See also : "stick match", "stick on", "stick store-request", section 2.2
David du Colombiera13d1b92011-03-17 10:40:22 +01006328 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006329
6330
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006331stick store-response <pattern> [table <table>] [{if | unless} <condition>]
6332 Define a request pattern used to create an entry in a stickiness table
6333 May be used in sections : defaults | frontend | listen | backend
6334 no | no | yes | yes
6335
6336 Arguments :
Willy Tarreau74ca5042013-06-11 23:12:07 +02006337 <pattern> is a pattern extraction rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006338 describes what elements of the response or connection will
6339 be analysed, extracted and stored in the table once a
6340 server is selected.
6341
6342 <table> is an optional stickiness table name. If unspecified, the same
6343 backend's table is used. A stickiness table is declared using
6344 the "stick-table" statement.
6345
6346 <cond> is an optional storage condition. It makes it possible to store
6347 certain criteria only when some conditions are met (or not met).
6348 For instance, it could be used to store the SSL session ID only
6349 when the response is a SSL server hello.
6350
6351 Some protocols or applications require complex stickiness rules and cannot
6352 always simply rely on cookies nor hashing. The "stick store-response"
6353 statement describes a rule to decide what to extract from the response and
6354 when to do it, in order to store it into a stickiness table for further
6355 requests to match it using the "stick match" statement. Obviously the
6356 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006357 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006358 See section 7 for a complete list of possible patterns and transformation
6359 rules.
6360
6361 The table has to be declared using the "stick-table" statement. It must be of
6362 a type compatible with the pattern. By default it is the one which is present
6363 in the same backend. It is possible to share a table with other backends by
6364 referencing it using the "table" keyword. If another table is referenced,
6365 the server's ID inside the backends are used. By default, all server IDs
6366 start at 1 in each backend, so the server ordering is enough. But in case of
6367 doubt, it is highly recommended to force server IDs using their "id" setting.
6368
6369 It is possible to restrict the conditions where a "stick store-response"
6370 statement will apply, using "if" or "unless" followed by a condition. This
6371 condition will be evaluated while parsing the response, so any criteria can
6372 be used. See section 7 for ACL based conditions.
6373
6374 There is no limit on the number of "stick store-response" statements, but
6375 there is a limit of 8 simultaneous stores per request or response. This
6376 makes it possible to store up to 8 criteria, all extracted from either the
6377 request or the response, regardless of the number of rules. Only the 8 first
6378 ones which match will be kept. Using this, it is possible to feed multiple
6379 tables at once in the hope to increase the chance to recognize a user on
6380 another protocol or access method.
6381
6382 The table will contain the real server that processed the request.
6383
6384 Example :
6385 # Learn SSL session ID from both request and response and create affinity.
6386 backend https
6387 mode tcp
6388 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +02006389 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006390 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006391
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006392 acl clienthello req_ssl_hello_type 1
6393 acl serverhello rep_ssl_hello_type 2
6394
6395 # use tcp content accepts to detects ssl client and server hello.
6396 tcp-request inspect-delay 5s
6397 tcp-request content accept if clienthello
6398
6399 # no timeout on response inspect delay by default.
6400 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006401
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006402 # SSL session ID (SSLID) may be present on a client or server hello.
6403 # Its length is coded on 1 byte at offset 43 and its value starts
6404 # at offset 44.
6405
6406 # Match and learn on request if client hello.
6407 stick on payload_lv(43,1) if clienthello
6408
6409 # Learn on response if server hello.
6410 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +02006411
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006412 server s1 192.168.1.1:443
6413 server s2 192.168.1.1:443
6414
6415 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
6416 extraction.
6417
6418
Willy Tarreaue9656522010-08-17 15:40:09 +02006419tcp-request connection <action> [{if | unless} <condition>]
6420 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +02006421 May be used in sections : defaults | frontend | listen | backend
6422 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +02006423 Arguments :
6424 <action> defines the action to perform if the condition applies. Valid
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006425 actions include : "accept", "reject", "track-sc0", "track-sc1",
6426 "track-sc2", and "expect-proxy". See below for more details.
Willy Tarreau1a687942010-05-23 22:40:30 +02006427
Willy Tarreaue9656522010-08-17 15:40:09 +02006428 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006429
6430 Immediately after acceptance of a new incoming connection, it is possible to
6431 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +02006432 or dropped or have its counters tracked. Those conditions cannot make use of
6433 any data contents because the connection has not been read from yet, and the
6434 buffers are not yet allocated. This is used to selectively and very quickly
6435 accept or drop connections from various sources with a very low overhead. If
6436 some contents need to be inspected in order to take the decision, the
6437 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006438
Willy Tarreaue9656522010-08-17 15:40:09 +02006439 The "tcp-request connection" rules are evaluated in their exact declaration
6440 order. If no rule matches or if there is no rule, the default action is to
6441 accept the incoming connection. There is no specific limit to the number of
6442 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006443
Willy Tarreaue9656522010-08-17 15:40:09 +02006444 Three types of actions are supported :
6445 - accept :
6446 accepts the connection if the condition is true (when used with "if")
6447 or false (when used with "unless"). The first such rule executed ends
6448 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006449
Willy Tarreaue9656522010-08-17 15:40:09 +02006450 - reject :
6451 rejects the connection if the condition is true (when used with "if")
6452 or false (when used with "unless"). The first such rule executed ends
6453 the rules evaluation. Rejected connections do not even become a
6454 session, which is why they are accounted separately for in the stats,
6455 as "denied connections". They are not considered for the session
6456 rate-limit and are not logged either. The reason is that these rules
6457 should only be used to filter extremely high connection rates such as
6458 the ones encountered during a massive DDoS attack. Under these extreme
6459 conditions, the simple action of logging each event would make the
6460 system collapse and would considerably lower the filtering capacity. If
6461 logging is absolutely desired, then "tcp-request content" rules should
6462 be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006463
Willy Tarreau4f0d9192013-06-11 20:40:55 +02006464 - expect-proxy layer4 :
6465 configures the client-facing connection to receive a PROXY protocol
6466 header before any byte is read from the socket. This is equivalent to
6467 having the "accept-proxy" keyword on the "bind" line, except that using
6468 the TCP rule allows the PROXY protocol to be accepted only for certain
6469 IP address ranges using an ACL. This is convenient when multiple layers
6470 of load balancers are passed through by traffic coming from public
6471 hosts.
6472
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006473 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +02006474 enables tracking of sticky counters from current connection. These
6475 rules do not stop evaluation and do not change default action. Two sets
6476 of counters may be simultaneously tracked by the same connection. The
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006477 first "track-sc0" rule executed enables tracking of the counters of the
6478 specified table as the first set. The first "track-sc1" rule executed
Willy Tarreaue9656522010-08-17 15:40:09 +02006479 enables tracking of the counters of the specified table as the second
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006480 set. The first "track-sc2" rule executed enables tracking of the
6481 counters of the specified table as the third set. It is a recommended
6482 practice to use the first set of counters for the per-frontend counters
6483 and the second set for the per-backend ones. But this is just a
6484 guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006485
Willy Tarreaue9656522010-08-17 15:40:09 +02006486 These actions take one or two arguments :
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006487 <key> is mandatory, and is a pattern extraction rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +02006488 in section 7.3. It describes what elements of the incoming
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006489 request or connection will be analysed, extracted, combined,
6490 and used to select which table entry to update the counters.
6491 Note that "tcp-request connection" cannot use content-based
6492 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006493
Willy Tarreaue9656522010-08-17 15:40:09 +02006494 <table> is an optional table to be used instead of the default one,
6495 which is the stick-table declared in the current proxy. All
6496 the counters for the matches and updates for the key will
6497 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006498
Willy Tarreaue9656522010-08-17 15:40:09 +02006499 Once a "track-sc*" rule is executed, the key is looked up in the table
6500 and if it is not found, an entry is allocated for it. Then a pointer to
6501 that entry is kept during all the session's life, and this entry's
6502 counters are updated as often as possible, every time the session's
6503 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006504 Counters are only updated for events that happen after the tracking has
6505 been started. For example, connection counters will not be updated when
6506 tracking layer 7 information, since the connection event happens before
6507 layer7 information is extracted.
6508
Willy Tarreaue9656522010-08-17 15:40:09 +02006509 If the entry tracks concurrent connection counters, one connection is
6510 counted for as long as the entry is tracked, and the entry will not
6511 expire during that time. Tracking counters also provides a performance
6512 advantage over just checking the keys, because only one table lookup is
6513 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006514
Willy Tarreaue9656522010-08-17 15:40:09 +02006515 Note that the "if/unless" condition is optional. If no condition is set on
6516 the action, it is simply performed unconditionally. That can be useful for
6517 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006518
Willy Tarreaue9656522010-08-17 15:40:09 +02006519 Example: accept all connections from white-listed hosts, reject too fast
6520 connection without counting them, and track accepted connections.
6521 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006522
Willy Tarreaue9656522010-08-17 15:40:09 +02006523 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006524 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006525 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006526
Willy Tarreaue9656522010-08-17 15:40:09 +02006527 Example: accept all connections from white-listed hosts, count all other
6528 connections and reject too fast ones. This results in abusive ones
6529 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006530
Willy Tarreaue9656522010-08-17 15:40:09 +02006531 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006532 tcp-request connection track-sc0 src
6533 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006534
Willy Tarreau4f0d9192013-06-11 20:40:55 +02006535 Example: enable the PROXY protocol for traffic coming from all known proxies.
6536
6537 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
6538
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006539 See section 7 about ACL usage.
6540
Willy Tarreaue9656522010-08-17 15:40:09 +02006541 See also : "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006542
6543
Willy Tarreaue9656522010-08-17 15:40:09 +02006544tcp-request content <action> [{if | unless} <condition>]
6545 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +02006546 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02006547 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +02006548 Arguments :
6549 <action> defines the action to perform if the condition applies. Valid
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006550 actions include : "accept", "reject", "track-sc0", "track-sc1",
6551 and "track-sc2". See "tcp-request connection" above for their
Willy Tarreaue25c9172013-05-28 18:32:20 +02006552 signification.
Willy Tarreau62644772008-07-16 18:36:06 +02006553
Willy Tarreaue9656522010-08-17 15:40:09 +02006554 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +02006555
Willy Tarreaue9656522010-08-17 15:40:09 +02006556 A request's contents can be analysed at an early stage of request processing
6557 called "TCP content inspection". During this stage, ACL-based rules are
6558 evaluated every time the request contents are updated, until either an
6559 "accept" or a "reject" rule matches, or the TCP request inspection delay
6560 expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +02006561
Willy Tarreaue9656522010-08-17 15:40:09 +02006562 The first difference between these rules and "tcp-request connection" rules
6563 is that "tcp-request content" rules can make use of contents to take a
6564 decision. Most often, these decisions will consider a protocol recognition or
6565 validity. The second difference is that content-based rules can be used in
6566 both frontends and backends. In frontends, they will be evaluated upon new
6567 connections. In backends, they will be evaluated once a session is assigned
6568 a backend. This means that a single frontend connection may be evaluated
6569 several times by one or multiple backends when a session gets reassigned
6570 (for instance after a client-side HTTP keep-alive request).
Willy Tarreau62644772008-07-16 18:36:06 +02006571
Willy Tarreaue9656522010-08-17 15:40:09 +02006572 Content-based rules are evaluated in their exact declaration order. If no
6573 rule matches or if there is no rule, the default action is to accept the
6574 contents. There is no specific limit to the number of rules which may be
6575 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +02006576
Willy Tarreaue9656522010-08-17 15:40:09 +02006577 Three types of actions are supported :
6578 - accept :
6579 - reject :
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006580 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Willy Tarreau62644772008-07-16 18:36:06 +02006581
Willy Tarreaue9656522010-08-17 15:40:09 +02006582 They have the same meaning as their counter-parts in "tcp-request connection"
6583 so please refer to that section for a complete description.
Willy Tarreau62644772008-07-16 18:36:06 +02006584
Willy Tarreaue9656522010-08-17 15:40:09 +02006585 Also, it is worth noting that if sticky counters are tracked from a rule
6586 defined in a backend, this tracking will automatically end when the session
6587 releases the backend. That allows per-backend counter tracking even in case
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006588 of HTTP keep-alive requests when the backend changes. This makes a subtle
6589 difference because tracking rules in "frontend" and "listen" section last for
6590 all the session, as opposed to the backend rules. The difference appears when
6591 some layer 7 information is tracked. While there is nothing mandatory about
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006592 it, it is recommended to use the track-sc0 pointer to track per-frontend
6593 counters and track-sc1 to track per-backend counters, but this is just a
Willy Tarreaue25c9172013-05-28 18:32:20 +02006594 guideline and all counters may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +02006595
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006596 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +02006597 the action, it is simply performed unconditionally. That can be useful for
6598 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +02006599
Willy Tarreaue9656522010-08-17 15:40:09 +02006600 It is perfectly possible to match layer 7 contents with "tcp-request content"
Willy Tarreauc0239e02012-04-16 14:42:55 +02006601 rules, since HTTP-specific ACL matches are able to preliminarily parse the
6602 contents of a buffer before extracting the required data. If the buffered
6603 contents do not parse as a valid HTTP message, then the ACL does not match.
6604 The parser which is involved there is exactly the same as for all other HTTP
6605 processing, so there is no risk of parsing something differently.
Willy Tarreau62644772008-07-16 18:36:06 +02006606
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006607 Tracking layer7 information is also possible provided that the information
6608 are present when the rule is processed. The current solution for making the
6609 rule engine wait for such information is to set an inspect delay and to
6610 condition its execution with an ACL relying on such information.
6611
Willy Tarreau62644772008-07-16 18:36:06 +02006612 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +02006613 # Accept HTTP requests containing a Host header saying "example.com"
6614 # and reject everything else.
6615 acl is_host_com hdr(Host) -i example.com
6616 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +02006617 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +02006618 tcp-request content reject
6619
6620 Example:
Willy Tarreau62644772008-07-16 18:36:06 +02006621 # reject SMTP connection if client speaks first
6622 tcp-request inspect-delay 30s
6623 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006624 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +02006625
6626 # Forward HTTPS connection only if client speaks
6627 tcp-request inspect-delay 30s
6628 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006629 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +02006630 tcp-request content reject
6631
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006632 Example:
6633 # Track the last IP from X-Forwarded-For
6634 tcp-request inspect-delay 10s
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006635 tcp-request content track-sc0 hdr(x-forwarded-for,-1) if HTTP
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006636
6637 Example:
6638 # track request counts per "base" (concatenation of Host+URL)
6639 tcp-request inspect-delay 10s
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006640 tcp-request content track-sc0 base table req-rate if HTTP
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006641
Willy Tarreaue9656522010-08-17 15:40:09 +02006642 Example: track per-frontend and per-backend counters, block abusers at the
6643 frontend when the backend detects abuse.
6644
6645 frontend http
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006646 # Use General Purpose Couter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +02006647 # protecting all our sites
6648 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006649 tcp-request connection track-sc0 src
6650 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +02006651 ...
6652 use_backend http_dynamic if { path_end .php }
6653
6654 backend http_dynamic
6655 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006656 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +02006657 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006658 acl click_too_fast sc1_http_req_rate gt 10
6659 acl mark_as_abuser sc0_inc_gpc0 gt 0
6660 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +02006661 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +02006662
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006663 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +02006664
Willy Tarreaue9656522010-08-17 15:40:09 +02006665 See also : "tcp-request connection", "tcp-request inspect-delay"
Willy Tarreau62644772008-07-16 18:36:06 +02006666
6667
6668tcp-request inspect-delay <timeout>
6669 Set the maximum allowed time to wait for data during content inspection
6670 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02006671 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +02006672 Arguments :
6673 <timeout> is the timeout value specified in milliseconds by default, but
6674 can be in any other unit if the number is suffixed by the unit,
6675 as explained at the top of this document.
6676
6677 People using haproxy primarily as a TCP relay are often worried about the
6678 risk of passing any type of protocol to a server without any analysis. In
6679 order to be able to analyze the request contents, we must first withhold
6680 the data then analyze them. This statement simply enables withholding of
6681 data for at most the specified amount of time.
6682
Willy Tarreaufb356202010-08-03 14:02:05 +02006683 TCP content inspection applies very early when a connection reaches a
6684 frontend, then very early when the connection is forwarded to a backend. This
6685 means that a connection may experience a first delay in the frontend and a
6686 second delay in the backend if both have tcp-request rules.
6687
Willy Tarreau62644772008-07-16 18:36:06 +02006688 Note that when performing content inspection, haproxy will evaluate the whole
6689 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006690 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +02006691 a last check is performed upon expiration, this time considering that the
Willy Tarreaud869b242009-03-15 14:43:58 +01006692 contents are definitive. If no delay is set, haproxy will not wait at all
6693 and will immediately apply a verdict based on the available information.
6694 Obviously this is unlikely to be very useful and might even be racy, so such
6695 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +02006696
6697 As soon as a rule matches, the request is released and continues as usual. If
6698 the timeout is reached and no rule matches, the default policy will be to let
6699 it pass through unaffected.
6700
6701 For most protocols, it is enough to set it to a few seconds, as most clients
6702 send the full request immediately upon connection. Add 3 or more seconds to
6703 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +01006704 to use large values, for instance to ensure that the client never talks
Willy Tarreau62644772008-07-16 18:36:06 +02006705 before the server (eg: SMTP), or to wait for a client to talk before passing
6706 data to the server (eg: SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +02006707 least the inspection delay, otherwise it will expire first. If the client
6708 closes the connection or if the buffer is full, the delay immediately expires
6709 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +02006710
Willy Tarreau55165fe2009-05-10 12:02:55 +02006711 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +02006712 "timeout client".
6713
6714
Emeric Brun0a3b67f2010-09-24 15:34:53 +02006715tcp-response content <action> [{if | unless} <condition>]
6716 Perform an action on a session response depending on a layer 4-7 condition
6717 May be used in sections : defaults | frontend | listen | backend
6718 no | no | yes | yes
6719 Arguments :
6720 <action> defines the action to perform if the condition applies. Valid
6721 actions include : "accept", "reject".
6722 See "tcp-request connection" above for their signification.
6723
6724 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
6725
6726 Response contents can be analysed at an early stage of response processing
6727 called "TCP content inspection". During this stage, ACL-based rules are
6728 evaluated every time the response contents are updated, until either an
6729 "accept" or a "reject" rule matches, or a TCP response inspection delay is
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006730 set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02006731
6732 Most often, these decisions will consider a protocol recognition or validity.
6733
6734 Content-based rules are evaluated in their exact declaration order. If no
6735 rule matches or if there is no rule, the default action is to accept the
6736 contents. There is no specific limit to the number of rules which may be
6737 inserted.
6738
6739 Two types of actions are supported :
6740 - accept :
6741 accepts the response if the condition is true (when used with "if")
6742 or false (when used with "unless"). The first such rule executed ends
6743 the rules evaluation.
6744
6745 - reject :
6746 rejects the response if the condition is true (when used with "if")
6747 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -04006748 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02006749
6750 Note that the "if/unless" condition is optional. If no condition is set on
6751 the action, it is simply performed unconditionally. That can be useful for
6752 for changing the default action to a reject.
6753
Jamie Gloudonaaa21002012-08-25 00:18:33 -04006754 It is perfectly possible to match layer 7 contents with "tcp-response
6755 content" rules, but then it is important to ensure that a full response has
6756 been buffered, otherwise no contents will match. In order to achieve this,
6757 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +02006758 period.
6759
6760 See section 7 about ACL usage.
6761
6762 See also : "tcp-request content", "tcp-response inspect-delay"
6763
6764
6765tcp-response inspect-delay <timeout>
6766 Set the maximum allowed time to wait for a response during content inspection
6767 May be used in sections : defaults | frontend | listen | backend
6768 no | no | yes | yes
6769 Arguments :
6770 <timeout> is the timeout value specified in milliseconds by default, but
6771 can be in any other unit if the number is suffixed by the unit,
6772 as explained at the top of this document.
6773
6774 See also : "tcp-response content", "tcp-request inspect-delay".
6775
6776
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01006777timeout check <timeout>
6778 Set additional check timeout, but only after a connection has been already
6779 established.
6780
6781 May be used in sections: defaults | frontend | listen | backend
6782 yes | no | yes | yes
6783 Arguments:
6784 <timeout> is the timeout value specified in milliseconds by default, but
6785 can be in any other unit if the number is suffixed by the unit,
6786 as explained at the top of this document.
6787
6788 If set, haproxy uses min("timeout connect", "inter") as a connect timeout
6789 for check and "timeout check" as an additional read timeout. The "min" is
6790 used so that people running with *very* long "timeout connect" (eg. those
6791 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +01006792 (Please also note that there is no valid reason to have such long connect
6793 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
6794 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01006795
6796 If "timeout check" is not set haproxy uses "inter" for complete check
6797 timeout (connect + read) exactly like all <1.3.15 version.
6798
6799 In most cases check request is much simpler and faster to handle than normal
6800 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +01006801 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01006802
6803 This parameter is specific to backends, but can be specified once for all in
6804 "defaults" sections. This is in fact one of the easiest solutions not to
6805 forget about it.
6806
Willy Tarreau41a340d2008-01-22 12:25:31 +01006807 See also: "timeout connect", "timeout queue", "timeout server",
6808 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01006809
6810
Willy Tarreau0ba27502007-12-24 16:55:16 +01006811timeout client <timeout>
6812timeout clitimeout <timeout> (deprecated)
6813 Set the maximum inactivity time on the client side.
6814 May be used in sections : defaults | frontend | listen | backend
6815 yes | yes | yes | no
6816 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01006817 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01006818 can be in any other unit if the number is suffixed by the unit,
6819 as explained at the top of this document.
6820
6821 The inactivity timeout applies when the client is expected to acknowledge or
6822 send data. In HTTP mode, this timeout is particularly important to consider
6823 during the first phase, when the client sends the request, and during the
6824 response while it is reading data sent by the server. The value is specified
6825 in milliseconds by default, but can be in any other unit if the number is
6826 suffixed by the unit, as specified at the top of this document. In TCP mode
6827 (and to a lesser extent, in HTTP mode), it is highly recommended that the
6828 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006829 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01006830 losses by specifying timeouts that are slightly above multiples of 3 seconds
Willy Tarreauce887fd2012-05-12 12:50:00 +02006831 (eg: 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
6832 sessions (eg: WebSocket and HTTP), it's worth considering "timeout tunnel",
6833 which overrides "timeout client" and "timeout server" for tunnels.
Willy Tarreau0ba27502007-12-24 16:55:16 +01006834
6835 This parameter is specific to frontends, but can be specified once for all in
6836 "defaults" sections. This is in fact one of the easiest solutions not to
6837 forget about it. An unspecified timeout results in an infinite timeout, which
6838 is not recommended. Such a usage is accepted and works but reports a warning
6839 during startup because it may results in accumulation of expired sessions in
6840 the system if the system's timeouts are not configured either.
6841
6842 This parameter replaces the old, deprecated "clitimeout". It is recommended
6843 to use it to write new configurations. The form "timeout clitimeout" is
6844 provided only by backwards compatibility but its use is strongly discouraged.
6845
Willy Tarreauce887fd2012-05-12 12:50:00 +02006846 See also : "clitimeout", "timeout server", "timeout tunnel".
Willy Tarreau0ba27502007-12-24 16:55:16 +01006847
6848
6849timeout connect <timeout>
6850timeout contimeout <timeout> (deprecated)
6851 Set the maximum time to wait for a connection attempt to a server to succeed.
6852 May be used in sections : defaults | frontend | listen | backend
6853 yes | no | yes | yes
6854 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01006855 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01006856 can be in any other unit if the number is suffixed by the unit,
6857 as explained at the top of this document.
6858
6859 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006860 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01006861 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01006862 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01006863 connect timeout also presets both queue and tarpit timeouts to the same value
6864 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +01006865
6866 This parameter is specific to backends, but can be specified once for all in
6867 "defaults" sections. This is in fact one of the easiest solutions not to
6868 forget about it. An unspecified timeout results in an infinite timeout, which
6869 is not recommended. Such a usage is accepted and works but reports a warning
6870 during startup because it may results in accumulation of failed sessions in
6871 the system if the system's timeouts are not configured either.
6872
6873 This parameter replaces the old, deprecated "contimeout". It is recommended
6874 to use it to write new configurations. The form "timeout contimeout" is
6875 provided only by backwards compatibility but its use is strongly discouraged.
6876
Willy Tarreau41a340d2008-01-22 12:25:31 +01006877 See also: "timeout check", "timeout queue", "timeout server", "contimeout",
6878 "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +01006879
6880
Willy Tarreaub16a5742010-01-10 14:46:16 +01006881timeout http-keep-alive <timeout>
6882 Set the maximum allowed time to wait for a new HTTP request to appear
6883 May be used in sections : defaults | frontend | listen | backend
6884 yes | yes | yes | yes
6885 Arguments :
6886 <timeout> is the timeout value specified in milliseconds by default, but
6887 can be in any other unit if the number is suffixed by the unit,
6888 as explained at the top of this document.
6889
6890 By default, the time to wait for a new request in case of keep-alive is set
6891 by "timeout http-request". However this is not always convenient because some
6892 people want very short keep-alive timeouts in order to release connections
6893 faster, and others prefer to have larger ones but still have short timeouts
6894 once the request has started to present itself.
6895
6896 The "http-keep-alive" timeout covers these needs. It will define how long to
6897 wait for a new HTTP request to start coming after a response was sent. Once
6898 the first byte of request has been seen, the "http-request" timeout is used
6899 to wait for the complete request to come. Note that empty lines prior to a
6900 new request do not refresh the timeout and are not counted as a new request.
6901
6902 There is also another difference between the two timeouts : when a connection
6903 expires during timeout http-keep-alive, no error is returned, the connection
6904 just closes. If the connection expires in "http-request" while waiting for a
6905 connection to complete, a HTTP 408 error is returned.
6906
6907 In general it is optimal to set this value to a few tens to hundreds of
6908 milliseconds, to allow users to fetch all objects of a page at once but
6909 without waiting for further clicks. Also, if set to a very small value (eg:
6910 1 millisecond) it will probably only accept pipelined requests but not the
6911 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +02006912 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +01006913
6914 If this parameter is not set, the "http-request" timeout applies, and if both
6915 are not set, "timeout client" still applies at the lower level. It should be
6916 set in the frontend to take effect, unless the frontend is in TCP mode, in
6917 which case the HTTP backend's timeout will be used.
6918
6919 See also : "timeout http-request", "timeout client".
6920
6921
Willy Tarreau036fae02008-01-06 13:24:40 +01006922timeout http-request <timeout>
6923 Set the maximum allowed time to wait for a complete HTTP request
6924 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +02006925 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +01006926 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01006927 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +01006928 can be in any other unit if the number is suffixed by the unit,
6929 as explained at the top of this document.
6930
6931 In order to offer DoS protection, it may be required to lower the maximum
6932 accepted time to receive a complete HTTP request without affecting the client
6933 timeout. This helps protecting against established connections on which
6934 nothing is sent. The client timeout cannot offer a good protection against
6935 this abuse because it is an inactivity timeout, which means that if the
6936 attacker sends one character every now and then, the timeout will not
6937 trigger. With the HTTP request timeout, no matter what speed the client
6938 types, the request will be aborted if it does not complete in time.
6939
6940 Note that this timeout only applies to the header part of the request, and
6941 not to any data. As soon as the empty line is received, this timeout is not
Willy Tarreaub16a5742010-01-10 14:46:16 +01006942 used anymore. It is used again on keep-alive connections to wait for a second
6943 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +01006944
6945 Generally it is enough to set it to a few seconds, as most clients send the
6946 full request immediately upon connection. Add 3 or more seconds to cover TCP
6947 retransmits but that's all. Setting it to very low values (eg: 50 ms) will
6948 generally work on local networks as long as there are no packet losses. This
6949 will prevent people from sending bare HTTP requests using telnet.
6950
6951 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +02006952 chunk of the incoming request. It should be set in the frontend to take
6953 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
6954 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +01006955
Willy Tarreaub16a5742010-01-10 14:46:16 +01006956 See also : "timeout http-keep-alive", "timeout client".
Willy Tarreau036fae02008-01-06 13:24:40 +01006957
Willy Tarreau844e3c52008-01-11 16:28:18 +01006958
6959timeout queue <timeout>
6960 Set the maximum time to wait in the queue for a connection slot to be free
6961 May be used in sections : defaults | frontend | listen | backend
6962 yes | no | yes | yes
6963 Arguments :
6964 <timeout> is the timeout value specified in milliseconds by default, but
6965 can be in any other unit if the number is suffixed by the unit,
6966 as explained at the top of this document.
6967
6968 When a server's maxconn is reached, connections are left pending in a queue
6969 which may be server-specific or global to the backend. In order not to wait
6970 indefinitely, a timeout is applied to requests pending in the queue. If the
6971 timeout is reached, it is considered that the request will almost never be
6972 served, so it is dropped and a 503 error is returned to the client.
6973
6974 The "timeout queue" statement allows to fix the maximum time for a request to
6975 be left pending in a queue. If unspecified, the same value as the backend's
6976 connection timeout ("timeout connect") is used, for backwards compatibility
6977 with older versions with no "timeout queue" parameter.
6978
6979 See also : "timeout connect", "contimeout".
6980
6981
6982timeout server <timeout>
6983timeout srvtimeout <timeout> (deprecated)
6984 Set the maximum inactivity time on the server side.
6985 May be used in sections : defaults | frontend | listen | backend
6986 yes | no | yes | yes
6987 Arguments :
6988 <timeout> is the timeout value specified in milliseconds by default, but
6989 can be in any other unit if the number is suffixed by the unit,
6990 as explained at the top of this document.
6991
6992 The inactivity timeout applies when the server is expected to acknowledge or
6993 send data. In HTTP mode, this timeout is particularly important to consider
6994 during the first phase of the server's response, when it has to send the
6995 headers, as it directly represents the server's processing time for the
6996 request. To find out what value to put there, it's often good to start with
6997 what would be considered as unacceptable response times, then check the logs
6998 to observe the response time distribution, and adjust the value accordingly.
6999
7000 The value is specified in milliseconds by default, but can be in any other
7001 unit if the number is suffixed by the unit, as specified at the top of this
7002 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
7003 recommended that the client timeout remains equal to the server timeout in
7004 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007005 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01007006 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreauce887fd2012-05-12 12:50:00 +02007007 seconds (eg: 4 or 5 seconds minimum). If some long-lived sessions are mixed
7008 with short-lived sessions (eg: WebSocket and HTTP), it's worth considering
7009 "timeout tunnel", which overrides "timeout client" and "timeout server" for
7010 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007011
7012 This parameter is specific to backends, but can be specified once for all in
7013 "defaults" sections. This is in fact one of the easiest solutions not to
7014 forget about it. An unspecified timeout results in an infinite timeout, which
7015 is not recommended. Such a usage is accepted and works but reports a warning
7016 during startup because it may results in accumulation of expired sessions in
7017 the system if the system's timeouts are not configured either.
7018
7019 This parameter replaces the old, deprecated "srvtimeout". It is recommended
7020 to use it to write new configurations. The form "timeout srvtimeout" is
7021 provided only by backwards compatibility but its use is strongly discouraged.
7022
Willy Tarreauce887fd2012-05-12 12:50:00 +02007023 See also : "srvtimeout", "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +01007024
7025
7026timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +01007027 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +01007028 May be used in sections : defaults | frontend | listen | backend
7029 yes | yes | yes | yes
7030 Arguments :
7031 <timeout> is the tarpit duration specified in milliseconds by default, but
7032 can be in any other unit if the number is suffixed by the unit,
7033 as explained at the top of this document.
7034
7035 When a connection is tarpitted using "reqtarpit", it is maintained open with
7036 no activity for a certain amount of time, then closed. "timeout tarpit"
7037 defines how long it will be maintained open.
7038
7039 The value is specified in milliseconds by default, but can be in any other
7040 unit if the number is suffixed by the unit, as specified at the top of this
7041 document. If unspecified, the same value as the backend's connection timeout
7042 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +01007043 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007044
7045 See also : "timeout connect", "contimeout".
7046
7047
Willy Tarreauce887fd2012-05-12 12:50:00 +02007048timeout tunnel <timeout>
7049 Set the maximum inactivity time on the client and server side for tunnels.
7050 May be used in sections : defaults | frontend | listen | backend
7051 yes | no | yes | yes
7052 Arguments :
7053 <timeout> is the timeout value specified in milliseconds by default, but
7054 can be in any other unit if the number is suffixed by the unit,
7055 as explained at the top of this document.
7056
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007057 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +02007058 between a client and a server, and the connection remains inactive in both
7059 directions. This timeout supersedes both the client and server timeouts once
7060 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
7061 analyser remains attached to either connection (eg: tcp content rules are
7062 accepted). In HTTP, this timeout is used when a connection is upgraded (eg:
7063 when switching to the WebSocket protocol, or forwarding a CONNECT request
7064 to a proxy), or after the first response when no keepalive/close option is
7065 specified.
7066
7067 The value is specified in milliseconds by default, but can be in any other
7068 unit if the number is suffixed by the unit, as specified at the top of this
7069 document. Whatever the expected normal idle time, it is a good practice to
7070 cover at least one or several TCP packet losses by specifying timeouts that
7071 are slightly above multiples of 3 seconds (eg: 4 or 5 seconds minimum).
7072
7073 This parameter is specific to backends, but can be specified once for all in
7074 "defaults" sections. This is in fact one of the easiest solutions not to
7075 forget about it.
7076
7077 Example :
7078 defaults http
7079 option http-server-close
7080 timeout connect 5s
7081 timeout client 30s
7082 timeout client 30s
7083 timeout server 30s
7084 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
7085
7086 See also : "timeout client", "timeout server".
7087
7088
Willy Tarreau844e3c52008-01-11 16:28:18 +01007089transparent (deprecated)
7090 Enable client-side transparent proxying
7091 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01007092 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01007093 Arguments : none
7094
7095 This keyword was introduced in order to provide layer 7 persistence to layer
7096 3 load balancers. The idea is to use the OS's ability to redirect an incoming
7097 connection for a remote address to a local process (here HAProxy), and let
7098 this process know what address was initially requested. When this option is
7099 used, sessions without cookies will be forwarded to the original destination
7100 IP address of the incoming request (which should match that of another
7101 equipment), while requests with cookies will still be forwarded to the
7102 appropriate server.
7103
7104 The "transparent" keyword is deprecated, use "option transparent" instead.
7105
7106 Note that contrary to a common belief, this option does NOT make HAProxy
7107 present the client's IP to the server when establishing the connection.
7108
Willy Tarreau844e3c52008-01-11 16:28:18 +01007109 See also: "option transparent"
7110
William Lallemanda73203e2012-03-12 12:48:57 +01007111unique-id-format <string>
7112 Generate a unique ID for each request.
7113 May be used in sections : defaults | frontend | listen | backend
7114 yes | yes | yes | no
7115 Arguments :
7116 <string> is a log-format string.
7117
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007118 This keyword creates a ID for each request using the custom log format. A
7119 unique ID is useful to trace a request passing through many components of
7120 a complex infrastructure. The newly created ID may also be logged using the
7121 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +01007122
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007123 The format should be composed from elements that are guaranteed to be
7124 unique when combined together. For instance, if multiple haproxy instances
7125 are involved, it might be important to include the node name. It is often
7126 needed to log the incoming connection's source and destination addresses
7127 and ports. Note that since multiple requests may be performed over the same
7128 connection, including a request counter may help differentiate them.
7129 Similarly, a timestamp may protect against a rollover of the counter.
7130 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +01007131
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007132 It is recommended to use hexadecimal notation for many fields since it
7133 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +01007134
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007135 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01007136
7137 unique-id-format %{+X}o\ %Ci:%Cp_%Fi:%Fp_%Ts_%rt:%pid
7138
7139 will generate:
7140
7141 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
7142
7143 See also: "unique-id-header"
7144
7145unique-id-header <name>
7146 Add a unique ID header in the HTTP request.
7147 May be used in sections : defaults | frontend | listen | backend
7148 yes | yes | yes | no
7149 Arguments :
7150 <name> is the name of the header.
7151
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007152 Add a unique-id header in the HTTP request sent to the server, using the
7153 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +01007154
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007155 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01007156
7157 unique-id-format %{+X}o\ %Ci:%Cp_%Fi:%Fp_%Ts_%rt:%pid
7158 unique-id-header X-Unique-ID
7159
7160 will generate:
7161
7162 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
7163
7164 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +01007165
7166use_backend <backend> if <condition>
7167use_backend <backend> unless <condition>
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02007168 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007169 May be used in sections : defaults | frontend | listen | backend
7170 no | yes | yes | no
7171 Arguments :
7172 <backend> is the name of a valid backend or "listen" section.
7173
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007174 <condition> is a condition composed of ACLs, as described in section 7.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007175
7176 When doing content-switching, connections arrive on a frontend and are then
7177 dispatched to various backends depending on a number of conditions. The
7178 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02007179 "use_backend" keyword. While it is normally used with HTTP processing, it can
7180 also be used in pure TCP, either without content using stateless ACLs (eg:
7181 source address validation) or combined with a "tcp-request" rule to wait for
7182 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007183
7184 There may be as many "use_backend" rules as desired. All of these rules are
7185 evaluated in their declaration order, and the first one which matches will
7186 assign the backend.
7187
7188 In the first form, the backend will be used if the condition is met. In the
7189 second form, the backend will be used if the condition is not met. If no
7190 condition is valid, the backend defined with "default_backend" will be used.
7191 If no default backend is defined, either the servers in the same section are
7192 used (in case of a "listen" section) or, in case of a frontend, no server is
7193 used and a 503 service unavailable response is returned.
7194
Willy Tarreau51aecc72009-07-12 09:47:04 +02007195 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007196 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +02007197 and backend processing will immediately follow, or the backend will wait for
7198 a complete HTTP request to get in. This feature is useful when a frontend
7199 must decode several protocols on a unique port, one of them being HTTP.
7200
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02007201 See also: "default_backend", "tcp-request", and section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +01007202
Willy Tarreau036fae02008-01-06 13:24:40 +01007203
Willy Tarreau4a5cade2012-04-05 21:09:48 +02007204use-server <server> if <condition>
7205use-server <server> unless <condition>
7206 Only use a specific server if/unless an ACL-based condition is matched.
7207 May be used in sections : defaults | frontend | listen | backend
7208 no | no | yes | yes
7209 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007210 <server> is the name of a valid server in the same backend section.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02007211
7212 <condition> is a condition composed of ACLs, as described in section 7.
7213
7214 By default, connections which arrive to a backend are load-balanced across
7215 the available servers according to the configured algorithm, unless a
7216 persistence mechanism such as a cookie is used and found in the request.
7217
7218 Sometimes it is desirable to forward a particular request to a specific
7219 server without having to declare a dedicated backend for this server. This
7220 can be achieved using the "use-server" rules. These rules are evaluated after
7221 the "redirect" rules and before evaluating cookies, and they have precedence
7222 on them. There may be as many "use-server" rules as desired. All of these
7223 rules are evaluated in their declaration order, and the first one which
7224 matches will assign the server.
7225
7226 If a rule designates a server which is down, and "option persist" is not used
7227 and no force-persist rule was validated, it is ignored and evaluation goes on
7228 with the next rules until one matches.
7229
7230 In the first form, the server will be used if the condition is met. In the
7231 second form, the server will be used if the condition is not met. If no
7232 condition is valid, the processing continues and the server will be assigned
7233 according to other persistence mechanisms.
7234
7235 Note that even if a rule is matched, cookie processing is still performed but
7236 does not assign the server. This allows prefixed cookies to have their prefix
7237 stripped.
7238
7239 The "use-server" statement works both in HTTP and TCP mode. This makes it
7240 suitable for use with content-based inspection. For instance, a server could
7241 be selected in a farm according to the TLS SNI field. And if these servers
7242 have their weight set to zero, they will not be used for other traffic.
7243
7244 Example :
7245 # intercept incoming TLS requests based on the SNI field
7246 use-server www if { req_ssl_sni -i www.example.com }
7247 server www 192.168.0.1:443 weight 0
7248 use-server mail if { req_ssl_sni -i mail.example.com }
7249 server mail 192.168.0.1:587 weight 0
7250 use-server imap if { req_ssl_sni -i imap.example.com }
7251 server mail 192.168.0.1:993 weight 0
7252 # all the rest is forwarded to this server
7253 server default 192.168.0.2:443 check
7254
7255 See also: "use_backend", serction 5 about server and section 7 about ACLs.
7256
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007257
72585. Bind and Server options
7259--------------------------
7260
7261The "bind", "server" and "default-server" keywords support a number of settings
7262depending on some build options and on the system HAProxy was built on. These
7263settings generally each consist in one word sometimes followed by a value,
7264written on the same line as the "bind" or "server" line. All these options are
7265described in this section.
7266
7267
72685.1. Bind options
7269-----------------
7270
7271The "bind" keyword supports a certain number of settings which are all passed
7272as arguments on the same line. The order in which those arguments appear makes
7273no importance, provided that they appear after the bind address. All of these
7274parameters are optional. Some of them consist in a single words (booleans),
7275while other ones expect a value after them. In this case, the value must be
7276provided immediately after the setting name.
7277
7278The currently supported settings are the following ones.
7279
7280accept-proxy
7281 Enforces the use of the PROXY protocol over any connection accepted by any of
7282 the sockets declared on the same line. The PROXY protocol dictates the layer
7283 3/4 addresses of the incoming connection to be used everywhere an address is
7284 used, with the only exception of "tcp-request connection" rules which will
7285 only see the real connection address. Logs will reflect the addresses
7286 indicated in the protocol, unless it is violated, in which case the real
7287 address will still be used. This keyword combined with support from external
7288 components can be used as an efficient and reliable alternative to the
7289 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +02007290 usable. See also "tcp-request connection expect-proxy" for a finer-grained
7291 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007292
Willy Tarreauab861d32013-04-02 02:30:41 +02007293alpn <protocols>
7294 This enables the TLS ALPN extension and advertises the specified protocol
7295 list as supported on top of ALPN. The protocol list consists in a comma-
7296 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
7297 quotes). This requires that the SSL library is build with support for TLS
7298 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
7299 initial NPN extension.
7300
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007301backlog <backlog>
7302 Sets the socket's backlog to this value. If unspecified, the frontend's
7303 backlog is used instead, which generally defaults to the maxconn value.
7304
Emeric Brun7fb34422012-09-28 15:26:15 +02007305ecdhe <named curve>
7306 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +01007307 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
7308 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +02007309
Emeric Brunfd33a262012-10-11 16:28:27 +02007310ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +02007311 This setting is only available when support for OpenSSL was built in. It
7312 designates a PEM file from which to load CA certificates used to verify
7313 client's certificate.
7314
Emeric Brunb6dc9342012-09-28 17:55:37 +02007315ca-ignore-err [all|<errorID>,...]
7316 This setting is only available when support for OpenSSL was built in.
7317 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
7318 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
7319 error is ignored.
7320
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007321ciphers <ciphers>
7322 This setting is only available when support for OpenSSL was built in. It sets
7323 the string describing the list of cipher algorithms ("cipher suite") that are
7324 negociated during the SSL/TLS handshake. The format of the string is defined
7325 in "man 1 ciphers" from OpenSSL man pages, and can be for instance a string
7326 such as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes).
7327
Emeric Brunfd33a262012-10-11 16:28:27 +02007328crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +02007329 This setting is only available when support for OpenSSL was built in. It
7330 designates a PEM file from which to load certificate revocation list used
7331 to verify client's certificate.
7332
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007333crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +00007334 This setting is only available when support for OpenSSL was built in. It
7335 designates a PEM file containing both the required certificates and any
7336 associated private keys. This file can be built by concatenating multiple
7337 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
7338 requires an intermediate certificate, this can also be concatenated into this
7339 file.
7340
7341 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
7342 are loaded.
7343
7344 If a directory name is used instead of a PEM file, then all files found in
7345 that directory will be loaded. This directive may be specified multiple times
7346 in order to load certificates from multiple files or directories. The
7347 certificates will be presented to clients who provide a valid TLS Server Name
7348 Indication field matching one of their CN or alt subjects. Wildcards are
7349 supported, where a wildcard character '*' is used instead of the first
7350 hostname component (eg: *.example.org matches www.example.org but not
7351 www.sub.example.org).
7352
7353 If no SNI is provided by the client or if the SSL library does not support
7354 TLS extensions, or if the client provides an SNI hostname which does not
7355 match any certificate, then the first loaded certificate will be presented.
7356 This means that when loading certificates from a directory, it is highly
7357 recommended to load the default one first as a file.
7358
Emeric Brune032bfa2012-09-28 13:01:45 +02007359 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007360
Alex Davies0fbf0162013-03-02 16:04:50 +00007361 Some CAs (such as Godaddy) offer a drop down list of server types that do not
7362 include HAProxy when obtaining a certificate. If this happens be sure to
7363 choose a webserver that the CA believes requires a intermediate CA (for
7364 Godaddy, selection Apache Tomcat will get the correct bundle, but many
7365 others, e.g. nginx, result in a wrong bundle that will not work for some
7366 clients).
7367
Emeric Brunb6dc9342012-09-28 17:55:37 +02007368crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +00007369 This setting is only available when support for OpenSSL was built in. Sets a
7370 comma separated list of errorIDs to ignore during verify at depth == 0. If
7371 set to 'all', all errors are ignored. SSL handshake is not abored if an error
7372 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +02007373
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01007374crt-list <file>
7375 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02007376 designates a list of PEM file with an optional list of SNI filter per
7377 certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01007378
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02007379 <crtfile> [[!]<snifilter> ...]
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01007380
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02007381 Wildcards are supported in the SNI filter. Negative filter are also supported,
7382 only useful in combination with a wildcard filter to exclude a particular SNI.
7383 The certificates will be presented to clients who provide a valid TLS Server
7384 Name Indication field matching one of the SNI filters. If no SNI filter is
7385 specified, the CN and alt subjects are used. This directive may be specified
7386 multiple times. See the "crt" option for more information. The default
7387 certificate is still needed to meet OpenSSL expectations. If it is not used,
7388 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01007389
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007390defer-accept
7391 Is an optional keyword which is supported only on certain Linux kernels. It
7392 states that a connection will only be accepted once some data arrive on it,
7393 or at worst after the first retransmit. This should be used only on protocols
7394 for which the client talks first (eg: HTTP). It can slightly improve
7395 performance by ensuring that most of the request is already available when
7396 the connection is accepted. On the other hand, it will not be able to detect
7397 connections which don't talk. It is important to note that this option is
7398 broken in all kernels up to 2.6.31, as the connection is never accepted until
7399 the client talks. This can cause issues with front firewalls which would see
7400 an established connection while the proxy will only see it in SYN_RECV. This
7401 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
7402
Emeric Brun2cb7ae52012-10-05 14:14:21 +02007403force-sslv3
7404 This option enforces use of SSLv3 only on SSL connections instanciated from
7405 this listener. SSLv3 is generally less expensive than the TLS counterparts
7406 for high connection rates. See also "force-tls*", "no-sslv3", and "no-tls*".
7407
7408force-tlsv10
7409 This option enforces use of TLSv1.0 only on SSL connections instanciated from
7410 this listener. See also "force-tls*", "no-sslv3", and "no-tls*".
7411
7412force-tlsv11
7413 This option enforces use of TLSv1.1 only on SSL connections instanciated from
7414 this listener. See also "force-tls*", "no-sslv3", and "no-tls*".
7415
7416force-tlsv12
7417 This option enforces use of TLSv1.2 only on SSL connections instanciated from
7418 this listener. See also "force-tls*", "no-sslv3", and "no-tls*".
7419
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007420gid <gid>
7421 Sets the group of the UNIX sockets to the designated system gid. It can also
7422 be set by default in the global section's "unix-bind" statement. Note that
7423 some platforms simply ignore this. This setting is equivalent to the "group"
7424 setting except that the group ID is used instead of its name. This setting is
7425 ignored by non UNIX sockets.
7426
7427group <group>
7428 Sets the group of the UNIX sockets to the designated system group. It can
7429 also be set by default in the global section's "unix-bind" statement. Note
7430 that some platforms simply ignore this. This setting is equivalent to the
7431 "gid" setting except that the group name is used instead of its gid. This
7432 setting is ignored by non UNIX sockets.
7433
7434id <id>
7435 Fixes the socket ID. By default, socket IDs are automatically assigned, but
7436 sometimes it is more convenient to fix them to ease monitoring. This value
7437 must be strictly positive and unique within the listener/frontend. This
7438 option can only be used when defining only a single socket.
7439
7440interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +01007441 Restricts the socket to a specific interface. When specified, only packets
7442 received from that particular interface are processed by the socket. This is
7443 currently only supported on Linux. The interface must be a primary system
7444 interface, not an aliased interface. It is also possible to bind multiple
7445 frontends to the same address if they are bound to different interfaces. Note
7446 that binding to a network interface requires root privileges. This parameter
7447 is only compatible with TCPv4/TCPv6 sockets.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007448
Willy Tarreauabb175f2012-09-24 12:43:26 +02007449level <level>
7450 This setting is used with the stats sockets only to restrict the nature of
7451 the commands that can be issued on the socket. It is ignored by other
7452 sockets. <level> can be one of :
7453 - "user" is the least privileged level ; only non-sensitive stats can be
7454 read, and no change is allowed. It would make sense on systems where it
7455 is not easy to restrict access to the socket.
7456 - "operator" is the default level and fits most common uses. All data can
7457 be read, and only non-sensitive changes are permitted (eg: clear max
7458 counters).
7459 - "admin" should be used with care, as everything is permitted (eg: clear
7460 all counters).
7461
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007462maxconn <maxconn>
7463 Limits the sockets to this number of concurrent connections. Extraneous
7464 connections will remain in the system's backlog until a connection is
7465 released. If unspecified, the limit will be the same as the frontend's
7466 maxconn. Note that in case of port ranges or multiple addresses, the same
7467 value will be applied to each socket. This setting enables different
7468 limitations on expensive sockets, for instance SSL entries which may easily
7469 eat all memory.
7470
7471mode <mode>
7472 Sets the octal mode used to define access permissions on the UNIX socket. It
7473 can also be set by default in the global section's "unix-bind" statement.
7474 Note that some platforms simply ignore this. This setting is ignored by non
7475 UNIX sockets.
7476
7477mss <maxseg>
7478 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
7479 connections. This can be used to force a lower MSS for certain specific
7480 ports, for instance for connections passing through a VPN. Note that this
7481 relies on a kernel feature which is theoretically supported under Linux but
7482 was buggy in all versions prior to 2.6.28. It may or may not work on other
7483 operating systems. It may also not change the advertised value but change the
7484 effective size of outgoing segments. The commonly advertised value for TCPv4
7485 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
7486 positive, it will be used as the advertised MSS. If it is negative, it will
7487 indicate by how much to reduce the incoming connection's advertised MSS for
7488 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
7489
7490name <name>
7491 Sets an optional name for these sockets, which will be reported on the stats
7492 page.
7493
7494nice <nice>
7495 Sets the 'niceness' of connections initiated from the socket. Value must be
7496 in the range -1024..1024 inclusive, and defaults to zero. Positive values
7497 means that such connections are more friendly to others and easily offer
7498 their place in the scheduler. On the opposite, negative values mean that
7499 connections want to run with a higher priority than others. The difference
7500 only happens under high loads when the system is close to saturation.
7501 Negative values are appropriate for low-latency or administration services,
7502 and high values are generally recommended for CPU intensive tasks such as SSL
7503 processing or bulk transfers which are less sensible to latency. For example,
7504 it may make sense to use a positive value for an SMTP socket and a negative
7505 one for an RDP socket.
7506
Emeric Brun9b3009b2012-10-05 11:55:06 +02007507no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007508 This setting is only available when support for OpenSSL was built in. It
7509 disables support for SSLv3 on any sockets instanciated from the listener when
7510 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2cb7ae52012-10-05 14:14:21 +02007511 be enabled using any configuration option. See also "force-tls*",
7512 and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007513
Emeric Brun90ad8722012-10-02 14:00:59 +02007514no-tls-tickets
7515 This setting is only available when support for OpenSSL was built in. It
7516 disables the stateless session resumption (RFC 5077 TLS Ticket
7517 extension) and force to use stateful session resumption. Stateless
7518 session resumption is more expensive in CPU usage.
7519
Emeric Brun9b3009b2012-10-05 11:55:06 +02007520no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007521 This setting is only available when support for OpenSSL was built in. It
Emeric Brun2cb7ae52012-10-05 14:14:21 +02007522 disables support for TLSv1.0 on any sockets instanciated from the listener
7523 when SSL is supported. Note that SSLv2 is forced disabled in the code and
7524 cannot be enabled using any configuration option. See also "force-tls*",
7525 and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007526
Emeric Brun9b3009b2012-10-05 11:55:06 +02007527no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +02007528 This setting is only available when support for OpenSSL was built in. It
Emeric Brun2cb7ae52012-10-05 14:14:21 +02007529 disables support for TLSv1.1 on any sockets instanciated from the listener
7530 when SSL is supported. Note that SSLv2 is forced disabled in the code and
7531 cannot be enabled using any configuration option. See also "force-tls*",
7532 and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +02007533
Emeric Brun9b3009b2012-10-05 11:55:06 +02007534no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +02007535 This setting is only available when support for OpenSSL was built in. It
Emeric Brun2cb7ae52012-10-05 14:14:21 +02007536 disables support for TLSv1.2 on any sockets instanciated from the listener
7537 when SSL is supported. Note that SSLv2 is forced disabled in the code and
7538 cannot be enabled using any configuration option. See also "force-tls*",
7539 and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +02007540
Willy Tarreau6c9a3d52012-10-18 18:57:14 +02007541npn <protocols>
7542 This enables the NPN TLS extension and advertises the specified protocol list
7543 as supported on top of NPN. The protocol list consists in a comma-delimited
7544 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
7545 This requires that the SSL library is build with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +02007546 enabled (check with haproxy -vv). Note that the NPN extension has been
7547 replaced with the ALPN extension (see the "alpn" keyword).
Willy Tarreau6c9a3d52012-10-18 18:57:14 +02007548
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007549ssl
7550 This setting is only available when support for OpenSSL was built in. It
7551 enables SSL deciphering on connections instanciated from this listener. A
7552 certificate is necessary (see "crt" above). All contents in the buffers will
7553 appear in clear text, so that ACLs and HTTP processing will only have access
7554 to deciphered contents.
7555
Emmanuel Hocdet65623372013-01-24 17:17:15 +01007556strict-sni
7557 This setting is only available when support for OpenSSL was built in. The
7558 SSL/TLS negotiation is allow only if the client provided an SNI which match
7559 a certificate. The default certificate is not used.
7560 See the "crt" option for more information.
7561
Willy Tarreau1c862c52012-10-05 16:21:00 +02007562tfo
Lukas Tribus0defb902013-02-13 23:35:39 +01007563 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +02007564 enables TCP Fast Open on the listening socket, which means that clients which
7565 support this feature will be able to send a request and receive a response
7566 during the 3-way handshake starting from second connection, thus saving one
7567 round-trip after the first connection. This only makes sense with protocols
7568 that use high connection rates and where each round trip matters. This can
7569 possibly cause issues with many firewalls which do not accept data on SYN
7570 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +02007571 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
7572 need to build HAProxy with USE_TFO=1 if your libc doesn't define
7573 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +02007574
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007575transparent
7576 Is an optional keyword which is supported only on certain Linux kernels. It
7577 indicates that the addresses will be bound even if they do not belong to the
7578 local machine, and that packets targeting any of these addresses will be
7579 intercepted just as if the addresses were locally configured. This normally
7580 requires that IP forwarding is enabled. Caution! do not use this with the
7581 default address '*', as it would redirect any traffic for the specified port.
7582 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
7583 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
7584 kernel version. Some distribution kernels include backports of the feature,
7585 so check for support with your vendor.
7586
Willy Tarreau77e3af92012-11-24 15:07:23 +01007587v4v6
7588 Is an optional keyword which is supported only on most recent systems
7589 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
7590 and IPv6 when it uses the default address. Doing so is sometimes necessary
7591 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
7592 sockets, and is overriden by the "v6only" option.
7593
Willy Tarreau9b6700f2012-11-24 11:55:28 +01007594v6only
7595 Is an optional keyword which is supported only on most recent systems
7596 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
7597 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +01007598 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
7599 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +01007600
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007601uid <uid>
7602 Sets the owner of the UNIX sockets to the designated system uid. It can also
7603 be set by default in the global section's "unix-bind" statement. Note that
7604 some platforms simply ignore this. This setting is equivalent to the "user"
7605 setting except that the user numeric ID is used instead of its name. This
7606 setting is ignored by non UNIX sockets.
7607
7608user <user>
7609 Sets the owner of the UNIX sockets to the designated system user. It can also
7610 be set by default in the global section's "unix-bind" statement. Note that
7611 some platforms simply ignore this. This setting is equivalent to the "uid"
7612 setting except that the user name is used instead of its uid. This setting is
7613 ignored by non UNIX sockets.
7614
Emeric Brun1a073b42012-09-28 17:07:34 +02007615verify [none|optional|required]
7616 This setting is only available when support for OpenSSL was built in. If set
7617 to 'none', client certificate is not requested. This is the default. In other
7618 cases, a client certificate is requested. If the client does not provide a
7619 certificate after the request and if 'verify' is set to 'required', then the
7620 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +02007621 certificate provided by the client is always verified using CAs from
7622 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
7623 is aborted, regardless of the 'verify' option, unless the error code exactly
7624 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02007625
Willy Tarreaub6205fd2012-09-24 12:27:33 +020076265.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +01007627------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02007628
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01007629The "server" and "default-server" keywords support a certain number of settings
7630which are all passed as arguments on the server line. The order in which those
7631arguments appear does not count, and they are all optional. Some of those
7632settings are single words (booleans) while others expect one or several values
7633after them. In this case, the values must immediately follow the setting name.
7634Except default-server, all those settings must be specified after the server's
7635address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +02007636
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007637 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01007638 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +02007639
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007640The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +01007641
Willy Tarreauceb4ac92012-04-28 00:41:46 +02007642addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007643 Using the "addr" parameter, it becomes possible to use a different IP address
7644 to send health-checks. On some servers, it may be desirable to dedicate an IP
7645 address to specific component able to perform complex tests which are more
7646 suitable to health-checks than the application. This parameter is ignored if
7647 the "check" parameter is not set. See also the "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +02007648
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007649 Supported in default-server: No
7650
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007651backup
7652 When "backup" is present on a server line, the server is only used in load
7653 balancing when all other non-backup servers are unavailable. Requests coming
7654 with a persistence cookie referencing the server will always be served
7655 though. By default, only the first operational backup server is used, unless
7656 the "allbackups" option is set in the backend. See also the "allbackups"
7657 option.
Willy Tarreau6a06a402007-07-15 20:15:28 +02007658
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007659 Supported in default-server: No
7660
Emeric Brunef42d922012-10-11 16:11:36 +02007661ca-file <cafile>
7662 This setting is only available when support for OpenSSL was built in. It
7663 designates a PEM file from which to load CA certificates used to verify
7664 server's certificate.
7665
7666 Supported in default-server: No
7667
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007668check
7669 This option enables health checks on the server. By default, a server is
Patrick Mézardb7aeec62012-01-22 16:01:22 +01007670 always considered available. If "check" is set, the server is available when
7671 accepting periodic TCP connections, to ensure that it is really able to serve
7672 requests. The default address and port to send the tests to are those of the
7673 server, and the default source is the same as the one defined in the
7674 backend. It is possible to change the address using the "addr" parameter, the
7675 port using the "port" parameter, the source address using the "source"
7676 address, and the interval and timers using the "inter", "rise" and "fall"
Simon Hormana2b9dad2013-02-12 10:45:54 +09007677 parameters. The request method is define in the backend using the
7678 "httpchk", "lb-agent-chk", "smtpchk", "mysql-check", "pgsql-check" and
7679 "ssl-hello-chk" options. Please refer to those options and parameters for
7680 more information.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007681
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007682 Supported in default-server: No
7683
Willy Tarreau6c16adc2012-10-05 00:04:16 +02007684check-send-proxy
7685 This option forces emission of a PROXY protocol line with outgoing health
7686 checks, regardless of whether the server uses send-proxy or not for the
7687 normal traffic. By default, the PROXY protocol is enabled for health checks
7688 if it is already enabled for normal traffic and if no "port" nor "addr"
7689 directive is present. However, if such a directive is present, the
7690 "check-send-proxy" option needs to be used to force the use of the
7691 protocol. See also the "send-proxy" option for more information.
7692
7693 Supported in default-server: No
7694
Willy Tarreau763a95b2012-10-04 23:15:39 +02007695check-ssl
7696 This option forces encryption of all health checks over SSL, regardless of
7697 whether the server uses SSL or not for the normal traffic. This is generally
7698 used when an explicit "port" or "addr" directive is specified and SSL health
7699 checks are not inherited. It is important to understand that this option
7700 inserts an SSL transport layer below the ckecks, so that a simple TCP connect
7701 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
7702 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
7703 All SSL settings are common to health checks and traffic (eg: ciphers).
7704 See the "ssl" option for more information.
7705
7706 Supported in default-server: No
7707
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02007708ciphers <ciphers>
7709 This option sets the string describing the list of cipher algorithms that is
7710 is negociated during the SSL/TLS handshake with the server. The format of the
7711 string is defined in "man 1 ciphers". When SSL is used to communicate with
7712 servers on the local network, it is common to see a weaker set of algorithms
7713 than what is used over the internet. Doing so reduces CPU usage on both the
7714 server and haproxy while still keeping it compatible with deployed software.
7715 Some algorithms such as RC4-SHA1 are reasonably cheap. If no security at all
7716 is needed and just connectivity, using DES can be appropriate.
7717
Willy Tarreau763a95b2012-10-04 23:15:39 +02007718 Supported in default-server: No
7719
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007720cookie <value>
7721 The "cookie" parameter sets the cookie value assigned to the server to
7722 <value>. This value will be checked in incoming requests, and the first
7723 operational server possessing the same value will be selected. In return, in
7724 cookie insertion or rewrite modes, this value will be assigned to the cookie
7725 sent to the client. There is nothing wrong in having several servers sharing
7726 the same cookie value, and it is in fact somewhat common between normal and
7727 backup servers. See also the "cookie" keyword in backend section.
7728
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007729 Supported in default-server: No
7730
Emeric Brunef42d922012-10-11 16:11:36 +02007731crl-file <crlfile>
7732 This setting is only available when support for OpenSSL was built in. It
7733 designates a PEM file from which to load certificate revocation list used
7734 to verify server's certificate.
7735
7736 Supported in default-server: No
7737
Emeric Bruna7aa3092012-10-26 12:58:00 +02007738crt <cert>
7739 This setting is only available when support for OpenSSL was built in.
7740 It designates a PEM file from which to load both a certificate and the
7741 associated private key. This file can be built by concatenating both PEM
7742 files into one. This certificate will be sent if the server send a client
7743 certificate request.
7744
7745 Supported in default-server: No
7746
Willy Tarreau96839092010-03-29 10:02:24 +02007747disabled
7748 The "disabled" keyword starts the server in the "disabled" state. That means
7749 that it is marked down in maintenance mode, and no connection other than the
7750 ones allowed by persist mode will reach it. It is very well suited to setup
7751 new servers, because normal traffic will never reach them, while it is still
7752 possible to test the service by making use of the force-persist mechanism.
7753
7754 Supported in default-server: No
7755
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007756error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +01007757 If health observing is enabled, the "error-limit" parameter specifies the
7758 number of consecutive errors that triggers event selected by the "on-error"
7759 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01007760
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007761 Supported in default-server: Yes
7762
7763 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01007764
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007765fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007766 The "fall" parameter states that a server will be considered as dead after
7767 <count> consecutive unsuccessful health checks. This value defaults to 3 if
7768 unspecified. See also the "check", "inter" and "rise" parameters.
7769
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007770 Supported in default-server: Yes
7771
Emeric Brun8694b9a2012-10-05 14:39:07 +02007772force-sslv3
7773 This option enforces use of SSLv3 only when SSL is used to communicate with
7774 the server. SSLv3 is generally less expensive than the TLS counterparts for
7775 high connection rates. See also "no-tlsv*", "no-sslv3".
7776
7777 Supported in default-server: No
7778
7779force-tlsv10
7780 This option enforces use of TLSv1.0 only when SSL is used to communicate with
7781 the server. See also "no-tlsv*", "no-sslv3".
7782
7783 Supported in default-server: No
7784
7785force-tlsv11
7786 This option enforces use of TLSv1.1 only when SSL is used to communicate with
7787 the server. See also "no-tlsv*", "no-sslv3".
7788
7789 Supported in default-server: No
7790
7791force-tlsv12
7792 This option enforces use of TLSv1.2 only when SSL is used to communicate with
7793 the server. See also "no-tlsv*", "no-sslv3".
7794
7795 Supported in default-server: No
7796
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007797id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02007798 Set a persistent ID for the server. This ID must be positive and unique for
7799 the proxy. An unused ID will automatically be assigned if unset. The first
7800 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007801
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007802 Supported in default-server: No
7803
7804inter <delay>
7805fastinter <delay>
7806downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007807 The "inter" parameter sets the interval between two consecutive health checks
7808 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
7809 It is also possible to use "fastinter" and "downinter" to optimize delays
7810 between checks depending on the server state :
7811
7812 Server state | Interval used
7813 ---------------------------------+-----------------------------------------
7814 UP 100% (non-transitional) | "inter"
7815 ---------------------------------+-----------------------------------------
7816 Transitionally UP (going down), |
7817 Transitionally DOWN (going up), | "fastinter" if set, "inter" otherwise.
7818 or yet unchecked. |
7819 ---------------------------------+-----------------------------------------
7820 DOWN 100% (non-transitional) | "downinter" if set, "inter" otherwise.
7821 ---------------------------------+-----------------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +01007822
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007823 Just as with every other time-based parameter, they can be entered in any
7824 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
7825 serves as a timeout for health checks sent to servers if "timeout check" is
7826 not set. In order to reduce "resonance" effects when multiple servers are
7827 hosted on the same hardware, the health-checks of all servers are started
7828 with a small time offset between them. It is also possible to add some random
7829 noise in the health checks interval using the global "spread-checks"
7830 keyword. This makes sense for instance when a lot of backends use the same
7831 servers.
7832
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007833 Supported in default-server: Yes
7834
7835maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007836 The "maxconn" parameter specifies the maximal number of concurrent
7837 connections that will be sent to this server. If the number of incoming
7838 concurrent requests goes higher than this value, they will be queued, waiting
7839 for a connection to be released. This parameter is very important as it can
7840 save fragile servers from going down under extreme loads. If a "minconn"
7841 parameter is specified, the limit becomes dynamic. The default value is "0"
7842 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
7843 the backend's "fullconn" keyword.
7844
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007845 Supported in default-server: Yes
7846
7847maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007848 The "maxqueue" parameter specifies the maximal number of connections which
7849 will wait in the queue for this server. If this limit is reached, next
7850 requests will be redispatched to other servers instead of indefinitely
7851 waiting to be served. This will break persistence but may allow people to
7852 quickly re-log in when the server they try to connect to is dying. The
7853 default value is "0" which means the queue is unlimited. See also the
7854 "maxconn" and "minconn" parameters.
7855
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007856 Supported in default-server: Yes
7857
7858minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007859 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
7860 limit following the backend's load. The server will always accept at least
7861 <minconn> connections, never more than <maxconn>, and the limit will be on
7862 the ramp between both values when the backend has less than <fullconn>
7863 concurrent connections. This makes it possible to limit the load on the
7864 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007865 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007866 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01007867
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007868 Supported in default-server: Yes
7869
Emeric Brun9b3009b2012-10-05 11:55:06 +02007870no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02007871 This option disables support for SSLv3 when SSL is used to communicate with
7872 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emeric Brun8694b9a2012-10-05 14:39:07 +02007873 using any configuration option. See also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02007874
Willy Tarreau763a95b2012-10-04 23:15:39 +02007875 Supported in default-server: No
7876
Emeric Brunf9c5c472012-10-11 15:28:34 +02007877no-tls-tickets
7878 This setting is only available when support for OpenSSL was built in. It
7879 disables the stateless session resumption (RFC 5077 TLS Ticket
7880 extension) and force to use stateful session resumption. Stateless
7881 session resumption is more expensive in CPU usage for servers.
7882
7883 Supported in default-server: No
7884
Emeric Brun9b3009b2012-10-05 11:55:06 +02007885no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +02007886 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +02007887 the server. Note that SSLv2 is disabled in the code and cannot be enabled
7888 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun8694b9a2012-10-05 14:39:07 +02007889 often makes sense to disable it when communicating with local servers. See
7890 also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +02007891
Willy Tarreau763a95b2012-10-04 23:15:39 +02007892 Supported in default-server: No
7893
Emeric Brun9b3009b2012-10-05 11:55:06 +02007894no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +02007895 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +02007896 the server. Note that SSLv2 is disabled in the code and cannot be enabled
7897 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun8694b9a2012-10-05 14:39:07 +02007898 often makes sense to disable it when communicating with local servers. See
7899 also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +02007900
Willy Tarreau763a95b2012-10-04 23:15:39 +02007901 Supported in default-server: No
7902
Emeric Brun9b3009b2012-10-05 11:55:06 +02007903no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +02007904 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02007905 the server. Note that SSLv2 is disabled in the code and cannot be enabled
7906 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun8694b9a2012-10-05 14:39:07 +02007907 often makes sense to disable it when communicating with local servers. See
7908 also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02007909
Willy Tarreau763a95b2012-10-04 23:15:39 +02007910 Supported in default-server: No
7911
Simon Hormanfa461682011-06-25 09:39:49 +09007912non-stick
7913 Never add connections allocated to this sever to a stick-table.
7914 This may be used in conjunction with backup to ensure that
7915 stick-table persistence is disabled for backup servers.
7916
Willy Tarreau763a95b2012-10-04 23:15:39 +02007917 Supported in default-server: No
7918
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01007919observe <mode>
7920 This option enables health adjusting based on observing communication with
7921 the server. By default this functionality is disabled and enabling it also
7922 requires to enable health checks. There are two supported modes: "layer4" and
7923 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
7924 significant. In layer7, which is only allowed for http proxies, responses
7925 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +01007926 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01007927
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007928 Supported in default-server: No
7929
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01007930 See also the "check", "on-error" and "error-limit".
7931
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007932on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01007933 Select what should happen when enough consecutive errors are detected.
7934 Currently, four modes are available:
7935 - fastinter: force fastinter
7936 - fail-check: simulate a failed check, also forces fastinter (default)
7937 - sudden-death: simulate a pre-fatal failed health check, one more failed
7938 check will mark a server down, forces fastinter
7939 - mark-down: mark the server immediately down and force fastinter
7940
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007941 Supported in default-server: Yes
7942
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01007943 See also the "check", "observe" and "error-limit".
7944
Simon Hormane0d1bfb2011-06-21 14:34:58 +09007945on-marked-down <action>
7946 Modify what occurs when a server is marked down.
7947 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -07007948 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
7949 all connections to the server are immediately terminated when the server
7950 goes down. It might be used if the health check detects more complex cases
7951 than a simple connection status, and long timeouts would cause the service
7952 to remain unresponsive for too long a time. For instance, a health check
7953 might detect that a database is stuck and that there's no chance to reuse
7954 existing connections anymore. Connections killed this way are logged with
7955 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +09007956
7957 Actions are disabled by default
7958
7959 Supported in default-server: Yes
7960
Justin Karnegeseb2c24a2012-05-24 15:28:52 -07007961on-marked-up <action>
7962 Modify what occurs when a server is marked up.
7963 Currently one action is available:
7964 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
7965 done only if the server is not in backup state and if it is not disabled
7966 (it must have an effective weight > 0). This can be used sometimes to force
7967 an active server to take all the traffic back after recovery when dealing
7968 with long sessions (eg: LDAP, SQL, ...). Doing this can cause more trouble
7969 than it tries to solve (eg: incomplete transactions), so use this feature
7970 with extreme care. Sessions killed because a server comes up are logged
7971 with an 'U' termination code (for "Up").
7972
7973 Actions are disabled by default
7974
7975 Supported in default-server: Yes
7976
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007977port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007978 Using the "port" parameter, it becomes possible to use a different port to
7979 send health-checks. On some servers, it may be desirable to dedicate a port
7980 to a specific component able to perform complex tests which are more suitable
7981 to health-checks than the application. It is common to run a simple script in
7982 inetd for instance. This parameter is ignored if the "check" parameter is not
7983 set. See also the "addr" parameter.
7984
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007985 Supported in default-server: Yes
7986
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007987redir <prefix>
7988 The "redir" parameter enables the redirection mode for all GET and HEAD
7989 requests addressing this server. This means that instead of having HAProxy
7990 forward the request to the server, it will send an "HTTP 302" response with
7991 the "Location" header composed of this prefix immediately followed by the
7992 requested URI beginning at the leading '/' of the path component. That means
7993 that no trailing slash should be used after <prefix>. All invalid requests
7994 will be rejected, and all non-GET or HEAD requests will be normally served by
7995 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007996 mangling nor cookie insertion is possible in the response. However, cookies in
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007997 requests are still analysed, making this solution completely usable to direct
7998 users to a remote location in case of local disaster. Main use consists in
7999 increasing bandwidth for static servers by having the clients directly
8000 connect to them. Note: never use a relative location here, it would cause a
8001 loop between the client and HAProxy!
8002
8003 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
8004
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008005 Supported in default-server: No
8006
8007rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008008 The "rise" parameter states that a server will be considered as operational
8009 after <count> consecutive successful health checks. This value defaults to 2
8010 if unspecified. See also the "check", "inter" and "fall" parameters.
8011
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008012 Supported in default-server: Yes
8013
Willy Tarreau5ab04ec2011-03-20 10:32:26 +01008014send-proxy
8015 The "send-proxy" parameter enforces use of the PROXY protocol over any
8016 connection established to this server. The PROXY protocol informs the other
8017 end about the layer 3/4 addresses of the incoming connection, so that it can
8018 know the client's address or the public address it accessed to, whatever the
8019 upper layer protocol. For connections accepted by an "accept-proxy" listener,
8020 the advertised address will be used. Only TCPv4 and TCPv6 address families
8021 are supported. Other families such as Unix sockets, will report an UNKNOWN
8022 family. Servers using this option can fully be chained to another instance of
8023 haproxy listening with an "accept-proxy" setting. This setting must not be
Willy Tarreau6c16adc2012-10-05 00:04:16 +02008024 used if the server isn't aware of the protocol. When health checks are sent
8025 to the server, the PROXY protocol is automatically used when this option is
8026 set, unless there is an explicit "port" or "addr" directive, in which case an
8027 explicit "check-send-proxy" directive would also be needed to use the PROXY
8028 protocol. See also the "accept-proxy" option of the "bind" keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +01008029
8030 Supported in default-server: No
8031
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008032slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008033 The "slowstart" parameter for a server accepts a value in milliseconds which
8034 indicates after how long a server which has just come back up will run at
8035 full speed. Just as with every other time-based parameter, it can be entered
8036 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
8037 linearly from 0 to 100% during this time. The limitation applies to two
8038 parameters :
8039
8040 - maxconn: the number of connections accepted by the server will grow from 1
8041 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
8042
8043 - weight: when the backend uses a dynamic weighted algorithm, the weight
8044 grows linearly from 1 to 100%. In this case, the weight is updated at every
8045 health-check. For this reason, it is important that the "inter" parameter
8046 is smaller than the "slowstart", in order to maximize the number of steps.
8047
8048 The slowstart never applies when haproxy starts, otherwise it would cause
8049 trouble to running servers. It only applies when a server has been previously
8050 seen as failed.
8051
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008052 Supported in default-server: Yes
8053
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02008054source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02008055source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02008056source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008057 The "source" parameter sets the source address which will be used when
8058 connecting to the server. It follows the exact same parameters and principle
8059 as the backend "source" keyword, except that it only applies to the server
8060 referencing it. Please consult the "source" keyword for details.
8061
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02008062 Additionally, the "source" statement on a server line allows one to specify a
8063 source port range by indicating the lower and higher bounds delimited by a
8064 dash ('-'). Some operating systems might require a valid IP address when a
8065 source port range is specified. It is permitted to have the same IP/range for
8066 several servers. Doing so makes it possible to bypass the maximum of 64k
8067 total concurrent connections. The limit will then reach 64k connections per
8068 server.
8069
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008070 Supported in default-server: No
8071
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02008072ssl
Willy Tarreau44f65392013-06-25 07:56:20 +02008073 This option enables SSL ciphering on outgoing connections to the server. It
8074 is critical to verify server certificates using "verify" when using SSL to
8075 connect to servers, otherwise the communication is prone to trivial man in
8076 the-middle attacks rendering SSL useless. When this option is used, health
8077 checks are automatically sent in SSL too unless there is a "port" or an
8078 "addr" directive indicating the check should be sent to a different location.
8079 See the "check-ssl" optino to force SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +02008080
8081 Supported in default-server: No
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02008082
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008083track [<proxy>/]<server>
8084 This option enables ability to set the current state of the server by
8085 tracking another one. Only a server with checks enabled can be tracked
8086 so it is not possible for example to track a server that tracks another
8087 one. If <proxy> is omitted the current one is used. If disable-on-404 is
8088 used, it has to be enabled on both proxies.
8089
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008090 Supported in default-server: No
8091
Emeric Brunef42d922012-10-11 16:11:36 +02008092verify [none|required]
8093 This setting is only available when support for OpenSSL was built in. If set
8094 to 'none', server certificate is not verified. This is the default. In the
8095 other case, The certificate provided by the server is verified using CAs from
8096 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
Willy Tarreau44f65392013-06-25 07:56:20 +02008097 is aborted. It is critically important to verify server certificates when
8098 using SSL to connect to servers, otherwise the communication is prone to
8099 trivial man-in-the-middle attacks rendering SSL totally useless.
Emeric Brunef42d922012-10-11 16:11:36 +02008100
8101 Supported in default-server: No
8102
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008103weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008104 The "weight" parameter is used to adjust the server's weight relative to
8105 other servers. All servers will receive a load proportional to their weight
8106 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +02008107 load. The default weight is 1, and the maximal value is 256. A value of 0
8108 means the server will not participate in load-balancing but will still accept
8109 persistent connections. If this parameter is used to distribute the load
8110 according to server's capacity, it is recommended to start with values which
8111 can both grow and shrink, for instance between 10 and 100 to leave enough
8112 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008113
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008114 Supported in default-server: Yes
8115
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008116
81176. HTTP header manipulation
8118---------------------------
8119
8120In HTTP mode, it is possible to rewrite, add or delete some of the request and
8121response headers based on regular expressions. It is also possible to block a
8122request or a response if a particular header matches a regular expression,
8123which is enough to stop most elementary protocol attacks, and to protect
8124against information leak from the internal network. But there is a limitation
8125to this : since HAProxy's HTTP engine does not support keep-alive, only headers
8126passed during the first request of a TCP session will be seen. All subsequent
8127headers will be considered data only and not analyzed. Furthermore, HAProxy
8128never touches data contents, it stops analysis at the end of headers.
8129
Willy Tarreau816b9792009-09-15 21:25:21 +02008130There is an exception though. If HAProxy encounters an "Informational Response"
8131(status code 1xx), it is able to process all rsp* rules which can allow, deny,
8132rewrite or delete a header, but it will refuse to add a header to any such
8133messages as this is not HTTP-compliant. The reason for still processing headers
8134in such responses is to stop and/or fix any possible information leak which may
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008135happen, for instance because another downstream equipment would unconditionally
Willy Tarreau816b9792009-09-15 21:25:21 +02008136add a header, or if a server name appears there. When such messages are seen,
8137normal processing still occurs on the next non-informational messages.
8138
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008139This section covers common usage of the following keywords, described in detail
8140in section 4.2 :
8141
8142 - reqadd <string>
8143 - reqallow <search>
8144 - reqiallow <search>
8145 - reqdel <search>
8146 - reqidel <search>
8147 - reqdeny <search>
8148 - reqideny <search>
8149 - reqpass <search>
8150 - reqipass <search>
8151 - reqrep <search> <replace>
8152 - reqirep <search> <replace>
8153 - reqtarpit <search>
8154 - reqitarpit <search>
8155 - rspadd <string>
8156 - rspdel <search>
8157 - rspidel <search>
8158 - rspdeny <search>
8159 - rspideny <search>
8160 - rsprep <search> <replace>
8161 - rspirep <search> <replace>
8162
8163With all these keywords, the same conventions are used. The <search> parameter
8164is a POSIX extended regular expression (regex) which supports grouping through
8165parenthesis (without the backslash). Spaces and other delimiters must be
8166prefixed with a backslash ('\') to avoid confusion with a field delimiter.
8167Other characters may be prefixed with a backslash to change their meaning :
8168
8169 \t for a tab
8170 \r for a carriage return (CR)
8171 \n for a new line (LF)
8172 \ to mark a space and differentiate it from a delimiter
8173 \# to mark a sharp and differentiate it from a comment
8174 \\ to use a backslash in a regex
8175 \\\\ to use a backslash in the text (*2 for regex, *2 for haproxy)
8176 \xXX to write the ASCII hex code XX as in the C language
8177
8178The <replace> parameter contains the string to be used to replace the largest
8179portion of text matching the regex. It can make use of the special characters
8180above, and can reference a substring which is delimited by parenthesis in the
8181regex, by writing a backslash ('\') immediately followed by one digit from 0 to
81829 indicating the group position (0 designating the entire line). This practice
8183is very common to users of the "sed" program.
8184
8185The <string> parameter represents the string which will systematically be added
8186after the last header line. It can also use special character sequences above.
8187
8188Notes related to these keywords :
8189---------------------------------
8190 - these keywords are not always convenient to allow/deny based on header
8191 contents. It is strongly recommended to use ACLs with the "block" keyword
8192 instead, resulting in far more flexible and manageable rules.
8193
8194 - lines are always considered as a whole. It is not possible to reference
8195 a header name only or a value only. This is important because of the way
8196 headers are written (notably the number of spaces after the colon).
8197
8198 - the first line is always considered as a header, which makes it possible to
8199 rewrite or filter HTTP requests URIs or response codes, but in turn makes
8200 it harder to distinguish between headers and request line. The regex prefix
8201 ^[^\ \t]*[\ \t] matches any HTTP method followed by a space, and the prefix
8202 ^[^ \t:]*: matches any header name followed by a colon.
8203
8204 - for performances reasons, the number of characters added to a request or to
8205 a response is limited at build time to values between 1 and 4 kB. This
8206 should normally be far more than enough for most usages. If it is too short
8207 on occasional usages, it is possible to gain some space by removing some
8208 useless headers before adding new ones.
8209
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008210 - keywords beginning with "reqi" and "rspi" are the same as their counterpart
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008211 without the 'i' letter except that they ignore case when matching patterns.
8212
8213 - when a request passes through a frontend then a backend, all req* rules
8214 from the frontend will be evaluated, then all req* rules from the backend
8215 will be evaluated. The reverse path is applied to responses.
8216
8217 - req* statements are applied after "block" statements, so that "block" is
8218 always the first one, but before "use_backend" in order to permit rewriting
Willy Tarreaud72758d2010-01-12 10:42:19 +01008219 before switching.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008220
8221
Willy Tarreau74ca5042013-06-11 23:12:07 +020082227. Using ACLs and fetching samples
8223----------------------------------
8224
8225Haproxy is capable of extracting data from request or response streams, from
8226client or server information, from tables, environmental information etc...
8227The action of extracting such data is called fetching a sample. Once retrieved,
8228these samples may be used for various purposes such as a key to a stick-table,
8229but most common usages consist in matching them against predefined constant
8230data called patterns.
8231
8232
82337.1. ACL basics
8234---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008235
8236The use of Access Control Lists (ACL) provides a flexible solution to perform
8237content switching and generally to take decisions based on content extracted
8238from the request, the response or any environmental status. The principle is
8239simple :
8240
Willy Tarreau74ca5042013-06-11 23:12:07 +02008241 - extract a data sample from a stream, table or the environment
8242 - apply one or multiple pattern matching methods on this sample
8243 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008244
Willy Tarreau74ca5042013-06-11 23:12:07 +02008245The actions generally consist in blocking a request, selecting a backend, or
8246adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008247
8248In order to define a test, the "acl" keyword is used. The syntax is :
8249
Willy Tarreau74ca5042013-06-11 23:12:07 +02008250 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008251
8252This creates a new ACL <aclname> or completes an existing one with new tests.
8253Those tests apply to the portion of request/response specified in <criterion>
8254and may be adjusted with optional flags [flags]. Some criteria also support
8255an operator which may be specified before the set of values. The values are
8256of the type supported by the criterion, and are separated by spaces.
8257
8258ACL names must be formed from upper and lower case letters, digits, '-' (dash),
8259'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
8260which means that "my_acl" and "My_Acl" are two different ACLs.
8261
8262There is no enforced limit to the number of ACLs. The unused ones do not affect
8263performance, they just consume a small amount of memory.
8264
Willy Tarreau74ca5042013-06-11 23:12:07 +02008265The criterion generally is the name of a sample fetch method, or one of its ACL
8266specific declinations. The default test method is implied by the output type of
8267this sample fetch method. The ACL declinations can describe alternate matching
8268methods of a same sample fetch method.
8269
8270Sample fetch methods return data which can be of the following types :
8271 - boolean
8272 - integer (signed or unsigned)
8273 - IPv4 or IPv6 address
8274 - string
8275 - data block
8276
8277The ACL engine can match these types against patterns of the following types :
8278 - boolean
8279 - integer or integer range
8280 - IP address / network
8281 - string (exact, substring, suffix, prefix, subdir, domain)
8282 - regular expression
8283 - hex block
8284
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008285The following ACL flags are currently supported :
8286
Willy Tarreau2b5285d2010-05-09 23:45:24 +02008287 -i : ignore case during matching of all subsequent patterns.
8288 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +02008289 -m : use a specific pattern matching method
Willy Tarreau6a06a402007-07-15 20:15:28 +02008290 -- : force end of flags. Useful when a string looks like one of the flags.
8291
Willy Tarreau74ca5042013-06-11 23:12:07 +02008292The "-f" flag is followed by the name of a file from which all lines will be
8293read as individual values. It is even possible to pass multiple "-f" arguments
8294if the patterns are to be loaded from multiple files. Empty lines as well as
8295lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
8296will be stripped. If it is absolutely necessary to insert a valid pattern
8297beginning with a sharp, just prefix it with a space so that it is not taken for
8298a comment. Depending on the data type and match method, haproxy may load the
8299lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
8300exact string matching. In this case, duplicates will automatically be removed.
8301
8302Also, note that the "-i" flag applies to subsequent entries and not to entries
8303loaded from files preceding it. For instance :
8304
8305 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
8306
8307In this example, each line of "exact-ua.lst" will be exactly matched against
8308the "user-agent" header of the request. Then each line of "generic-ua" will be
8309case-insensitively matched. Then the word "test" will be insensitively matched
8310as well.
8311
8312The "-m" flag is used to select a specific pattern matching method on the input
8313sample. All ACL-specific criteria imply a pattern matching method and generally
8314do not need this flag. However, this flag is useful with generic sample fetch
8315methods to describe how they're going to be matched against the patterns. This
8316is required for sample fetches which return data type for which there is no
8317obvious matching method (eg: string or binary). When "-m" is specified and
8318followed by a pattern matching method name, this method is used instead of the
8319default one for the criterion. This makes it possible to match contents in ways
8320that were not initially planned, or with sample fetch methods which return a
8321string. The matching method also affects the way the patterns are parsed.
8322
8323There are some restrictions however. Not all methods can be used with all
8324sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
8325be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +02008326
8327 - "found" : only check if the requested sample could be found in the stream,
8328 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +02008329 to pass any pattern to avoid confusion. This matching method is
8330 particularly useful to detect presence of certain contents such
8331 as headers, cookies, etc... even if they are empty and without
8332 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008333
8334 - "bool" : check the value as a boolean. It can only be applied to fetches
8335 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +02008336 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008337
8338 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +02008339 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008340
8341 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +02008342 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008343
8344 - "bin" : match the contents against an hexadecimal string representing a
8345 binary sequence. This may be used with binary or string samples.
8346
8347 - "len" : match the sample's length as an integer. This may be used with
8348 binary or string samples.
8349
Willy Tarreau74ca5042013-06-11 23:12:07 +02008350 - "str" : exact match : match the contents against a string. This may be
8351 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008352
Willy Tarreau74ca5042013-06-11 23:12:07 +02008353 - "sub" : substring match : check that the contents contain at least one of
8354 the provided string patterns. This may be used with binary or
8355 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008356
Willy Tarreau74ca5042013-06-11 23:12:07 +02008357 - "reg" : regex match : match the contents against a list of regular
8358 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008359
Willy Tarreau74ca5042013-06-11 23:12:07 +02008360 - "beg" : prefix match : check that the contents begin like the provided
8361 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008362
Willy Tarreau74ca5042013-06-11 23:12:07 +02008363 - "end" : suffix match : check that the contents end like the provided
8364 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008365
Willy Tarreau74ca5042013-06-11 23:12:07 +02008366 - "dir" : subdir match : check that a slash-delimited portion of the
8367 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008368 This may be used with binary or string samples.
8369
Willy Tarreau74ca5042013-06-11 23:12:07 +02008370 - "dom" : domain match : check that a dot-delimited portion of the contents
8371 exactly match one of the provided string patterns. This may be
8372 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008373
8374For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
8375request, it is possible to do :
8376
8377 acl jsess_present cook(JSESSIONID) -m found
8378
8379In order to apply a regular expression on the 500 first bytes of data in the
8380buffer, one would use the following acl :
8381
8382 acl script_tag payload(0,500) -m reg -i <script>
8383
Willy Tarreau74ca5042013-06-11 23:12:07 +02008384All ACL-specific criteria imply a default matching method. Most often, these
8385criteria are composed by concatenating the name of the original sample fetch
8386method and the matching method. For example, "hdr_beg" applies the "beg" match
8387to samples retrieved using the "hdr" fetch method. Since all ACL-specific
8388criteria rely on a sample fetch method, it is always possible instead to use
8389the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +02008390
Willy Tarreau74ca5042013-06-11 23:12:07 +02008391If an alternate match is specified using "-m" on an ACL-specific criterion,
8392the mathing method is simply applied to the underlying sample fetch method. For
8393example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +02008394
Willy Tarreau74ca5042013-06-11 23:12:07 +02008395 acl short_form hdr_beg(host) www.
8396 acl alternate1 hdr_beg(host) -m beg www.
8397 acl alternate2 hdr_dom(host) -m beg www.
8398 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +02008399
Willy Tarreau2b5285d2010-05-09 23:45:24 +02008400
Willy Tarreau74ca5042013-06-11 23:12:07 +02008401The table below summarizes the compatibility matrix between sample fetch types
8402and the pattern types to fetch against. It indicates for each compatible
8403combination the name of the matching method to be used, prefixed with "*" when
8404the method is implicit and will work by default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +01008405
Willy Tarreau74ca5042013-06-11 23:12:07 +02008406 +-------------------------------------------------+
8407 | Input sample type |
8408 +----------------------+---------+---------+---------+---------+---------+
8409 | pattern type | boolean | integer | IP | string | binary |
8410 +----------------------+---------+---------+---------+---------+---------+
8411 | none (presence only) | found | found | found | found | found |
8412 +----------------------+---------+---------+---------+---------+---------+
8413 | none (boolean value) | *bool | bool | | | |
8414 +----------------------+---------+---------+---------+---------+---------+
8415 | integer (value) | int | *int | | | |
8416 +----------------------+---------+---------+---------+---------+---------+
8417 | integer (length) | | | | len | len |
8418 +----------------------+---------+---------+---------+---------+---------+
8419 | IP address | | | *ip | | |
8420 +----------------------+---------+---------+---------+---------+---------+
8421 | exact string | | | | str | str |
8422 +----------------------+---------+---------+---------+---------+---------+
8423 | prefix | | | | beg | beg |
8424 +----------------------+---------+---------+---------+---------+---------+
8425 | suffix | | | | end | end |
8426 +----------------------+---------+---------+---------+---------+---------+
8427 | substring | | | | sub | sub |
8428 +----------------------+---------+---------+---------+---------+---------+
8429 | subdir | | | | dir | dir |
8430 +----------------------+---------+---------+---------+---------+---------+
8431 | domain | | | | dom | dom |
8432 +----------------------+---------+---------+---------+---------+---------+
8433 | regex | | | | reg | reg |
8434 +----------------------+---------+---------+---------+---------+---------+
8435 | hex block | | | | bin | bin |
8436 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +02008437
8438
Willy Tarreau74ca5042013-06-11 23:12:07 +020084397.1.1. Matching booleans
8440------------------------
8441
8442In order to match a boolean, no value is needed and all values are ignored.
8443Boolean matching is used by default for all fetch methods of type "boolean".
8444When boolean matching is used, the fetched value is returned as-is, which means
8445that a boolean "true" will always match and a boolean "false" will never match.
8446
8447Boolean matching may also be enforced using "-m bool" on fetch methods which
8448return an integer value. Then, integer value 0 is converted to the boolean
8449"false" and all other values are converted to "true".
8450
Willy Tarreau6a06a402007-07-15 20:15:28 +02008451
Willy Tarreau74ca5042013-06-11 23:12:07 +020084527.1.2. Matching integers
8453------------------------
8454
8455Integer matching applies by default to integer fetch methods. It can also be
8456enforced on boolean fetches using "-m int". In this case, "false" is converted
8457to the integer 0, and "true" is converted to the integer 1.
8458
8459Integer matching also supports integer ranges and operators. Note that integer
8460matching only applies to positive values. A range is a value expressed with a
8461lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +02008462
8463For instance, "1024:65535" is a valid range to represent a range of
8464unprivileged ports, and "1024:" would also work. "0:1023" is a valid
8465representation of privileged ports, and ":1023" would also work.
8466
Willy Tarreau62644772008-07-16 18:36:06 +02008467As a special case, some ACL functions support decimal numbers which are in fact
8468two integers separated by a dot. This is used with some version checks for
8469instance. All integer properties apply to those decimal numbers, including
8470ranges and operators.
8471
Willy Tarreau6a06a402007-07-15 20:15:28 +02008472For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +01008473operators with ranges does not make much sense and is strongly discouraged.
8474Similarly, it does not make much sense to perform order comparisons with a set
8475of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +02008476
Willy Tarreau0ba27502007-12-24 16:55:16 +01008477Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +02008478
8479 eq : true if the tested value equals at least one value
8480 ge : true if the tested value is greater than or equal to at least one value
8481 gt : true if the tested value is greater than at least one value
8482 le : true if the tested value is less than or equal to at least one value
8483 lt : true if the tested value is less than at least one value
8484
Willy Tarreau0ba27502007-12-24 16:55:16 +01008485For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +02008486
8487 acl negative-length hdr_val(content-length) lt 0
8488
Willy Tarreau62644772008-07-16 18:36:06 +02008489This one matches SSL versions between 3.0 and 3.1 (inclusive) :
8490
8491 acl sslv3 req_ssl_ver 3:3.1
8492
Willy Tarreau6a06a402007-07-15 20:15:28 +02008493
Willy Tarreau74ca5042013-06-11 23:12:07 +020084947.1.3. Matching strings
8495-----------------------
8496
8497String matching applies to string or binary fetch methods, and exists in 6
8498different forms :
8499
8500 - exact match (-m str) : the extracted string must exactly match the
8501 patterns ;
8502
8503 - substring match (-m sub) : the patterns are looked up inside the
8504 extracted string, and the ACL matches if any of them is found inside ;
8505
8506 - prefix match (-m beg) : the patterns are compared with the beginning of
8507 the extracted string, and the ACL matches if any of them matches.
8508
8509 - suffix match (-m end) : the patterns are compared with the end of the
8510 extracted string, and the ACL matches if any of them matches.
8511
8512 - subdir match (-m sub) : the patterns are looked up inside the extracted
8513 string, delimited with slashes ("/"), and the ACL matches if any of them
8514 matches.
8515
8516 - domain match (-m dom) : the patterns are looked up inside the extracted
8517 string, delimited with dots ("."), and the ACL matches if any of them
8518 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +02008519
8520String matching applies to verbatim strings as they are passed, with the
8521exception of the backslash ("\") which makes it possible to escape some
8522characters such as the space. If the "-i" flag is passed before the first
8523string, then the matching will be performed ignoring the case. In order
8524to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +01008525before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +02008526
8527
Willy Tarreau74ca5042013-06-11 23:12:07 +020085287.1.4. Matching regular expressions (regexes)
8529---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02008530
8531Just like with string matching, regex matching applies to verbatim strings as
8532they are passed, with the exception of the backslash ("\") which makes it
8533possible to escape some characters such as the space. If the "-i" flag is
8534passed before the first regex, then the matching will be performed ignoring
8535the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +01008536the "--" flag before the first string. Same principle applies of course to
8537match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +02008538
8539
Willy Tarreau74ca5042013-06-11 23:12:07 +020085407.1.5. Matching arbitrary data blocks
8541-------------------------------------
8542
8543It is possible to match some extracted samples against a binary block which may
8544not safely be represented as a string. For this, the patterns must be passed as
8545a series of hexadecimal digits in an even number, when the match method is set
8546to binary. Each sequence of two digits will represent a byte. The hexadecimal
8547digits may be used upper or lower case.
8548
8549Example :
8550 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
8551 acl hello payload(0,6) -m bin 48656c6c6f0a
8552
8553
85547.1.6. Matching IPv4 and IPv6 addresses
8555---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02008556
8557IPv4 addresses values can be specified either as plain addresses or with a
8558netmask appended, in which case the IPv4 address matches whenever it is
8559within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01008560host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +01008561difficult to read and debug configurations. If hostnames are used, you should
8562at least ensure that they are present in /etc/hosts so that the configuration
8563does not depend on any random DNS match at the moment the configuration is
8564parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +02008565
Willy Tarreauceb4ac92012-04-28 00:41:46 +02008566IPv6 may be entered in their usual form, with or without a netmask appended.
8567Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
8568trouble with randomly resolved IP addresses, host names are never allowed in
8569IPv6 patterns.
8570
8571HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
8572following situations :
8573 - tested address is IPv4, pattern address is IPv4, the match applies
8574 in IPv4 using the supplied mask if any.
8575 - tested address is IPv6, pattern address is IPv6, the match applies
8576 in IPv6 using the supplied mask if any.
8577 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
8578 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
8579 ::IPV4 or ::ffff:IPV4, otherwise it fails.
8580 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
8581 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
8582 applied in IPv6 using the supplied IPv6 mask.
8583
Willy Tarreau74ca5042013-06-11 23:12:07 +02008584
85857.2. Using ACLs to form conditions
8586----------------------------------
8587
8588Some actions are only performed upon a valid condition. A condition is a
8589combination of ACLs with operators. 3 operators are supported :
8590
8591 - AND (implicit)
8592 - OR (explicit with the "or" keyword or the "||" operator)
8593 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +02008594
Willy Tarreau74ca5042013-06-11 23:12:07 +02008595A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +02008596
Willy Tarreau74ca5042013-06-11 23:12:07 +02008597 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +02008598
Willy Tarreau74ca5042013-06-11 23:12:07 +02008599Such conditions are generally used after an "if" or "unless" statement,
8600indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +02008601
Willy Tarreau74ca5042013-06-11 23:12:07 +02008602For instance, to block HTTP requests to the "*" URL with methods other than
8603"OPTIONS", as well as POST requests without content-length, and GET or HEAD
8604requests with a content-length greater than 0, and finally every request which
8605is not either GET/HEAD/POST/OPTIONS !
8606
8607 acl missing_cl hdr_cnt(Content-length) eq 0
8608 block if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
8609 block if METH_GET HTTP_CONTENT
8610 block unless METH_GET or METH_POST or METH_OPTIONS
8611
8612To select a different backend for requests to static contents on the "www" site
8613and to every request on the "img", "video", "download" and "ftp" hosts :
8614
8615 acl url_static path_beg /static /images /img /css
8616 acl url_static path_end .gif .png .jpg .css .js
8617 acl host_www hdr_beg(host) -i www
8618 acl host_static hdr_beg(host) -i img. video. download. ftp.
8619
8620 # now use backend "static" for all static-only hosts, and for static urls
8621 # of host "www". Use backend "www" for the rest.
8622 use_backend static if host_static or host_www url_static
8623 use_backend www if host_www
8624
8625It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
8626expressions that are built on the fly without needing to be declared. They must
8627be enclosed between braces, with a space before and after each brace (because
8628the braces must be seen as independent words). Example :
8629
8630 The following rule :
8631
8632 acl missing_cl hdr_cnt(Content-length) eq 0
8633 block if METH_POST missing_cl
8634
8635 Can also be written that way :
8636
8637 block if METH_POST { hdr_cnt(Content-length) eq 0 }
8638
8639It is generally not recommended to use this construct because it's a lot easier
8640to leave errors in the configuration when written that way. However, for very
8641simple rules matching only one source IP address for instance, it can make more
8642sense to use them than to declare ACLs with random names. Another example of
8643good use is the following :
8644
8645 With named ACLs :
8646
8647 acl site_dead nbsrv(dynamic) lt 2
8648 acl site_dead nbsrv(static) lt 2
8649 monitor fail if site_dead
8650
8651 With anonymous ACLs :
8652
8653 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
8654
8655See section 4.2 for detailed help on the "block" and "use_backend" keywords.
8656
8657
86587.3. Fetching samples
8659---------------------
8660
8661Historically, sample fetch methods were only used to retrieve data to match
8662against patterns using ACLs. With the arrival of stick-tables, a new class of
8663sample fetch methods was created, most often sharing the same syntax as their
8664ACL counterpart. These sample fetch methods are also known as "fetches". As
8665of now, ACLs and fetches have converged. All ACL fetch methods have been made
8666available as fetch methods, and ACLs may use any sample fetch method as well.
8667
8668This section details all available sample fetch methods and their output type.
8669Some sample fetch methods have deprecated aliases that are used to maintain
8670compatibility with existing configurations. They are then explicitly marked as
8671deprecated and should not be used in new setups.
8672
8673The ACL derivatives are also indicated when available, with their respective
8674matching methods. These ones all have a well defined default pattern matching
8675method, so it is never necessary (though allowed) to pass the "-m" option to
8676indicate how the sample will be matched using ACLs.
8677
8678As indicated in the sample type versus matching compatibility matrix above,
8679when using a generic sample fetch method in an ACL, the "-m" option is
8680mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
8681the same keyword exists as an ACL keyword and as a standard fetch method, the
8682ACL engine will automatically pick the ACL-only one by default.
8683
8684Some of these keywords support one or multiple mandatory arguments, and one or
8685multiple optional arguments. These arguments are strongly typed and are checked
8686when the configuration is parsed so that there is no risk of running with an
8687incorrect argument (eg: an unresolved backend name). Fetch function arguments
8688are passed between parenthesis and are delimited by commas. When an argument
8689is optional, it will be indicated below between square brackets ('[ ]'). When
8690all arguments are optional, the parenthesis may be omitted.
8691
8692Thus, the syntax of a standard sample fetch method is one of the following :
8693 - name
8694 - name(arg1)
8695 - name(arg1,arg2)
8696
8697At the moment, the stickiness features are the most advanced users of the
8698sample fetch system. The "stick on", and "stick store-request" directives
8699support sample fetch rules which allow a list of transformations to be applied
8700on top of the fetched sample, and the finaly result is automatically converted
8701to the type of the table. These transformations are enumerated as a series
Willy Tarreau833cc792013-07-24 15:34:19 +02008702of specific keywords after the sample fetch method. These keywords may equally
8703be appended immediately after the fetch keyword's argument, delimited by a
8704comma. These keywords can also support some arguments (eg: a netmask) which
8705must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +01008706
Willy Tarreau74ca5042013-06-11 23:12:07 +02008707The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +01008708
Willy Tarreau74ca5042013-06-11 23:12:07 +02008709 lower Convert a string sample to lower case. This can only be placed
8710 after a string sample fetch function or after a transformation
8711 keyword returning a string type. The result is of type string.
Willy Tarreau6a06a402007-07-15 20:15:28 +02008712
Willy Tarreau74ca5042013-06-11 23:12:07 +02008713 upper Convert a string sample to upper case. This can only be placed
8714 after a string sample fetch function or after a transformation
8715 keyword returning a string type. The result is of type string.
Willy Tarreau6a06a402007-07-15 20:15:28 +02008716
Willy Tarreau74ca5042013-06-11 23:12:07 +02008717 ipmask(<mask>) Apply a mask to an IPv4 address, and use the result for
8718 lookups and storage. This can be used to make all hosts within
8719 a certain mask to share the same table entries and as such use
8720 the same server. The mask can be passed in dotted form (eg:
8721 255.255.255.0) or in CIDR form (eg: 24).
8722
Willy Tarreau276fae92013-07-25 14:36:01 +02008723 http_date([<offset>])
8724 Converts an integer supposed to contain a date since epoch to
8725 a string representing this date in a format suitable for use
8726 in HTTP header fields. If an offset value is specified, then
8727 it is a number of seconds that is added to the date before the
8728 conversion is operated. This is particularly useful to emit
8729 Date header fields, Expires values in responses when combined
8730 with a positive offset, or Last-Modified values when the
8731 offset is negative.
Willy Tarreau74ca5042013-06-11 23:12:07 +02008732
87337.3.1. Fetching samples from internal states
8734--------------------------------------------
8735
8736A first set of sample fetch methods applies to internal information which does
8737not even relate to any client information. These ones are sometimes used with
8738"monitor-fail" directives to report an internal status to external watchers.
8739The sample fetch methods described in this section are usable anywhere.
8740
8741always_false : boolean
8742 Always returns the boolean "false" value. It may be used with ACLs as a
8743 temporary replacement for another one when adjusting configurations.
8744
8745always_true : boolean
8746 Always returns the boolean "true" value. It may be used with ACLs as a
8747 temporary replacement for another one when adjusting configurations.
8748
8749avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +01008750 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +02008751 divided by the number of active servers. The current backend is used if no
8752 backend is specified. This is very similar to "queue" except that the size of
8753 the farm is considered, in order to give a more accurate measurement of the
8754 time it may take for a new connection to be processed. The main usage is with
8755 ACL to return a sorry page to new users when it becomes certain they will get
8756 a degraded service, or to pass to the backend servers in a header so that
8757 they decide to work in degraded mode or to disable some functions to speed up
8758 the processing a bit. Note that in the event there would not be any active
8759 server anymore, twice the number of queued connections would be considered as
8760 the measured value. This is a fair estimate, as we expect one server to get
8761 back soon anyway, but we still prefer to send new traffic to another backend
8762 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
8763 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +01008764
Willy Tarreau74ca5042013-06-11 23:12:07 +02008765be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +02008766 Applies to the number of currently established connections on the backend,
8767 possibly including the connection being evaluated. If no backend name is
8768 specified, the current one is used. But it is also possible to check another
8769 backend. It can be used to use a specific farm when the nominal one is full.
8770 See also the "fe_conn", "queue" and "be_sess_rate" criteria.
Willy Tarreau6a06a402007-07-15 20:15:28 +02008771
Willy Tarreau74ca5042013-06-11 23:12:07 +02008772be_sess_rate([<backend>]) : integer
8773 Returns an integer value corresponding to the sessions creation rate on the
8774 backend, in number of new sessions per second. This is used with ACLs to
8775 switch to an alternate backend when an expensive or fragile one reaches too
8776 high a session rate, or to limit abuse of service (eg. prevent sucking of an
8777 online dictionary). It can also be useful to add this element to logs using a
8778 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008779
8780 Example :
8781 # Redirect to an error page if the dictionary is requested too often
8782 backend dynamic
8783 mode http
8784 acl being_scanned be_sess_rate gt 100
8785 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +01008786
Willy Tarreau74ca5042013-06-11 23:12:07 +02008787connslots([<backend>]) : integer
8788 Returns an integer value corresponding to the number of connection slots
8789 still available in the backend, by totalizing the maximum amount of
8790 connections on all servers and the maximum queue size. This is probably only
8791 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -05008792
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +08008793 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +02008794 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +08008795 usage; see "use_backend" keyword) can be redirected to a different backend.
8796
Willy Tarreau55165fe2009-05-10 12:02:55 +02008797 'connslots' = number of available server connection slots, + number of
8798 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +08008799
Willy Tarreaua36af912009-10-10 12:02:45 +02008800 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +02008801 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +02008802 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +02008803 you want to be able to differentiate between different backends, and their
8804 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +02008805 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +02008806 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +08008807
Willy Tarreau55165fe2009-05-10 12:02:55 +02008808 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
8809 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +02008810 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +02008811 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +08008812
Willy Tarreau6236d3a2013-07-25 14:28:25 +02008813date([<offset>]) : integer
8814 Returns the current date as the epoch (number of seconds since 01/01/1970).
8815 If an offset value is specified, then it is a number of seconds that is added
8816 to the current date before returning the value. This is particularly useful
8817 to compute relative dates, as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +02008818 It is useful combined with the http_date converter.
8819
8820 Example :
8821
8822 # set an expires header to now+1 hour in every response
8823 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +02008824
Willy Tarreau595ec542013-06-12 21:34:28 +02008825env(<name>) : string
8826 Returns a string containing the value of environment variable <name>. As a
8827 reminder, environment variables are per-process and are sampled when the
8828 process starts. This can be useful to pass some information to a next hop
8829 server, or with ACLs to take specific action when the process is started a
8830 certain way.
8831
8832 Examples :
8833 # Pass the Via header to next hop with the local hostname in it
8834 http-request add-header Via 1.1\ %[env(HOSTNAME)]
8835
8836 # reject cookie-less requests when the STOP environment variable is set
8837 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
8838
Willy Tarreau74ca5042013-06-11 23:12:07 +02008839fe_conn([<frontend>]) : integer
8840 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +01008841 possibly including the connection being evaluated. If no frontend name is
8842 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +02008843 frontend. It can be used to return a sorry page before hard-blocking, or to
8844 use a specific backend to drain new requests when the farm is considered
8845 full. This is mostly used with ACLs but can also be used to pass some
8846 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
8847 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +02008848
Willy Tarreau74ca5042013-06-11 23:12:07 +02008849fe_sess_rate([<frontend>]) : integer
8850 Returns an integer value corresponding to the sessions creation rate on the
8851 frontend, in number of new sessions per second. This is used with ACLs to
8852 limit the incoming session rate to an acceptable range in order to prevent
8853 abuse of service at the earliest moment, for example when combined with other
8854 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
8855 down below the limit. It can also be useful to add this element to logs using
8856 a log-format directive. See also the "rate-limit sessions" directive for use
8857 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +01008858
8859 Example :
8860 # This frontend limits incoming mails to 10/s with a max of 100
8861 # concurrent connections. We accept any connection below 10/s, and
8862 # force excess clients to wait for 100 ms. Since clients are limited to
8863 # 100 max, there cannot be more than 10 incoming mails per second.
8864 frontend mail
8865 bind :25
8866 mode tcp
8867 maxconn 100
8868 acl too_fast fe_sess_rate ge 10
8869 tcp-request inspect-delay 100ms
8870 tcp-request content accept if ! too_fast
8871 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +01008872
Willy Tarreau74ca5042013-06-11 23:12:07 +02008873nbsrv([<backend>]) : integer
8874 Returns an integer value corresponding to the number of usable servers of
8875 either the current backend or the named backend. This is mostly used with
8876 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +01008877 switch to an alternate backend when the number of servers is too low to
8878 to handle some load. It is useful to report a failure when combined with
8879 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +01008880
Willy Tarreau74ca5042013-06-11 23:12:07 +02008881queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +01008882 Returns the total number of queued connections of the designated backend,
8883 including all the connections in server queues. If no backend name is
8884 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +02008885 one. This is useful with ACLs or to pass statistics to backend servers. This
8886 can be used to take actions when queuing goes above a known level, generally
8887 indicating a surge of traffic or a massive slowdown on the servers. One
8888 possible action could be to reject new users but still accept old ones. See
8889 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
8890
8891srv_conn([<backend>/]<server>) : integer
8892 Returns an integer value corresponding to the number of currently established
8893 connections on the designated server, possibly including the connection being
8894 evaluated. If <backend> is omitted, then the server is looked up in the
8895 current backend. It can be used to use a specific farm when one server is
8896 full, or to inform the server about our view of the number of active
8897 connections with it. See also the "fe_conn", "be_conn" and "queue" fetch
8898 methods.
8899
8900srv_is_up([<backend>/]<server>) : boolean
8901 Returns true when the designated server is UP, and false when it is either
8902 DOWN or in maintenance mode. If <backend> is omitted, then the server is
8903 looked up in the current backend. It is mainly used to take action based on
8904 an external status reported via a health check (eg: a geographical site's
8905 availability). Another possible use which is more of a hack consists in
8906 using dummy servers as boolean variables that can be enabled or disabled from
8907 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
8908
8909srv_sess_rate([<backend>/]<server>) : integer
8910 Returns an integer corresponding to the sessions creation rate on the
8911 designated server, in number of new sessions per second. If <backend> is
8912 omitted, then the server is looked up in the current backend. This is mosly
8913 used with ACLs but can make sense with logs too. This is used to switch to an
8914 alternate backend when an expensive or fragile one reaches too high a session
8915 rate, or to limit abuse of service (eg. prevent latent requests from
8916 overloading servers).
8917
8918 Example :
8919 # Redirect to a separate back
8920 acl srv1_full srv_sess_rate(be1/srv1) gt 50
8921 acl srv2_full srv_sess_rate(be1/srv2) gt 50
8922 use_backend be2 if srv1_full or srv2_full
8923
8924table_avl([<table>]) : integer
8925 Returns the total number of available entries in the current proxy's
8926 stick-table or in the designated stick-table. See also table_cnt.
8927
8928table_cnt([<table>]) : integer
8929 Returns the total number of entries currently in use in the current proxy's
8930 stick-table or in the designated stick-table. See also src_conn_cnt and
8931 table_avl for other entry counting methods.
8932
8933
89347.3.2. Fetching samples at Layer 4
8935----------------------------------
8936
8937The layer 4 usually describes just the transport layer which in haproxy is
8938closest to the connection, where no content is yet made available. The fetch
8939methods described here are usable as low as the "tcp-request connection" rule
8940sets unless they require some future information. Those generally include
8941TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +02008942the incoming connection. For retrieving a value from a sticky counters, the
8943counter number can be explicitly set as 0, 1, or 2 using the pre-defined
8944"sc0_", "sc1_", or "sc2_" prefix, or it can be specified as the first integer
Willy Tarreau0f791d42013-07-23 19:56:43 +02008945argument when using the "sc_" prefix. An optional table may be specified with
8946the "sc*" form, in which case the currently tracked key will be looked up into
8947this alternate table instead of the table currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +02008948
8949be_id : integer
8950 Returns an integer containing the current backend's id. It can be used in
8951 frontends with responses to check which backend processed the request.
8952
8953dst : ip
8954 This is the destination IPv4 address of the connection on the client side,
8955 which is the address the client connected to. It can be useful when running
8956 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
8957 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
8958 RFC 4291.
8959
8960dst_conn : integer
8961 Returns an integer value corresponding to the number of currently established
8962 connections on the same socket including the one being evaluated. It is
8963 normally used with ACLs but can as well be used to pass the information to
8964 servers in an HTTP header or in logs. It can be used to either return a sorry
8965 page before hard-blocking, or to use a specific backend to drain new requests
8966 when the socket is considered saturated. This offers the ability to assign
8967 different limits to different listening ports or addresses. See also the
8968 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008969
Willy Tarreau74ca5042013-06-11 23:12:07 +02008970dst_port : integer
8971 Returns an integer value corresponding to the destination TCP port of the
8972 connection on the client side, which is the port the client connected to.
8973 This might be used when running in transparent mode, when assigning dynamic
8974 ports to some clients for a whole application session, to stick all users to
8975 a same server, or to pass the destination port information to a server using
8976 an HTTP header.
8977
8978fe_id : integer
8979 Returns an integer containing the current frontend's id. It can be used in
8980 backends to check from which backend it was called, or to stick all users
8981 coming via a same frontend to the same server.
8982
Willy Tarreau0f791d42013-07-23 19:56:43 +02008983sc_bytes_in_rate(<ctr>,[<table>]) : integer
8984sc0_bytes_in_rate([<table>]) : integer
8985sc1_bytes_in_rate([<table>]) : integer
8986sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02008987 Returns the average client-to-server bytes rate from the currently tracked
8988 counters, measured in amount of bytes over the period configured in the
8989 table. See also src_bytes_in_rate.
8990
Willy Tarreau0f791d42013-07-23 19:56:43 +02008991sc_bytes_out_rate(<ctr>,[<table>]) : integer
8992sc0_bytes_out_rate([<table>]) : integer
8993sc1_bytes_out_rate([<table>]) : integer
8994sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02008995 Returns the average server-to-client bytes rate from the currently tracked
8996 counters, measured in amount of bytes over the period configured in the
8997 table. See also src_bytes_out_rate.
8998
Willy Tarreau0f791d42013-07-23 19:56:43 +02008999sc_clr_gpc0(<ctr>,[<table>]) : integer
9000sc0_clr_gpc0([<table>]) : integer
9001sc1_clr_gpc0([<table>]) : integer
9002sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +02009003 Clears the first General Purpose Counter associated to the currently tracked
9004 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +01009005 stored value is zero, so first invocation will always return zero. This is
9006 typically used as a second ACL in an expression in order to mark a connection
9007 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +02009008
9009 # block if 5 consecutive requests continue to come faster than 10 sess
9010 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009011 acl abuse sc0_http_req_rate gt 10
9012 acl kill sc0_inc_gpc0 gt 5
9013 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +02009014 tcp-request connection accept if !abuse save
9015 tcp-request connection reject if abuse kill
9016
Willy Tarreau0f791d42013-07-23 19:56:43 +02009017sc_conn_cnt(<ctr>,[<table>]) : integer
9018sc0_conn_cnt([<table>]) : integer
9019sc1_conn_cnt([<table>]) : integer
9020sc2_conn_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009021 Returns the cumulated number of incoming connections from currently tracked
9022 counters. See also src_conn_cnt.
9023
Willy Tarreau0f791d42013-07-23 19:56:43 +02009024sc_conn_cur(<ctr>,[<table>]) : integer
9025sc0_conn_cur([<table>]) : integer
9026sc1_conn_cur([<table>]) : integer
9027sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009028 Returns the current amount of concurrent connections tracking the same
9029 tracked counters. This number is automatically incremented when tracking
9030 begins and decremented when tracking stops. See also src_conn_cur.
9031
Willy Tarreau0f791d42013-07-23 19:56:43 +02009032sc_conn_rate(<ctr>,[<table>]) : integer
9033sc0_conn_rate([<table>]) : integer
9034sc1_conn_rate([<table>]) : integer
9035sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009036 Returns the average connection rate from the currently tracked counters,
9037 measured in amount of connections over the period configured in the table.
9038 See also src_conn_rate.
9039
Willy Tarreau0f791d42013-07-23 19:56:43 +02009040sc_get_gpc0(<ctr>,[<table>]) : integer
9041sc0_get_gpc0([<table>]) : integer
9042sc1_get_gpc0([<table>]) : integer
9043sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009044 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009045 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009046
Willy Tarreau0f791d42013-07-23 19:56:43 +02009047sc_gpc0_rate(<ctr>,[<table>]) : integer
9048sc0_gpc0_rate([<table>]) : integer
9049sc1_gpc0_rate([<table>]) : integer
9050sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009051 Returns the average increment rate of the first General Purpose Counter
9052 associated to the currently tracked counters. It reports the frequency
9053 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009054 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
9055 that the "gpc0_rate" counter must be stored in the stick-table for a value to
9056 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +02009057
Willy Tarreau0f791d42013-07-23 19:56:43 +02009058sc_http_err_cnt(<ctr>,[<table>]) : integer
9059sc0_http_err_cnt([<table>]) : integer
9060sc1_http_err_cnt([<table>]) : integer
9061sc2_http_err_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009062 Returns the cumulated number of HTTP errors from the currently tracked
9063 counters. This includes the both request errors and 4xx error responses.
9064 See also src_http_err_cnt.
9065
Willy Tarreau0f791d42013-07-23 19:56:43 +02009066sc_http_err_rate(<ctr>,[<table>]) : integer
9067sc0_http_err_rate([<table>]) : integer
9068sc1_http_err_rate([<table>]) : integer
9069sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009070 Returns the average rate of HTTP errors from the currently tracked counters,
9071 measured in amount of errors over the period configured in the table. This
9072 includes the both request errors and 4xx error responses. See also
9073 src_http_err_rate.
9074
Willy Tarreau0f791d42013-07-23 19:56:43 +02009075sc_http_req_cnt(<ctr>,[<table>]) : integer
9076sc0_http_req_cnt([<table>]) : integer
9077sc1_http_req_cnt([<table>]) : integer
9078sc2_http_req_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009079 Returns the cumulated number of HTTP requests from the currently tracked
9080 counters. This includes every started request, valid or not. See also
9081 src_http_req_cnt.
9082
Willy Tarreau0f791d42013-07-23 19:56:43 +02009083sc_http_req_rate(<ctr>,[<table>]) : integer
9084sc0_http_req_rate([<table>]) : integer
9085sc1_http_req_rate([<table>]) : integer
9086sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009087 Returns the average rate of HTTP requests from the currently tracked
9088 counters, measured in amount of requests over the period configured in
9089 the table. This includes every started request, valid or not. See also
9090 src_http_req_rate.
9091
Willy Tarreau0f791d42013-07-23 19:56:43 +02009092sc_inc_gpc0(<ctr>,[<table>]) : integer
9093sc0_inc_gpc0([<table>]) : integer
9094sc1_inc_gpc0([<table>]) : integer
9095sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009096 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +01009097 tracked counters, and returns its new value. Before the first invocation,
9098 the stored value is zero, so first invocation will increase it to 1 and will
9099 return 1. This is typically used as a second ACL in an expression in order
9100 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +02009101
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009102 acl abuse sc0_http_req_rate gt 10
9103 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +02009104 tcp-request connection reject if abuse kill
9105
Willy Tarreau0f791d42013-07-23 19:56:43 +02009106sc_kbytes_in(<ctr>,[<table>]) : integer
9107sc0_kbytes_in([<table>]) : integer
9108sc1_kbytes_in([<table>]) : integer
9109sc2_kbytes_in([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009110 Returns the amount of client-to-server data from the currently tracked
9111 counters, measured in kilobytes over the period configured in the table. The
9112 test is currently performed on 32-bit integers, which limits values to 4
9113 terabytes. See also src_kbytes_in.
9114
Willy Tarreau0f791d42013-07-23 19:56:43 +02009115sc_kbytes_out(<ctr>,[<table>]) : integer
9116sc0_kbytes_out([<table>]) : integer
9117sc1_kbytes_out([<table>]) : integer
9118sc2_kbytes_out([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009119 Returns the amount of server-to-client data from the currently tracked
9120 counters, measured in kilobytes over the period configured in the table. The
9121 test is currently performed on 32-bit integers, which limits values to 4
9122 terabytes. See also src_kbytes_out.
9123
Willy Tarreau0f791d42013-07-23 19:56:43 +02009124sc_sess_cnt(<ctr>,[<table>]) : integer
9125sc0_sess_cnt([<table>]) : integer
9126sc1_sess_cnt([<table>]) : integer
9127sc2_sess_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009128 Returns the cumulated number of incoming connections that were transformed
9129 into sessions, which means that they were accepted by a "tcp-request
9130 connection" rule, from the currently tracked counters. A backend may count
9131 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009132 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +02009133 with the client. See also src_sess_cnt.
9134
Willy Tarreau0f791d42013-07-23 19:56:43 +02009135sc_sess_rate(<ctr>,[<table>]) : integer
9136sc0_sess_rate([<table>]) : integer
9137sc1_sess_rate([<table>]) : integer
9138sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009139 Returns the average session rate from the currently tracked counters,
9140 measured in amount of sessions over the period configured in the table. A
9141 session is a connection that got past the early "tcp-request connection"
9142 rules. A backend may count more sessions than connections because each
9143 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009144 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +02009145
Willy Tarreau0f791d42013-07-23 19:56:43 +02009146sc_tracked(<ctr>,[<table>]) : boolean
9147sc0_tracked([<table>]) : boolean
9148sc1_tracked([<table>]) : boolean
9149sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +02009150 Returns true if the designated session counter is currently being tracked by
9151 the current session. This can be useful when deciding whether or not we want
9152 to set some values in a header passed to the server.
9153
Willy Tarreau0f791d42013-07-23 19:56:43 +02009154sc_trackers(<ctr>,[<table>]) : integer
9155sc0_trackers([<table>]) : integer
9156sc1_trackers([<table>]) : integer
9157sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +01009158 Returns the current amount of concurrent connections tracking the same
9159 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009160 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +01009161 that it does not rely on any stored information but on the table's reference
9162 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +02009163 may sometimes be more suited for layer7 tracking. It can be used to tell a
9164 server how many concurrent connections there are from a given address for
9165 example.
Willy Tarreau2406db42012-12-09 12:16:43 +01009166
Willy Tarreau74ca5042013-06-11 23:12:07 +02009167so_id : integer
9168 Returns an integer containing the current listening socket's id. It is useful
9169 in frontends involving many "bind" lines, or to stick all users coming via a
9170 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009171
Willy Tarreau74ca5042013-06-11 23:12:07 +02009172src : ip
9173 This is the source IPv4 address of the client of the session. It is of type
9174 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
9175 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
9176 TCP-level source address which is used, and not the address of a client
9177 behind a proxy. However if the "accept-proxy" bind directive is used, it can
9178 be the address of a client behind another PROXY-protocol compatible component
9179 for all rule sets except "tcp-request connection" which sees the real address.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009180
Willy Tarreau74ca5042013-06-11 23:12:07 +02009181src_bytes_in_rate([<table>]) : integer
9182 Returns the average bytes rate from the incoming connection's source address
9183 in the current proxy's stick-table or in the designated stick-table, measured
9184 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009185 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009186
Willy Tarreau74ca5042013-06-11 23:12:07 +02009187src_bytes_out_rate([<table>]) : integer
9188 Returns the average bytes rate to the incoming connection's source address in
9189 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +02009190 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009191 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009192
Willy Tarreau74ca5042013-06-11 23:12:07 +02009193src_clr_gpc0([<table>]) : integer
9194 Clears the first General Purpose Counter associated to the incoming
9195 connection's source address in the current proxy's stick-table or in the
9196 designated stick-table, and returns its previous value. If the address is not
9197 found, an entry is created and 0 is returned. This is typically used as a
9198 second ACL in an expression in order to mark a connection when a first ACL
9199 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +02009200
9201 # block if 5 consecutive requests continue to come faster than 10 sess
9202 # per second, and reset the counter as soon as the traffic slows down.
9203 acl abuse src_http_req_rate gt 10
9204 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +01009205 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +02009206 tcp-request connection accept if !abuse save
9207 tcp-request connection reject if abuse kill
9208
Willy Tarreau74ca5042013-06-11 23:12:07 +02009209src_conn_cnt([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +02009210 Returns the cumulated number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +02009211 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +02009212 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009213 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009214
Willy Tarreau74ca5042013-06-11 23:12:07 +02009215src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +02009216 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +02009217 current incoming connection's source address in the current proxy's
9218 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009219 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009220
Willy Tarreau74ca5042013-06-11 23:12:07 +02009221src_conn_rate([<table>]) : integer
9222 Returns the average connection rate from the incoming connection's source
9223 address in the current proxy's stick-table or in the designated stick-table,
9224 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009225 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009226
Willy Tarreau74ca5042013-06-11 23:12:07 +02009227src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +02009228 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +02009229 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +02009230 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009231 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009232
Willy Tarreau74ca5042013-06-11 23:12:07 +02009233src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009234 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +02009235 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009236 stick-table or in the designated stick-table. It reports the frequency
9237 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009238 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
9239 that the "gpc0_rate" counter must be stored in the stick-table for a value to
9240 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009241
Willy Tarreau74ca5042013-06-11 23:12:07 +02009242src_http_err_cnt([<table>]) : integer
9243 Returns the cumulated number of HTTP errors from the incoming connection's
9244 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +02009245 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009246 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +02009247 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009248
Willy Tarreau74ca5042013-06-11 23:12:07 +02009249src_http_err_rate([<table>]) : integer
9250 Returns the average rate of HTTP errors from the incoming connection's source
9251 address in the current proxy's stick-table or in the designated stick-table,
9252 measured in amount of errors over the period configured in the table. This
9253 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009254 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009255
Willy Tarreau74ca5042013-06-11 23:12:07 +02009256src_http_req_cnt([<table>]) : integer
9257 Returns the cumulated number of HTTP requests from the incoming connection's
9258 source address in the current proxy's stick-table or in the designated stick-
9259 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009260 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009261
Willy Tarreau74ca5042013-06-11 23:12:07 +02009262src_http_req_rate([<table>]) : integer
9263 Returns the average rate of HTTP requests from the incoming connection's
9264 source address in the current proxy's stick-table or in the designated stick-
9265 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +02009266 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009267 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009268
Willy Tarreau74ca5042013-06-11 23:12:07 +02009269src_inc_gpc0([<table>]) : integer
9270 Increments the first General Purpose Counter associated to the incoming
9271 connection's source address in the current proxy's stick-table or in the
9272 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009273 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +02009274 This is typically used as a second ACL in an expression in order to mark a
9275 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +02009276
9277 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +01009278 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +02009279 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +02009280
Willy Tarreau74ca5042013-06-11 23:12:07 +02009281src_kbytes_in([<table>]) : integer
9282 Returns the amount of data received from the incoming connection's source
9283 address in the current proxy's stick-table or in the designated stick-table,
9284 measured in kilobytes over the period configured in the table. If the address
9285 is not found, zero is returned. The test is currently performed on 32-bit
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009286 integers, which limits values to 4 terabytes. See also
9287 sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009288
Willy Tarreau74ca5042013-06-11 23:12:07 +02009289src_kbytes_out([<table>]) : integer
9290 Returns the amount of data sent to the incoming connection's source address
9291 in the current proxy's stick-table or in the designated stick-table, measured
Willy Tarreauc9705a12010-07-27 20:05:50 +02009292 in kilobytes over the period configured in the table. If the address is not
9293 found, zero is returned. The test is currently performed on 32-bit integers,
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009294 which limits values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +02009295
Willy Tarreau74ca5042013-06-11 23:12:07 +02009296src_port : integer
9297 Returns an integer value corresponding to the TCP source port of the
9298 connection on the client side, which is the port the client connected from.
9299 Usage of this function is very limited as modern protocols do not care much
9300 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +01009301
Willy Tarreau74ca5042013-06-11 23:12:07 +02009302src_sess_cnt([<table>]) : integer
9303 Returns the cumulated number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +02009304 connection's source IPv4 address in the current proxy's stick-table or in the
9305 designated stick-table, that were transformed into sessions, which means that
9306 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009307 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009308
Willy Tarreau74ca5042013-06-11 23:12:07 +02009309src_sess_rate([<table>]) : integer
9310 Returns the average session rate from the incoming connection's source
9311 address in the current proxy's stick-table or in the designated stick-table,
9312 measured in amount of sessions over the period configured in the table. A
9313 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009314 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009315
Willy Tarreau74ca5042013-06-11 23:12:07 +02009316src_updt_conn_cnt([<table>]) : integer
9317 Creates or updates the entry associated to the incoming connection's source
9318 address in the current proxy's stick-table or in the designated stick-table.
9319 This table must be configured to store the "conn_cnt" data type, otherwise
9320 the match will be ignored. The current count is incremented by one, and the
9321 expiration timer refreshed. The updated count is returned, so this match
9322 can't return zero. This was used to reject service abusers based on their
9323 source address. Note: it is recommended to use the more complete "track-sc*"
9324 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +02009325
9326 Example :
9327 # This frontend limits incoming SSH connections to 3 per 10 second for
9328 # each source address, and rejects excess connections until a 10 second
9329 # silence is observed. At most 20 addresses are tracked.
9330 listen ssh
9331 bind :22
9332 mode tcp
9333 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +02009334 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +02009335 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +02009336 server local 127.0.0.1:22
9337
Willy Tarreau74ca5042013-06-11 23:12:07 +02009338srv_id : integer
9339 Returns an integer containing the server's id when processing the response.
9340 While it's almost only used with ACLs, it may be used for logging or
9341 debugging.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +02009342
Hervé COMMOWICK35ed8012010-12-15 14:04:51 +01009343
Willy Tarreau74ca5042013-06-11 23:12:07 +020093447.3.3. Fetching samples at Layer 5
9345----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +02009346
Willy Tarreau74ca5042013-06-11 23:12:07 +02009347The layer 5 usually describes just the session layer which in haproxy is
9348closest to the session once all the connection handshakes are finished, but
9349when no content is yet made available. The fetch methods described here are
9350usable as low as the "tcp-request content" rule sets unless they require some
9351future information. Those generally include the results of SSL negociations.
Willy Tarreauc735a072011-03-29 00:57:02 +02009352
Willy Tarreau74ca5042013-06-11 23:12:07 +02009353ssl_c_ca_err : integer
9354 When the incoming connection was made over an SSL/TLS transport layer,
9355 returns the ID of the first error detected during verification of the client
9356 certificate at depth > 0, or 0 if no error was encountered during this
9357 verification process. Please refer to your SSL library's documentation to
9358 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +02009359
Willy Tarreau74ca5042013-06-11 23:12:07 +02009360ssl_c_ca_err_depth : integer
9361 When the incoming connection was made over an SSL/TLS transport layer,
9362 returns the depth in the CA chain of the first error detected during the
9363 verification of the client certificate. If no error is encountered, 0 is
9364 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +01009365
Willy Tarreau74ca5042013-06-11 23:12:07 +02009366ssl_c_err : integer
9367 When the incoming connection was made over an SSL/TLS transport layer,
9368 returns the ID of the first error detected during verification at depth 0, or
9369 0 if no error was encountered during this verification process. Please refer
9370 to your SSL library's documentation to find the exhaustive list of error
9371 codes.
Willy Tarreau62644772008-07-16 18:36:06 +02009372
Willy Tarreau74ca5042013-06-11 23:12:07 +02009373ssl_c_i_dn([<entry>[,<occ>]]) : string
9374 When the incoming connection was made over an SSL/TLS transport layer,
9375 returns the full distinguished name of the issuer of the certificate
9376 presented by the client when no <entry> is specified, or the value of the
9377 first given entry found from the beginning of the DN. If a positive/negative
9378 occurrence number is specified as the optional second argument, it returns
9379 the value of the nth given entry value from the beginning/end of the DN.
9380 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
9381 "ssl_c_i_dn(CN)" retrieves the common name.
Willy Tarreau62644772008-07-16 18:36:06 +02009382
Willy Tarreau74ca5042013-06-11 23:12:07 +02009383 ACL derivatives :
9384 ssl_c_i_dn([<entry>[,<occ>]]) : exact string match
Willy Tarreaub6672b52011-12-12 17:23:41 +01009385
Willy Tarreau74ca5042013-06-11 23:12:07 +02009386ssl_c_key_alg : string
9387 Returns the name of the algorithm used to generate the key of the certificate
9388 presented by the client when the incoming connection was made over an SSL/TLS
9389 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +02009390
Willy Tarreau74ca5042013-06-11 23:12:07 +02009391 ACL derivatives :
9392 ssl_c_key_alg : exact string match
Willy Tarreau2492d5b2009-07-11 00:06:00 +02009393
Willy Tarreau74ca5042013-06-11 23:12:07 +02009394ssl_c_notafter : string
9395 Returns the end date presented by the client as a formatted string
9396 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
9397 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +02009398
Willy Tarreau74ca5042013-06-11 23:12:07 +02009399 ACL derivatives :
9400 ssl_c_notafter : exact string match
Emeric Brunbede3d02009-06-30 17:54:00 +02009401
Willy Tarreau74ca5042013-06-11 23:12:07 +02009402ssl_c_notbefore : string
9403 Returns the start date presented by the client as a formatted string
9404 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
9405 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +01009406
Willy Tarreau74ca5042013-06-11 23:12:07 +02009407 ACL derivatives :
9408 ssl_c_notbefore : exact string match
Willy Tarreaub6672b52011-12-12 17:23:41 +01009409
Willy Tarreau74ca5042013-06-11 23:12:07 +02009410ssl_c_s_dn([<entry>[,<occ>]]) : string
9411 When the incoming connection was made over an SSL/TLS transport layer,
9412 returns the full distinguished name of the subject of the certificate
9413 presented by the client when no <entry> is specified, or the value of the
9414 first given entry found from the beginning of the DN. If a positive/negative
9415 occurrence number is specified as the optional second argument, it returns
9416 the value of the nth given entry value from the beginning/end of the DN.
9417 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
9418 "ssl_c_s_dn(CN)" retrieves the common name.
Willy Tarreaub6672b52011-12-12 17:23:41 +01009419
Willy Tarreau74ca5042013-06-11 23:12:07 +02009420 ACL derivatives :
9421 ssl_c_s_dn([<entry>[,<occ>]]) : exact string match
Willy Tarreau7875d092012-09-10 08:20:03 +02009422
Willy Tarreau74ca5042013-06-11 23:12:07 +02009423ssl_c_serial : binary
9424 Returns the serial of the certificate presented by the client when the
9425 incoming connection was made over an SSL/TLS transport layer. When used for
9426 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +02009427
Willy Tarreau74ca5042013-06-11 23:12:07 +02009428 ACL derivatives :
9429 ssl_c_serial : hex block match
Emeric Brun2525b6b2012-10-18 15:59:43 +02009430
Willy Tarreau74ca5042013-06-11 23:12:07 +02009431ssl_c_sha1 : binary
9432 Returns the SHA-1 fingerprint of the certificate presented by the client when
9433 the incoming connection was made over an SSL/TLS transport layer. This can be
9434 used to stick a client to a server, or to pass this information to a server.
Emeric Brun2525b6b2012-10-18 15:59:43 +02009435
Willy Tarreau74ca5042013-06-11 23:12:07 +02009436ssl_c_sig_alg : string
9437 Returns the name of the algorithm used to sign the certificate presented by
9438 the client when the incoming connection was made over an SSL/TLS transport
9439 layer.
Emeric Brun87855892012-10-17 17:39:35 +02009440
Willy Tarreau74ca5042013-06-11 23:12:07 +02009441 ACL derivatives :
9442 ssl_c_sig_alg : exact string match
9443
9444ssl_c_used : boolean
9445 Returns true if current SSL session uses a client certificate even if current
9446 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +02009447
Willy Tarreau74ca5042013-06-11 23:12:07 +02009448ssl_c_verify : integer
9449 Returns the verify result error ID when the incoming connection was made over
9450 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
9451 refer to your SSL library's documentation for an exhaustive list of error
9452 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +02009453
Willy Tarreau74ca5042013-06-11 23:12:07 +02009454ssl_c_version : integer
9455 Returns the version of the certificate presented by the client when the
9456 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +02009457
Willy Tarreau74ca5042013-06-11 23:12:07 +02009458ssl_f_i_dn([<entry>[,<occ>]]) : string
9459 When the incoming connection was made over an SSL/TLS transport layer,
9460 returns the full distinguished name of the issuer of the certificate
9461 presented by the frontend when no <entry> is specified, or the value of the
9462 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +02009463 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +02009464 the value of the nth given entry value from the beginning/end of the DN.
9465 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
9466 "ssl_f_i_dn(CN)" retrieves the common name.
Emeric Brun87855892012-10-17 17:39:35 +02009467
Willy Tarreau74ca5042013-06-11 23:12:07 +02009468 ACL derivatives :
9469 ssl_f_i_dn([<entry>[,<occ>]]) : exact string match
Willy Tarreau8d598402012-10-22 17:58:39 +02009470
Willy Tarreau74ca5042013-06-11 23:12:07 +02009471ssl_f_key_alg : string
9472 Returns the name of the algorithm used to generate the key of the certificate
9473 presented by the frontend when the incoming connection was made over an
9474 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +02009475
Willy Tarreau74ca5042013-06-11 23:12:07 +02009476 ACL derivatives :
9477 ssl_f_key_alg : exact string match
Emeric Brun9143d372012-12-20 15:44:16 +01009478
Willy Tarreau74ca5042013-06-11 23:12:07 +02009479ssl_f_notafter : string
9480 Returns the end date presented by the frontend as a formatted string
9481 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
9482 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +02009483
Willy Tarreau74ca5042013-06-11 23:12:07 +02009484 ACL derivatives :
9485 ssl_f_notafter : exact string match
Emeric Bruna7359fd2012-10-17 15:03:11 +02009486
Willy Tarreau74ca5042013-06-11 23:12:07 +02009487ssl_f_notbefore : string
9488 Returns the start date presented by the frontend as a formatted string
9489 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
9490 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +02009491
Willy Tarreau74ca5042013-06-11 23:12:07 +02009492 ACL derivatives :
9493 ssl_f_notbefore : exact string match
Emeric Brun7f56e742012-10-19 18:15:40 +02009494
Willy Tarreau74ca5042013-06-11 23:12:07 +02009495ssl_f_s_dn([<entry>[,<occ>]]) : string
9496 When the incoming connection was made over an SSL/TLS transport layer,
9497 returns the full distinguished name of the subject of the certificate
9498 presented by the frontend when no <entry> is specified, or the value of the
9499 first given entry found from the beginning of the DN. If a positive/negative
9500 occurrence number is specified as the optional second argument, it returns
9501 the value of the nth given entry value from the beginning/end of the DN.
9502 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
9503 "ssl_f_s_dn(CN)" retrieves the common name.
Emeric Brunce5ad802012-10-22 14:11:22 +02009504
Willy Tarreau74ca5042013-06-11 23:12:07 +02009505 ACL derivatives :
9506 ssl_f_s_dn([<entry>[,<occ>]]) : exact string match
Emeric Brunce5ad802012-10-22 14:11:22 +02009507
Willy Tarreau74ca5042013-06-11 23:12:07 +02009508ssl_f_serial : binary
9509 Returns the serial of the certificate presented by the frontend when the
9510 incoming connection was made over an SSL/TLS transport layer. When used for
9511 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +02009512
Willy Tarreau74ca5042013-06-11 23:12:07 +02009513 ACL derivatives :
9514 ssl_f_serial : hex block match
Willy Tarreau8d598402012-10-22 17:58:39 +02009515
Willy Tarreau74ca5042013-06-11 23:12:07 +02009516ssl_f_sig_alg : string
9517 Returns the name of the algorithm used to sign the certificate presented by
9518 the frontend when the incoming connection was made over an SSL/TLS transport
9519 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +02009520
Willy Tarreau74ca5042013-06-11 23:12:07 +02009521 ACL derivatives :
9522 ssl_f_sig_alg : exact string match
Emeric Bruna7359fd2012-10-17 15:03:11 +02009523
Willy Tarreau74ca5042013-06-11 23:12:07 +02009524ssl_f_version : integer
9525 Returns the version of the certificate presented by the frontend when the
9526 incoming connection was made over an SSL/TLS transport layer.
9527
9528ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +02009529 Returns true when the front connection was made via an SSL/TLS transport
9530 layer and is locally deciphered. This means it has matched a socket declared
9531 with a "bind" line having the "ssl" option.
9532
Willy Tarreau74ca5042013-06-11 23:12:07 +02009533 Example :
9534 # This passes "X-Proto: https" to servers when client connects over SSL
9535 listen http-https
9536 bind :80
9537 bind :443 ssl crt /etc/haproxy.pem
9538 http-request add-header X-Proto https if { ssl_fc }
9539
9540ssl_fc_alg_keysize : integer
9541 Returns the symmetric cipher key size supported in bits when the incoming
9542 connection was made over an SSL/TLS transport layer.
9543
9544ssl_fc_alpn : string
9545 This extracts the Application Layer Protocol Negociation field from an
9546 incoming connection made via a TLS transport layer and locally deciphered by
9547 haproxy. The result is a string containing the protocol name advertised by
9548 the client. The SSL library must have been built with support for TLS
9549 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
9550 not advertised unless the "alpn" keyword on the "bind" line specifies a
9551 protocol list. Also, nothing forces the client to pick a protocol from this
9552 list, any other one may be requested. The TLS ALPN extension is meant to
9553 replace the TLS NPN extension. See also "ssl_fc_npn".
9554
9555 ACL derivatives :
9556 ssl_fc_alpn : exact string match
Emeric Brun589fcad2012-10-16 14:13:26 +02009557
Willy Tarreau74ca5042013-06-11 23:12:07 +02009558ssl_fc_cipher : string
9559 Returns the name of the used cipher when the incoming connection was made
9560 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +02009561
Willy Tarreau74ca5042013-06-11 23:12:07 +02009562 ACL derivatives :
9563 ssl_fc_cipher : exact string match
Emeric Brun589fcad2012-10-16 14:13:26 +02009564
Willy Tarreau74ca5042013-06-11 23:12:07 +02009565ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +02009566 Returns true if a client certificate is present in an incoming connection over
9567 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +01009568 Note: on SSL session resumption with Session ID or TLS ticket, client
9569 certificate is not present in the current connection but may be retrieved
9570 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
9571 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +02009572
Willy Tarreau74ca5042013-06-11 23:12:07 +02009573ssl_fc_has_sni : boolean
9574 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +02009575 in an incoming connection was made over an SSL/TLS transport layer. Returns
9576 true when the incoming connection presents a TLS SNI field. This requires
9577 that the SSL library is build with support for TLS extensions enabled (check
9578 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +02009579
Willy Tarreau74ca5042013-06-11 23:12:07 +02009580ssl_fc_npn : string
9581 This extracts the Next Protocol Negociation field from an incoming connection
9582 made via a TLS transport layer and locally deciphered by haproxy. The result
9583 is a string containing the protocol name advertised by the client. The SSL
9584 library must have been built with support for TLS extensions enabled (check
9585 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
9586 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
9587 forces the client to pick a protocol from this list, any other one may be
9588 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +02009589
Willy Tarreau74ca5042013-06-11 23:12:07 +02009590 ACL derivatives :
9591 ssl_fc_npn : exact string match
Emeric Brun589fcad2012-10-16 14:13:26 +02009592
Willy Tarreau74ca5042013-06-11 23:12:07 +02009593ssl_fc_protocol : string
9594 Returns the name of the used protocol when the incoming connection was made
9595 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +02009596
Willy Tarreau74ca5042013-06-11 23:12:07 +02009597 ACL derivatives :
9598 ssl_fc_protocol : exact string match
9599
9600ssl_fc_session_id : binary
9601 Returns the SSL ID of the front connection when the incoming connection was
9602 made over an SSL/TLS transport layer. It is useful to stick a given client to
9603 a server. It is important to note that some browsers refresh their session ID
9604 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +02009605
Willy Tarreau74ca5042013-06-11 23:12:07 +02009606ssl_fc_sni : string
9607 This extracts the Server Name Indication TLS extension (SNI) field from an
9608 incoming connection made via an SSL/TLS transport layer and locally
9609 deciphered by haproxy. The result (when present) typically is a string
9610 matching the HTTPS host name (253 chars or less). The SSL library must have
9611 been built with support for TLS extensions enabled (check haproxy -vv).
9612
9613 This fetch is different from "req_ssl_sni" above in that it applies to the
9614 connection being deciphered by haproxy and not to SSL contents being blindly
9615 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +02009616 requires that the SSL library is build with support for TLS extensions
9617 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +02009618
Willy Tarreau74ca5042013-06-11 23:12:07 +02009619 ACL derivatives :
9620 ssl_fc_sni : exact string match
9621 ssl_fc_sni_end : suffix match
9622 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +02009623
Willy Tarreau74ca5042013-06-11 23:12:07 +02009624ssl_fc_use_keysize : integer
9625 Returns the symmetric cipher key size used in bits when the incoming
9626 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +02009627
Willy Tarreaub6fb4202008-07-20 11:18:28 +02009628
Willy Tarreau74ca5042013-06-11 23:12:07 +020096297.3.4. Fetching samples from buffer contents (Layer 6)
9630------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +02009631
Willy Tarreau74ca5042013-06-11 23:12:07 +02009632Fetching samples from buffer contents is a bit different from the previous
9633sample fetches above because the sampled data are ephemeral. These data can
9634only be used when they're available and will be lost when they're forwarded.
9635For this reason, samples fetched from buffer contents during a request cannot
9636be used in a response for example. Even while the data are being fetched, they
9637can change. Sometimes it is necessary to set some delays or combine multiple
9638sample fetch methods to ensure that the expected data are complete and usable,
9639for example through TCP request content inspection. Please see the "tcp-request
9640content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +02009641
Willy Tarreau74ca5042013-06-11 23:12:07 +02009642payload(<offset>,<length>) : binary (deprecated)
9643 This is an alias for "req.payload" when used in the context of a request (eg:
9644 "stick on", "stick match"), and for "res.payload" when used in the context of
9645 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +01009646
Willy Tarreau74ca5042013-06-11 23:12:07 +02009647payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
9648 This is an alias for "req.payload_lv" when used in the context of a request
9649 (eg: "stick on", "stick match"), and for "res.payload_lv" when used in the
9650 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +01009651
Willy Tarreau74ca5042013-06-11 23:12:07 +02009652req.len : integer
9653req_len : integer (deprecated)
9654 Returns an integer value corresponding to the number of bytes present in the
9655 request buffer. This is mostly used in ACL. It is important to understand
9656 that this test does not return false as long as the buffer is changing. This
9657 means that a check with equality to zero will almost always immediately match
9658 at the beginning of the session, while a test for more data will wait for
9659 that data to come in and return false only when haproxy is certain that no
9660 more data will come in. This test was designed to be used with TCP request
9661 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +02009662
Willy Tarreau74ca5042013-06-11 23:12:07 +02009663req.payload(<offset>,<length>) : binary
9664 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +02009665 in the request buffer. As a special case, if the <length> argument is zero,
9666 the the whole buffer from <offset> to the end is extracted. This can be used
9667 with ACLs in order to check for the presence of some content in a buffer at
9668 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +02009669
Willy Tarreau74ca5042013-06-11 23:12:07 +02009670 ACL alternatives :
9671 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +02009672
Willy Tarreau74ca5042013-06-11 23:12:07 +02009673req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
9674 This extracts a binary block whose size is specified at <offset1> for <length>
9675 bytes, and which starts at <offset2> if specified or just after the length in
9676 the request buffer. The <offset2> parameter also supports relative offsets if
9677 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +02009678
Willy Tarreau74ca5042013-06-11 23:12:07 +02009679 ACL alternatives :
9680 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +02009681
Willy Tarreau74ca5042013-06-11 23:12:07 +02009682 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +02009683
Willy Tarreau74ca5042013-06-11 23:12:07 +02009684req.proto_http : boolean
9685req_proto_http : boolean (deprecated)
9686 Returns true when data in the request buffer look like HTTP and correctly
9687 parses as such. It is the same parser as the common HTTP request parser which
9688 is used so there should be no surprises. The test does not match until the
9689 request is complete, failed or timed out. This test may be used to report the
9690 protocol in TCP logs, but the biggest use is to block TCP request analysis
9691 until a complete HTTP request is present in the buffer, for example to track
9692 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +02009693
Willy Tarreau74ca5042013-06-11 23:12:07 +02009694 Example:
9695 # track request counts per "base" (concatenation of Host+URL)
9696 tcp-request inspect-delay 10s
9697 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009698 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +02009699
Willy Tarreau74ca5042013-06-11 23:12:07 +02009700req.rdp_cookie([<name>]) : string
9701rdp_cookie([<name>]) : string (deprecated)
9702 When the request buffer looks like the RDP protocol, extracts the RDP cookie
9703 <name>, or any cookie if unspecified. The parser only checks for the first
9704 cookie, as illustrated in the RDP protocol specification. The cookie name is
9705 case insensitive. Generally the "MSTS" cookie name will be used, as it can
9706 contain the user name of the client connecting to the server if properly
9707 configured on the client. The "MSTSHASH" cookie is often used as well for
9708 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +02009709
Willy Tarreau74ca5042013-06-11 23:12:07 +02009710 This differs from "balance rdp-cookie" in that any balancing algorithm may be
9711 used and thus the distribution of clients to backend servers is not linked to
9712 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
9713 such as "balance roundrobin" or "balance leastconn" will lead to a more even
9714 distribution of clients to backend servers than the hash used by "balance
9715 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +02009716
Willy Tarreau74ca5042013-06-11 23:12:07 +02009717 ACL derivatives :
9718 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +02009719
Willy Tarreau74ca5042013-06-11 23:12:07 +02009720 Example :
9721 listen tse-farm
9722 bind 0.0.0.0:3389
9723 # wait up to 5s for an RDP cookie in the request
9724 tcp-request inspect-delay 5s
9725 tcp-request content accept if RDP_COOKIE
9726 # apply RDP cookie persistence
9727 persist rdp-cookie
9728 # Persist based on the mstshash cookie
9729 # This is only useful makes sense if
9730 # balance rdp-cookie is not used
9731 stick-table type string size 204800
9732 stick on req.rdp_cookie(mstshash)
9733 server srv1 1.1.1.1:3389
9734 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +02009735
Willy Tarreau74ca5042013-06-11 23:12:07 +02009736 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
9737 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +02009738
Willy Tarreau74ca5042013-06-11 23:12:07 +02009739req.rdp_cookie_cnt([name]) : integer
9740rdp_cookie_cnt([name]) : integer (deprecated)
9741 Tries to parse the request buffer as RDP protocol, then returns an integer
9742 corresponding to the number of RDP cookies found. If an optional cookie name
9743 is passed, only cookies matching this name are considered. This is mostly
9744 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +02009745
Willy Tarreau74ca5042013-06-11 23:12:07 +02009746 ACL derivatives :
9747 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +02009748
Willy Tarreau74ca5042013-06-11 23:12:07 +02009749req.ssl_hello_type : integer
9750req_ssl_hello_type : integer (deprecated)
9751 Returns an integer value containing the type of the SSL hello message found
9752 in the request buffer if the buffer contains data that parse as a complete
9753 SSL (v3 or superior) client hello message. Note that this only applies to raw
9754 contents found in the request buffer and not to contents deciphered via an
9755 SSL data layer, so this will not work with "bind" lines having the "ssl"
9756 option. This is mostly used in ACL to detect presence of an SSL hello message
9757 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +02009758
Willy Tarreau74ca5042013-06-11 23:12:07 +02009759req.ssl_sni : string
9760req_ssl_sni : string (deprecated)
9761 Returns a string containing the value of the Server Name TLS extension sent
9762 by a client in a TLS stream passing through the request buffer if the buffer
9763 contains data that parse as a complete SSL (v3 or superior) client hello
9764 message. Note that this only applies to raw contents found in the request
9765 buffer and not to contents deciphered via an SSL data layer, so this will not
9766 work with "bind" lines having the "ssl" option. SNI normally contains the
9767 name of the host the client tries to connect to (for recent browsers). SNI is
9768 useful for allowing or denying access to certain hosts when SSL/TLS is used
9769 by the client. This test was designed to be used with TCP request content
9770 inspection. If content switching is needed, it is recommended to first wait
9771 for a complete client hello (type 1), like in the example below. See also
9772 "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +02009773
Willy Tarreau74ca5042013-06-11 23:12:07 +02009774 ACL derivatives :
9775 req_ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +02009776
Willy Tarreau74ca5042013-06-11 23:12:07 +02009777 Examples :
9778 # Wait for a client hello for at most 5 seconds
9779 tcp-request inspect-delay 5s
9780 tcp-request content accept if { req_ssl_hello_type 1 }
9781 use_backend bk_allow if { req_ssl_sni -f allowed_sites }
9782 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +02009783
Willy Tarreau74ca5042013-06-11 23:12:07 +02009784res.ssl_hello_type : integer
9785rep_ssl_hello_type : integer (deprecated)
9786 Returns an integer value containing the type of the SSL hello message found
9787 in the response buffer if the buffer contains data that parses as a complete
9788 SSL (v3 or superior) hello message. Note that this only applies to raw
9789 contents found in the response buffer and not to contents deciphered via an
9790 SSL data layer, so this will not work with "server" lines having the "ssl"
9791 option. This is mostly used in ACL to detect presence of an SSL hello message
9792 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau51539362012-05-08 12:46:28 +02009793
Willy Tarreau74ca5042013-06-11 23:12:07 +02009794req.ssl_ver : integer
9795req_ssl_ver : integer (deprecated)
9796 Returns an integer value containing the version of the SSL/TLS protocol of a
9797 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
9798 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
9799 composed of the major version multiplied by 65536, added to the minor
9800 version. Note that this only applies to raw contents found in the request
9801 buffer and not to contents deciphered via an SSL data layer, so this will not
9802 work with "bind" lines having the "ssl" option. The ACL version of the test
9803 matches against a decimal notation in the form MAJOR.MINOR (eg: 3.1). This
9804 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009805
Willy Tarreau74ca5042013-06-11 23:12:07 +02009806 ACL derivatives :
9807 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +01009808
Willy Tarreau74ca5042013-06-11 23:12:07 +02009809res.payload(<offset>,<length>) : binary
9810 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +02009811 in the response buffer. As a special case, if the <length> argument is zero,
9812 the the whole buffer from <offset> to the end is extracted. This can be used
9813 with ACLs in order to check for the presence of some content in a buffer at
9814 any location.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009815
Willy Tarreau74ca5042013-06-11 23:12:07 +02009816res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
9817 This extracts a binary block whose size is specified at <offset1> for <length>
9818 bytes, and which starts at <offset2> if specified or just after the length in
9819 the response buffer. The <offset2> parameter also supports relative offsets
9820 if prepended with a '+' or '-' sign.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009821
Willy Tarreau74ca5042013-06-11 23:12:07 +02009822 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009823
Willy Tarreau74ca5042013-06-11 23:12:07 +02009824wait_end : boolean
9825 This fetch either returns true when the inspection period is over, or does
9826 not fetch. It is only used in ACLs, in conjunction with content analysis to
9827 avoid returning a wrong verdict early. It may also be used to delay some
9828 actions, such as a delayed reject for some special addresses. Since it either
9829 stops the rules evaluation or immediately returns true, it is recommended to
9830 use this acl as the last one in a rule. Please note that the default ACL
9831 "WAIT_END" is always usable without prior declaration. This test was designed
9832 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009833
Willy Tarreau74ca5042013-06-11 23:12:07 +02009834 Examples :
9835 # delay every incoming request by 2 seconds
9836 tcp-request inspect-delay 2s
9837 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +01009838
Willy Tarreau74ca5042013-06-11 23:12:07 +02009839 # don't immediately tell bad guys they are rejected
9840 tcp-request inspect-delay 10s
9841 acl goodguys src 10.0.0.0/24
9842 acl badguys src 10.0.1.0/24
9843 tcp-request content accept if goodguys
9844 tcp-request content reject if badguys WAIT_END
9845 tcp-request content reject
9846
9847
98487.3.5. Fetching HTTP samples (Layer 7)
9849--------------------------------------
9850
9851It is possible to fetch samples from HTTP contents, requests and responses.
9852This application layer is also called layer 7. It is only possible to fetch the
9853data in this section when a full HTTP request or response has been parsed from
9854its respective request or response buffer. This is always the case with all
9855HTTP specific rules and for sections running with "mode http". When using TCP
9856content inspection, it may be necessary to support an inspection delay in order
9857to let the request or response come in first. These fetches may require a bit
9858more CPU resources than the layer 4 ones, but not much since the request and
9859response are indexed.
9860
9861base : string
9862 This returns the concatenation of the first Host header and the path part of
9863 the request, which starts at the first slash and ends before the question
9864 mark. It can be useful in virtual hosted environments to detect URL abuses as
9865 well as to improve shared caches efficiency. Using this with a limited size
9866 stick table also allows one to collect statistics about most commonly
9867 requested objects by host/path. With ACLs it can allow simple content
9868 switching rules involving the host and the path at the same time, such as
9869 "www.example.com/favicon.ico". See also "path" and "uri".
9870
9871 ACL derivatives :
9872 base : exact string match
9873 base_beg : prefix match
9874 base_dir : subdir match
9875 base_dom : domain match
9876 base_end : suffix match
9877 base_len : length match
9878 base_reg : regex match
9879 base_sub : substring match
9880
9881base32 : integer
9882 This returns a 32-bit hash of the value returned by the "base" fetch method
9883 above. This is useful to track per-URL activity on high traffic sites without
9884 having to store all URLs. Instead a shorter hash is stored, saving a lot of
9885 memory. The output type is an unsigned integer.
9886
9887base32+src : binary
9888 This returns the concatenation of the base32 fetch above and the src fetch
9889 below. The resulting type is of type binary, with a size of 8 or 20 bytes
9890 depending on the source address family. This can be used to track per-IP,
9891 per-URL counters.
9892
9893req.cook([<name>]) : string
9894cook([<name>]) : string (deprecated)
9895 This extracts the last occurrence of the cookie name <name> on a "Cookie"
9896 header line from the request, and returns its value as string. If no name is
9897 specified, the first cookie value is returned. When used with ACLs, all
9898 matching cookies are evaluated. Spaces around the name and the value are
9899 ignored as requested by the Cookie header specification (RFC6265). The cookie
9900 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
9901 well return an empty value if it is present. Use the "found" match to detect
9902 presence. Use the res.cook() variant for response cookies sent by the server.
9903
9904 ACL derivatives :
9905 cook([<name>]) : exact string match
9906 cook_beg([<name>]) : prefix match
9907 cook_dir([<name>]) : subdir match
9908 cook_dom([<name>]) : domain match
9909 cook_end([<name>]) : suffix match
9910 cook_len([<name>]) : length match
9911 cook_reg([<name>]) : regex match
9912 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +01009913
Willy Tarreau74ca5042013-06-11 23:12:07 +02009914req.cook_cnt([<name>]) : integer
9915cook_cnt([<name>]) : integer (deprecated)
9916 Returns an integer value representing the number of occurrences of the cookie
9917 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009918
Willy Tarreau74ca5042013-06-11 23:12:07 +02009919req.cook_val([<name>]) : integer
9920cook_val([<name>]) : integer (deprecated)
9921 This extracts the last occurrence of the cookie name <name> on a "Cookie"
9922 header line from the request, and converts its value to an integer which is
9923 returned. If no name is specified, the first cookie value is returned. When
9924 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +02009925
Willy Tarreau74ca5042013-06-11 23:12:07 +02009926cookie([<name>]) : string (deprecated)
9927 This extracts the last occurrence of the cookie name <name> on a "Cookie"
9928 header line from the request, or a "Set-Cookie" header from the response, and
9929 returns its value as a string. A typical use is to get multiple clients
9930 sharing a same profile use the same server. This can be similar to what
9931 "appsession" does with the "request-learn" statement, but with support for
9932 multi-peer synchronization and state keeping across restarts. If no name is
9933 specified, the first cookie value is returned. This fetch should not be used
9934 anymore and should be replaced by req.cook() or res.cook() instead as it
9935 ambiguously uses the direction based on the context where it is used.
9936 See also : "appsession".
Willy Tarreaud63335a2010-02-26 12:56:52 +01009937
Willy Tarreau74ca5042013-06-11 23:12:07 +02009938hdr([<name>[,<occ>]]) : string
9939 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
9940 used on responses. Please refer to these respective fetches for more details.
9941 In case of doubt about the fetch direction, please use the explicit ones.
9942 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
9943 unambiguouslly apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009944
Willy Tarreau74ca5042013-06-11 23:12:07 +02009945req.fhdr(<name>[,<occ>]) : string
9946 This extracts the last occurrence of header <name> in an HTTP request. When
9947 used from an ACL, all occurrences are iterated over until a match is found.
9948 Optionally, a specific occurrence might be specified as a position number.
9949 Positive values indicate a position from the first occurrence, with 1 being
9950 the first one. Negative values indicate positions relative to the last one,
9951 with -1 being the last one. It differs from req.hdr() in that any commas
9952 present in the value are returned and are not used as delimiters. This is
9953 sometimes useful with headers such as User-Agent.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009954
Willy Tarreau74ca5042013-06-11 23:12:07 +02009955req.fhdr_cnt([<name>]) : integer
9956 Returns an integer value representing the number of occurrences of request
9957 header field name <name>, or the total number of header fields if <name> is
9958 not specified. Contrary to its req.hdr_cnt() cousin, this function returns
9959 the number of full line headers and does not stop on commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009960
Willy Tarreau74ca5042013-06-11 23:12:07 +02009961req.hdr([<name>[,<occ>]]) : string
9962 This extracts the last occurrence of header <name> in an HTTP request. When
9963 used from an ACL, all occurrences are iterated over until a match is found.
9964 Optionally, a specific occurrence might be specified as a position number.
9965 Positive values indicate a position from the first occurrence, with 1 being
9966 the first one. Negative values indicate positions relative to the last one,
9967 with -1 being the last one. A typical use is with the X-Forwarded-For header
9968 once converted to IP, associated with an IP stick-table. The function
9969 considers any comma as a delimiter for distinct values. If full-line headers
9970 are desired instead, use req.fhdr(). Please carefully check RFC2616 to know
9971 how certain headers are supposed to be parsed. Also, some of them are case
9972 insensitive (eg: Connection).
Willy Tarreaud63335a2010-02-26 12:56:52 +01009973
Willy Tarreau74ca5042013-06-11 23:12:07 +02009974 ACL derivatives :
9975 hdr([<name>[,<occ>]]) : exact string match
9976 hdr_beg([<name>[,<occ>]]) : prefix match
9977 hdr_dir([<name>[,<occ>]]) : subdir match
9978 hdr_dom([<name>[,<occ>]]) : domain match
9979 hdr_end([<name>[,<occ>]]) : suffix match
9980 hdr_len([<name>[,<occ>]]) : length match
9981 hdr_reg([<name>[,<occ>]]) : regex match
9982 hdr_sub([<name>[,<occ>]]) : substring match
9983
9984req.hdr_cnt([<name>]) : integer
9985hdr_cnt([<header>]) : integer (deprecated)
9986 Returns an integer value representing the number of occurrences of request
9987 header field name <name>, or the total number of header field values if
9988 <name> is not specified. It is important to remember that one header line may
9989 count as several headers if it has several values. The function considers any
9990 comma as a delimiter for distinct values. If full-line headers are desired
9991 instead, req.fhdr_cnt() should be used instead. With ACLs, it can be used to
9992 detect presence, absence or abuse of a specific header, as well as to block
9993 request smuggling attacks by rejecting requests which contain more than one
9994 of certain headers. See "req.hdr" for more information on header matching.
9995
9996req.hdr_ip([<name>[,<occ>]]) : ip
9997hdr_ip([<name>[,<occ>]]) : ip (deprecated)
9998 This extracts the last occurrence of header <name> in an HTTP request,
9999 converts it to an IPv4 or IPv6 address and returns this address. When used
10000 with ACLs, all occurrences are checked, and if <name> is omitted, every value
10001 of every header is checked. Optionally, a specific occurrence might be
10002 specified as a position number. Positive values indicate a position from the
10003 first occurrence, with 1 being the first one. Negative values indicate
10004 positions relative to the last one, with -1 being the last one. A typical use
10005 is with the X-Forwarded-For and X-Client-IP headers.
10006
10007req.hdr_val([<name>[,<occ>]]) : integer
10008hdr_val([<name>[,<occ>]]) : integer (deprecated)
10009 This extracts the last occurrence of header <name> in an HTTP request, and
10010 converts it to an integer value. When used with ACLs, all occurrences are
10011 checked, and if <name> is omitted, every value of every header is checked.
10012 Optionally, a specific occurrence might be specified as a position number.
10013 Positive values indicate a position from the first occurrence, with 1 being
10014 the first one. Negative values indicate positions relative to the last one,
10015 with -1 being the last one. A typical use is with the X-Forwarded-For header.
10016
10017http_auth(<userlist>) : boolean
10018 Returns a boolean indicating whether the authentication data received from
10019 the client match a username & password stored in the specified userlist. This
10020 fetch function is not really useful outside of ACLs. Currently only http
10021 basic auth is supported.
10022
10023http_auth_group(<userlist>) : group
10024 Returns a boolean indicating whether the authentication data received from
10025 the client match a username & password stored in the specified userlist, and
10026 whether that username belongs to one of the groups supplied in ACL patterns.
10027 This fetch function is not really useful outside of ACLs. Currently only http
10028 basic auth is supported.
10029
10030 ACL derivatives :
10031 http_auth_group(<userlist>) : user group match
10032
10033http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020010034 Returns true when the request being processed is the first one of the
10035 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020010036 from some requests when a request is not the first one, or to help grouping
10037 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020010038
Willy Tarreau74ca5042013-06-11 23:12:07 +020010039method : integer + string
10040 Returns an integer value corresponding to the method in the HTTP request. For
10041 example, "GET" equals 1 (check sources to establish the matching). Value 9
10042 means "other method" and may be converted to a string extracted from the
10043 stream. This should not be used directly as a sample, this is only meant to
10044 be used from ACLs, which transparently convert methods from patterns to these
10045 integer + string values. Some predefined ACL already check for most common
10046 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010047
Willy Tarreau74ca5042013-06-11 23:12:07 +020010048 ACL derivatives :
10049 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020010050
Willy Tarreau74ca5042013-06-11 23:12:07 +020010051 Example :
10052 # only accept GET and HEAD requests
10053 acl valid_method method GET HEAD
10054 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020010055
Willy Tarreau74ca5042013-06-11 23:12:07 +020010056path : string
10057 This extracts the request's URL path, which starts at the first slash and
10058 ends before the question mark (without the host part). A typical use is with
10059 prefetch-capable caches, and with portals which need to aggregate multiple
10060 information from databases and keep them in caches. Note that with outgoing
10061 caches, it would be wiser to use "url" instead. With ACLs, it's typically
10062 used to match exact file names (eg: "/login.php"), or directory parts using
10063 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010064
Willy Tarreau74ca5042013-06-11 23:12:07 +020010065 ACL derivatives :
10066 path : exact string match
10067 path_beg : prefix match
10068 path_dir : subdir match
10069 path_dom : domain match
10070 path_end : suffix match
10071 path_len : length match
10072 path_reg : regex match
10073 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020010074
Willy Tarreau74ca5042013-06-11 23:12:07 +020010075req.ver : string
10076req_ver : string (deprecated)
10077 Returns the version string from the HTTP request, for example "1.1". This can
10078 be useful for logs, but is mostly there for ACL. Some predefined ACL already
10079 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010080
Willy Tarreau74ca5042013-06-11 23:12:07 +020010081 ACL derivatives :
10082 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020010083
Willy Tarreau74ca5042013-06-11 23:12:07 +020010084res.comp : boolean
10085 Returns the boolean "true" value if the response has been compressed by
10086 HAProxy, otherwise returns boolean "false". This may be used to add
10087 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010088
Willy Tarreau74ca5042013-06-11 23:12:07 +020010089res.comp_algo : string
10090 Returns a string containing the name of the algorithm used if the response
10091 was compressed by HAProxy, for example : "deflate". This may be used to add
10092 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010093
Willy Tarreau74ca5042013-06-11 23:12:07 +020010094res.cook([<name>]) : string
10095scook([<name>]) : string (deprecated)
10096 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
10097 header line from the response, and returns its value as string. If no name is
10098 specified, the first cookie value is returned.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020010099
Willy Tarreau74ca5042013-06-11 23:12:07 +020010100 ACL derivatives :
10101 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020010102
Willy Tarreau74ca5042013-06-11 23:12:07 +020010103res.cook_cnt([<name>]) : integer
10104scook_cnt([<name>]) : integer (deprecated)
10105 Returns an integer value representing the number of occurrences of the cookie
10106 <name> in the response, or all cookies if <name> is not specified. This is
10107 mostly useful when combined with ACLs to detect suspicious responses.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010108
Willy Tarreau74ca5042013-06-11 23:12:07 +020010109res.cook_val([<name>]) : integer
10110scook_val([<name>]) : integer (deprecated)
10111 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
10112 header line from the response, and converts its value to an integer which is
10113 returned. If no name is specified, the first cookie value is returned.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010114
Willy Tarreau74ca5042013-06-11 23:12:07 +020010115res.fhdr([<name>[,<occ>]]) : string
10116 This extracts the last occurrence of header <name> in an HTTP response, or of
10117 the last header if no <name> is specified. Optionally, a specific occurrence
10118 might be specified as a position number. Positive values indicate a position
10119 from the first occurrence, with 1 being the first one. Negative values
10120 indicate positions relative to the last one, with -1 being the last one. It
10121 differs from res.hdr() in that any commas present in the value are returned
10122 and are not used as delimiters. If this is not desired, the res.hdr() fetch
10123 should be used instead. This is sometimes useful with headers such as Date or
10124 Expires.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010125
Willy Tarreau74ca5042013-06-11 23:12:07 +020010126res.fhdr_cnt([<name>]) : integer
10127 Returns an integer value representing the number of occurrences of response
10128 header field name <name>, or the total number of header fields if <name> is
10129 not specified. Contrary to its res.hdr_cnt() cousin, this function returns
10130 the number of full line headers and does not stop on commas. If this is not
10131 desired, the res.hdr_cnt() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010132
Willy Tarreau74ca5042013-06-11 23:12:07 +020010133res.hdr([<name>[,<occ>]]) : string
10134shdr([<name>[,<occ>]]) : string (deprecated)
10135 This extracts the last occurrence of header <name> in an HTTP response, or of
10136 the last header if no <name> is specified. Optionally, a specific occurrence
10137 might be specified as a position number. Positive values indicate a position
10138 from the first occurrence, with 1 being the first one. Negative values
10139 indicate positions relative to the last one, with -1 being the last one. This
10140 can be useful to learn some data into a stick-table. The function considers
10141 any comma as a delimiter for distinct values. If this is not desired, the
10142 res.fhdr() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010143
Willy Tarreau74ca5042013-06-11 23:12:07 +020010144 ACL derivatives :
10145 shdr([<name>[,<occ>]]) : exact string match
10146 shdr_beg([<name>[,<occ>]]) : prefix match
10147 shdr_dir([<name>[,<occ>]]) : subdir match
10148 shdr_dom([<name>[,<occ>]]) : domain match
10149 shdr_end([<name>[,<occ>]]) : suffix match
10150 shdr_len([<name>[,<occ>]]) : length match
10151 shdr_reg([<name>[,<occ>]]) : regex match
10152 shdr_sub([<name>[,<occ>]]) : substring match
10153
10154res.hdr_cnt([<name>]) : integer
10155shdr_cnt([<name>]) : integer (deprecated)
10156 Returns an integer value representing the number of occurrences of response
10157 header field name <name>, or the total number of header fields if <name> is
10158 not specified. The function considers any comma as a delimiter for distinct
10159 values. If this is not desired, the res.fhdr_cnt() fetch should be used
10160 instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010161
Willy Tarreau74ca5042013-06-11 23:12:07 +020010162res.hdr_ip([<name>[,<occ>]]) : ip
10163shdr_ip([<name>[,<occ>]]) : ip (deprecated)
10164 This extracts the last occurrence of header <name> in an HTTP response,
10165 convert it to an IPv4 or IPv6 address and returns this address. Optionally, a
10166 specific occurrence might be specified as a position number. Positive values
10167 indicate a position from the first occurrence, with 1 being the first one.
10168 Negative values indicate positions relative to the last one, with -1 being
10169 the last one. This can be useful to learn some data into a stick table.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010170
Willy Tarreau74ca5042013-06-11 23:12:07 +020010171res.hdr_val([<name>[,<occ>]]) : integer
10172shdr_val([<name>[,<occ>]]) : integer (deprecated)
10173 This extracts the last occurrence of header <name> in an HTTP response, and
10174 converts it to an integer value. Optionally, a specific occurrence might be
10175 specified as a position number. Positive values indicate a position from the
10176 first occurrence, with 1 being the first one. Negative values indicate
10177 positions relative to the last one, with -1 being the last one. This can be
10178 useful to learn some data into a stick table.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010010179
Willy Tarreau74ca5042013-06-11 23:12:07 +020010180res.ver : string
10181resp_ver : string (deprecated)
10182 Returns the version string from the HTTP response, for example "1.1". This
10183 can be useful for logs, but is mostly there for ACL.
Willy Tarreau0e698542011-09-16 08:32:32 +020010184
Willy Tarreau74ca5042013-06-11 23:12:07 +020010185 ACL derivatives :
10186 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010010187
Willy Tarreau74ca5042013-06-11 23:12:07 +020010188set-cookie([<name>]) : string (deprecated)
10189 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
10190 header line from the response and uses the corresponding value to match. This
10191 can be comparable to what "appsession" does with default options, but with
10192 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010010193
Willy Tarreau74ca5042013-06-11 23:12:07 +020010194 This fetch function is deprecated and has been superseded by the "res.cook"
10195 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010010196
Willy Tarreau74ca5042013-06-11 23:12:07 +020010197 See also : "appsession"
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010198
Willy Tarreau74ca5042013-06-11 23:12:07 +020010199status : integer
10200 Returns an integer containing the HTTP status code in the HTTP response, for
10201 example, 302. It is mostly used within ACLs and integer ranges, for example,
10202 to remove any Location header if the response is not a 3xx.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010203
Willy Tarreau74ca5042013-06-11 23:12:07 +020010204url : string
10205 This extracts the request's URL as presented in the request. A typical use is
10206 with prefetch-capable caches, and with portals which need to aggregate
10207 multiple information from databases and keep them in caches. With ACLs, using
10208 "path" is preferred over using "url", because clients may send a full URL as
10209 is normally done with proxies. The only real use is to match "*" which does
10210 not match in "path", and for which there is already a predefined ACL. See
10211 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010212
Willy Tarreau74ca5042013-06-11 23:12:07 +020010213 ACL derivatives :
10214 url : exact string match
10215 url_beg : prefix match
10216 url_dir : subdir match
10217 url_dom : domain match
10218 url_end : suffix match
10219 url_len : length match
10220 url_reg : regex match
10221 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010222
Willy Tarreau74ca5042013-06-11 23:12:07 +020010223url_ip : ip
10224 This extracts the IP address from the request's URL when the host part is
10225 presented as an IP address. Its use is very limited. For instance, a
10226 monitoring system might use this field as an alternative for the source IP in
10227 order to test what path a given source address would follow, or to force an
10228 entry in a table for a given source address. With ACLs it can be used to
10229 restrict access to certain systems through a proxy, for example when combined
10230 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010231
Willy Tarreau74ca5042013-06-11 23:12:07 +020010232url_port : integer
10233 This extracts the port part from the request's URL. Note that if the port is
10234 not specified in the request, port 80 is assumed. With ACLs it can be used to
10235 restrict access to certain systems through a proxy, for example when combined
10236 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010237
Willy Tarreau74ca5042013-06-11 23:12:07 +020010238urlp(<name>[,<delim>]) : string
10239url_param(<name>[,<delim>]) : string
10240 This extracts the first occurrence of the parameter <name> in the query
10241 string, which begins after either '?' or <delim>, and which ends before '&',
10242 ';' or <delim>. The parameter name is case-sensitive. The result is a string
10243 corresponding to the value of the parameter <name> as presented in the
10244 request (no URL decoding is performed). This can be used for session
10245 stickiness based on a client ID, to extract an application cookie passed as a
10246 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
10247 this fetch do not iterate over multiple parameters and stop at the first one
10248 as well.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010249
Willy Tarreau74ca5042013-06-11 23:12:07 +020010250 ACL derivatives :
10251 urlp(<name>[,<delim>]) : exact string match
10252 urlp_beg(<name>[,<delim>]) : prefix match
10253 urlp_dir(<name>[,<delim>]) : subdir match
10254 urlp_dom(<name>[,<delim>]) : domain match
10255 urlp_end(<name>[,<delim>]) : suffix match
10256 urlp_len(<name>[,<delim>]) : length match
10257 urlp_reg(<name>[,<delim>]) : regex match
10258 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010259
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010260
Willy Tarreau74ca5042013-06-11 23:12:07 +020010261 Example :
10262 # match http://example.com/foo?PHPSESSIONID=some_id
10263 stick on urlp(PHPSESSIONID)
10264 # match http://example.com/foo;JSESSIONID=some_id
10265 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010266
Willy Tarreau74ca5042013-06-11 23:12:07 +020010267urlp_val(<name>[,<delim>]) : integer
10268 See "urlp" above. This one extracts the URL parameter <name> in the request
10269 and converts it to an integer value. This can be used for session stickiness
10270 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020010271
Willy Tarreau198a7442008-01-17 12:05:32 +010010272
Willy Tarreau74ca5042013-06-11 23:12:07 +0200102737.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010274---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010010275
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010276Some predefined ACLs are hard-coded so that they do not have to be declared in
10277every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020010278order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010010279
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010280ACL name Equivalent to Usage
10281---------------+-----------------------------+---------------------------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010282FALSE always_false never match
Willy Tarreau2492d5b2009-07-11 00:06:00 +020010283HTTP req_proto_http match if protocol is valid HTTP
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010284HTTP_1.0 req_ver 1.0 match HTTP version 1.0
10285HTTP_1.1 req_ver 1.1 match HTTP version 1.1
Willy Tarreaud63335a2010-02-26 12:56:52 +010010286HTTP_CONTENT hdr_val(content-length) gt 0 match an existing content-length
10287HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
10288HTTP_URL_SLASH url_beg / match URL beginning with "/"
10289HTTP_URL_STAR url * match URL equal to "*"
10290LOCALHOST src 127.0.0.1/8 match connection from local host
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010291METH_CONNECT method CONNECT match HTTP CONNECT method
10292METH_GET method GET HEAD match HTTP GET or HEAD method
10293METH_HEAD method HEAD match HTTP HEAD method
10294METH_OPTIONS method OPTIONS match HTTP OPTIONS method
10295METH_POST method POST match HTTP POST method
10296METH_TRACE method TRACE match HTTP TRACE method
Emeric Brunbede3d02009-06-30 17:54:00 +020010297RDP_COOKIE req_rdp_cookie_cnt gt 0 match presence of an RDP cookie
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010298REQ_CONTENT req_len gt 0 match data in the request buffer
Willy Tarreaud63335a2010-02-26 12:56:52 +010010299TRUE always_true always match
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010300WAIT_END wait_end wait for end of content analysis
10301---------------+-----------------------------+---------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010010302
Willy Tarreaub937b7e2010-01-12 15:27:54 +010010303
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200103048. Logging
10305----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010010306
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010307One of HAProxy's strong points certainly lies is its precise logs. It probably
10308provides the finest level of information available for such a product, which is
10309very important for troubleshooting complex environments. Standard information
10310provided in logs include client ports, TCP/HTTP state timers, precise session
10311state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010312to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010313headers.
10314
10315In order to improve administrators reactivity, it offers a great transparency
10316about encountered problems, both internal and external, and it is possible to
10317send logs to different sources at the same time with different level filters :
10318
10319 - global process-level logs (system errors, start/stop, etc..)
10320 - per-instance system and internal errors (lack of resource, bugs, ...)
10321 - per-instance external troubles (servers up/down, max connections)
10322 - per-instance activity (client connections), either at the establishment or
10323 at the termination.
10324
10325The ability to distribute different levels of logs to different log servers
10326allow several production teams to interact and to fix their problems as soon
10327as possible. For example, the system team might monitor system-wide errors,
10328while the application team might be monitoring the up/down for their servers in
10329real time, and the security team might analyze the activity logs with one hour
10330delay.
10331
10332
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200103338.1. Log levels
10334---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010335
Simon Hormandf791f52011-05-29 15:01:10 +090010336TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010337source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090010338HTTP request, HTTP return code, number of bytes transmitted, conditions
10339in which the session ended, and even exchanged cookies values. For example
10340track a particular user's problems. All messages may be sent to up to two
10341syslog servers. Check the "log" keyword in section 4.2 for more information
10342about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010343
10344
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200103458.2. Log formats
10346----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010347
William Lallemand48940402012-01-30 16:47:22 +010010348HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090010349and will be detailed in the following sections. A few of them may vary
10350slightly with the configuration, due to indicators specific to certain
10351options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010352
10353 - the default format, which is very basic and very rarely used. It only
10354 provides very basic information about the incoming connection at the moment
10355 it is accepted : source IP:port, destination IP:port, and frontend-name.
10356 This mode will eventually disappear so it will not be described to great
10357 extents.
10358
10359 - the TCP format, which is more advanced. This format is enabled when "option
10360 tcplog" is set on the frontend. HAProxy will then usually wait for the
10361 connection to terminate before logging. This format provides much richer
10362 information, such as timers, connection counts, queue size, etc... This
10363 format is recommended for pure TCP proxies.
10364
10365 - the HTTP format, which is the most advanced for HTTP proxying. This format
10366 is enabled when "option httplog" is set on the frontend. It provides the
10367 same information as the TCP format with some HTTP-specific fields such as
10368 the request, the status code, and captures of headers and cookies. This
10369 format is recommended for HTTP proxies.
10370
Emeric Brun3a058f32009-06-30 18:26:00 +020010371 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
10372 fields arranged in the same order as the CLF format. In this mode, all
10373 timers, captures, flags, etc... appear one per field after the end of the
10374 common fields, in the same order they appear in the standard HTTP format.
10375
William Lallemand48940402012-01-30 16:47:22 +010010376 - the custom log format, allows you to make your own log line.
10377
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010378Next sections will go deeper into details for each of these formats. Format
10379specification will be performed on a "field" basis. Unless stated otherwise, a
10380field is a portion of text delimited by any number of spaces. Since syslog
10381servers are susceptible of inserting fields at the beginning of a line, it is
10382always assumed that the first field is the one containing the process name and
10383identifier.
10384
10385Note : Since log lines may be quite long, the log examples in sections below
10386 might be broken into multiple lines. The example log lines will be
10387 prefixed with 3 closing angle brackets ('>>>') and each time a log is
10388 broken into multiple lines, each non-final line will end with a
10389 backslash ('\') and the next line will start indented by two characters.
10390
10391
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200103928.2.1. Default log format
10393-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010394
10395This format is used when no specific option is set. The log is emitted as soon
10396as the connection is accepted. One should note that this currently is the only
10397format which logs the request's destination IP and ports.
10398
10399 Example :
10400 listen www
10401 mode http
10402 log global
10403 server srv1 127.0.0.1:8000
10404
10405 >>> Feb 6 12:12:09 localhost \
10406 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
10407 (www/HTTP)
10408
10409 Field Format Extract from the example above
10410 1 process_name '[' pid ']:' haproxy[14385]:
10411 2 'Connect from' Connect from
10412 3 source_ip ':' source_port 10.0.1.2:33312
10413 4 'to' to
10414 5 destination_ip ':' destination_port 10.0.3.31:8012
10415 6 '(' frontend_name '/' mode ')' (www/HTTP)
10416
10417Detailed fields description :
10418 - "source_ip" is the IP address of the client which initiated the connection.
10419 - "source_port" is the TCP port of the client which initiated the connection.
10420 - "destination_ip" is the IP address the client connected to.
10421 - "destination_port" is the TCP port the client connected to.
10422 - "frontend_name" is the name of the frontend (or listener) which received
10423 and processed the connection.
10424 - "mode is the mode the frontend is operating (TCP or HTTP).
10425
Willy Tarreauceb24bc2010-11-09 12:46:41 +010010426In case of a UNIX socket, the source and destination addresses are marked as
10427"unix:" and the ports reflect the internal ID of the socket which accepted the
10428connection (the same ID as reported in the stats).
10429
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010430It is advised not to use this deprecated format for newer installations as it
10431will eventually disappear.
10432
10433
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200104348.2.2. TCP log format
10435---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010436
10437The TCP format is used when "option tcplog" is specified in the frontend, and
10438is the recommended format for pure TCP proxies. It provides a lot of precious
10439information for troubleshooting. Since this format includes timers and byte
10440counts, the log is normally emitted at the end of the session. It can be
10441emitted earlier if "option logasap" is specified, which makes sense in most
10442environments with long sessions such as remote terminals. Sessions which match
10443the "monitor" rules are never logged. It is also possible not to emit logs for
10444sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020010445specifying "option dontlognull" in the frontend. Successful connections will
10446not be logged if "option dontlog-normal" is specified in the frontend. A few
10447fields may slightly vary depending on some configuration options, those are
10448marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010449
10450 Example :
10451 frontend fnt
10452 mode tcp
10453 option tcplog
10454 log global
10455 default_backend bck
10456
10457 backend bck
10458 server srv1 127.0.0.1:8000
10459
10460 >>> Feb 6 12:12:56 localhost \
10461 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
10462 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
10463
10464 Field Format Extract from the example above
10465 1 process_name '[' pid ']:' haproxy[14387]:
10466 2 client_ip ':' client_port 10.0.1.2:33313
10467 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
10468 4 frontend_name fnt
10469 5 backend_name '/' server_name bck/srv1
10470 6 Tw '/' Tc '/' Tt* 0/0/5007
10471 7 bytes_read* 212
10472 8 termination_state --
10473 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
10474 10 srv_queue '/' backend_queue 0/0
10475
10476Detailed fields description :
10477 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010010478 connection to haproxy. If the connection was accepted on a UNIX socket
10479 instead, the IP address would be replaced with the word "unix". Note that
10480 when the connection is accepted on a socket configured with "accept-proxy"
10481 and the PROXY protocol is correctly used, then the logs will reflect the
10482 forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010483
10484 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010010485 If the connection was accepted on a UNIX socket instead, the port would be
10486 replaced with the ID of the accepting socket, which is also reported in the
10487 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010488
10489 - "accept_date" is the exact date when the connection was received by haproxy
10490 (which might be very slightly different from the date observed on the
10491 network if there was some queuing in the system's backlog). This is usually
10492 the same date which may appear in any upstream firewall's log.
10493
10494 - "frontend_name" is the name of the frontend (or listener) which received
10495 and processed the connection.
10496
10497 - "backend_name" is the name of the backend (or listener) which was selected
10498 to manage the connection to the server. This will be the same as the
10499 frontend if no switching rule has been applied, which is common for TCP
10500 applications.
10501
10502 - "server_name" is the name of the last server to which the connection was
10503 sent, which might differ from the first one if there were connection errors
10504 and a redispatch occurred. Note that this server belongs to the backend
10505 which processed the request. If the connection was aborted before reaching
10506 a server, "<NOSRV>" is indicated instead of a server name.
10507
10508 - "Tw" is the total time in milliseconds spent waiting in the various queues.
10509 It can be "-1" if the connection was aborted before reaching the queue.
10510 See "Timers" below for more details.
10511
10512 - "Tc" is the total time in milliseconds spent waiting for the connection to
10513 establish to the final server, including retries. It can be "-1" if the
10514 connection was aborted before a connection could be established. See
10515 "Timers" below for more details.
10516
10517 - "Tt" is the total time in milliseconds elapsed between the accept and the
10518 last close. It covers all possible processings. There is one exception, if
10519 "option logasap" was specified, then the time counting stops at the moment
10520 the log is emitted. In this case, a '+' sign is prepended before the value,
10521 indicating that the final one will be larger. See "Timers" below for more
10522 details.
10523
10524 - "bytes_read" is the total number of bytes transmitted from the server to
10525 the client when the log is emitted. If "option logasap" is specified, the
10526 this value will be prefixed with a '+' sign indicating that the final one
10527 may be larger. Please note that this value is a 64-bit counter, so log
10528 analysis tools must be able to handle it without overflowing.
10529
10530 - "termination_state" is the condition the session was in when the session
10531 ended. This indicates the session state, which side caused the end of
10532 session to happen, and for what reason (timeout, error, ...). The normal
10533 flags should be "--", indicating the session was closed by either end with
10534 no data remaining in buffers. See below "Session state at disconnection"
10535 for more details.
10536
10537 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010538 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010539 limits have been reached. For instance, if actconn is close to 512 when
10540 multiple connection errors occur, chances are high that the system limits
10541 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010542 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010543
10544 - "feconn" is the total number of concurrent connections on the frontend when
10545 the session was logged. It is useful to estimate the amount of resource
10546 required to sustain high loads, and to detect when the frontend's "maxconn"
10547 has been reached. Most often when this value increases by huge jumps, it is
10548 because there is congestion on the backend servers, but sometimes it can be
10549 caused by a denial of service attack.
10550
10551 - "beconn" is the total number of concurrent connections handled by the
10552 backend when the session was logged. It includes the total number of
10553 concurrent connections active on servers as well as the number of
10554 connections pending in queues. It is useful to estimate the amount of
10555 additional servers needed to support high loads for a given application.
10556 Most often when this value increases by huge jumps, it is because there is
10557 congestion on the backend servers, but sometimes it can be caused by a
10558 denial of service attack.
10559
10560 - "srv_conn" is the total number of concurrent connections still active on
10561 the server when the session was logged. It can never exceed the server's
10562 configured "maxconn" parameter. If this value is very often close or equal
10563 to the server's "maxconn", it means that traffic regulation is involved a
10564 lot, meaning that either the server's maxconn value is too low, or that
10565 there aren't enough servers to process the load with an optimal response
10566 time. When only one of the server's "srv_conn" is high, it usually means
10567 that this server has some trouble causing the connections to take longer to
10568 be processed than on other servers.
10569
10570 - "retries" is the number of connection retries experienced by this session
10571 when trying to connect to the server. It must normally be zero, unless a
10572 server is being stopped at the same moment the connection was attempted.
10573 Frequent retries generally indicate either a network problem between
10574 haproxy and the server, or a misconfigured system backlog on the server
10575 preventing new connections from being queued. This field may optionally be
10576 prefixed with a '+' sign, indicating that the session has experienced a
10577 redispatch after the maximal retry count has been reached on the initial
10578 server. In this case, the server name appearing in the log is the one the
10579 connection was redispatched to, and not the first one, though both may
10580 sometimes be the same in case of hashing for instance. So as a general rule
10581 of thumb, when a '+' is present in front of the retry count, this count
10582 should not be attributed to the logged server.
10583
10584 - "srv_queue" is the total number of requests which were processed before
10585 this one in the server queue. It is zero when the request has not gone
10586 through the server queue. It makes it possible to estimate the approximate
10587 server's response time by dividing the time spent in queue by the number of
10588 requests in the queue. It is worth noting that if a session experiences a
10589 redispatch and passes through two server queues, their positions will be
10590 cumulated. A request should not pass through both the server queue and the
10591 backend queue unless a redispatch occurs.
10592
10593 - "backend_queue" is the total number of requests which were processed before
10594 this one in the backend's global queue. It is zero when the request has not
10595 gone through the global queue. It makes it possible to estimate the average
10596 queue length, which easily translates into a number of missing servers when
10597 divided by a server's "maxconn" parameter. It is worth noting that if a
10598 session experiences a redispatch, it may pass twice in the backend's queue,
10599 and then both positions will be cumulated. A request should not pass
10600 through both the server queue and the backend queue unless a redispatch
10601 occurs.
10602
10603
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200106048.2.3. HTTP log format
10605----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010606
10607The HTTP format is the most complete and the best suited for HTTP proxies. It
10608is enabled by when "option httplog" is specified in the frontend. It provides
10609the same level of information as the TCP format with additional features which
10610are specific to the HTTP protocol. Just like the TCP format, the log is usually
10611emitted at the end of the session, unless "option logasap" is specified, which
10612generally only makes sense for download sites. A session which matches the
10613"monitor" rules will never logged. It is also possible not to log sessions for
10614which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020010615frontend. Successful connections will not be logged if "option dontlog-normal"
10616is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010617
10618Most fields are shared with the TCP log, some being different. A few fields may
10619slightly vary depending on some configuration options. Those ones are marked
10620with a star ('*') after the field name below.
10621
10622 Example :
10623 frontend http-in
10624 mode http
10625 option httplog
10626 log global
10627 default_backend bck
10628
10629 backend static
10630 server srv1 127.0.0.1:8000
10631
10632 >>> Feb 6 12:14:14 localhost \
10633 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
10634 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 +010010635 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010636
10637 Field Format Extract from the example above
10638 1 process_name '[' pid ']:' haproxy[14389]:
10639 2 client_ip ':' client_port 10.0.1.2:33317
10640 3 '[' accept_date ']' [06/Feb/2009:12:14:14.655]
10641 4 frontend_name http-in
10642 5 backend_name '/' server_name static/srv1
10643 6 Tq '/' Tw '/' Tc '/' Tr '/' Tt* 10/0/30/69/109
10644 7 status_code 200
10645 8 bytes_read* 2750
10646 9 captured_request_cookie -
10647 10 captured_response_cookie -
10648 11 termination_state ----
10649 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
10650 13 srv_queue '/' backend_queue 0/0
10651 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
10652 15 '{' captured_response_headers* '}' {}
10653 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010010654
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010655
10656Detailed fields description :
10657 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010010658 connection to haproxy. If the connection was accepted on a UNIX socket
10659 instead, the IP address would be replaced with the word "unix". Note that
10660 when the connection is accepted on a socket configured with "accept-proxy"
10661 and the PROXY protocol is correctly used, then the logs will reflect the
10662 forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010663
10664 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010010665 If the connection was accepted on a UNIX socket instead, the port would be
10666 replaced with the ID of the accepting socket, which is also reported in the
10667 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010668
10669 - "accept_date" is the exact date when the TCP connection was received by
10670 haproxy (which might be very slightly different from the date observed on
10671 the network if there was some queuing in the system's backlog). This is
10672 usually the same date which may appear in any upstream firewall's log. This
10673 does not depend on the fact that the client has sent the request or not.
10674
10675 - "frontend_name" is the name of the frontend (or listener) which received
10676 and processed the connection.
10677
10678 - "backend_name" is the name of the backend (or listener) which was selected
10679 to manage the connection to the server. This will be the same as the
10680 frontend if no switching rule has been applied.
10681
10682 - "server_name" is the name of the last server to which the connection was
10683 sent, which might differ from the first one if there were connection errors
10684 and a redispatch occurred. Note that this server belongs to the backend
10685 which processed the request. If the request was aborted before reaching a
10686 server, "<NOSRV>" is indicated instead of a server name. If the request was
10687 intercepted by the stats subsystem, "<STATS>" is indicated instead.
10688
10689 - "Tq" is the total time in milliseconds spent waiting for the client to send
10690 a full HTTP request, not counting data. It can be "-1" if the connection
10691 was aborted before a complete request could be received. It should always
10692 be very small because a request generally fits in one single packet. Large
10693 times here generally indicate network trouble between the client and
10694 haproxy. See "Timers" below for more details.
10695
10696 - "Tw" is the total time in milliseconds spent waiting in the various queues.
10697 It can be "-1" if the connection was aborted before reaching the queue.
10698 See "Timers" below for more details.
10699
10700 - "Tc" is the total time in milliseconds spent waiting for the connection to
10701 establish to the final server, including retries. It can be "-1" if the
10702 request was aborted before a connection could be established. See "Timers"
10703 below for more details.
10704
10705 - "Tr" is the total time in milliseconds spent waiting for the server to send
10706 a full HTTP response, not counting data. It can be "-1" if the request was
10707 aborted before a complete response could be received. It generally matches
10708 the server's processing time for the request, though it may be altered by
10709 the amount of data sent by the client to the server. Large times here on
10710 "GET" requests generally indicate an overloaded server. See "Timers" below
10711 for more details.
10712
10713 - "Tt" is the total time in milliseconds elapsed between the accept and the
10714 last close. It covers all possible processings. There is one exception, if
10715 "option logasap" was specified, then the time counting stops at the moment
10716 the log is emitted. In this case, a '+' sign is prepended before the value,
10717 indicating that the final one will be larger. See "Timers" below for more
10718 details.
10719
10720 - "status_code" is the HTTP status code returned to the client. This status
10721 is generally set by the server, but it might also be set by haproxy when
10722 the server cannot be reached or when its response is blocked by haproxy.
10723
10724 - "bytes_read" is the total number of bytes transmitted to the client when
10725 the log is emitted. This does include HTTP headers. If "option logasap" is
10726 specified, the this value will be prefixed with a '+' sign indicating that
10727 the final one may be larger. Please note that this value is a 64-bit
10728 counter, so log analysis tools must be able to handle it without
10729 overflowing.
10730
10731 - "captured_request_cookie" is an optional "name=value" entry indicating that
10732 the client had this cookie in the request. The cookie name and its maximum
10733 length are defined by the "capture cookie" statement in the frontend
10734 configuration. The field is a single dash ('-') when the option is not
10735 set. Only one cookie may be captured, it is generally used to track session
10736 ID exchanges between a client and a server to detect session crossing
10737 between clients due to application bugs. For more details, please consult
10738 the section "Capturing HTTP headers and cookies" below.
10739
10740 - "captured_response_cookie" is an optional "name=value" entry indicating
10741 that the server has returned a cookie with its response. The cookie name
10742 and its maximum length are defined by the "capture cookie" statement in the
10743 frontend configuration. The field is a single dash ('-') when the option is
10744 not set. Only one cookie may be captured, it is generally used to track
10745 session ID exchanges between a client and a server to detect session
10746 crossing between clients due to application bugs. For more details, please
10747 consult the section "Capturing HTTP headers and cookies" below.
10748
10749 - "termination_state" is the condition the session was in when the session
10750 ended. This indicates the session state, which side caused the end of
10751 session to happen, for what reason (timeout, error, ...), just like in TCP
10752 logs, and information about persistence operations on cookies in the last
10753 two characters. The normal flags should begin with "--", indicating the
10754 session was closed by either end with no data remaining in buffers. See
10755 below "Session state at disconnection" for more details.
10756
10757 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010758 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010759 limits have been reached. For instance, if actconn is close to 512 or 1024
10760 when multiple connection errors occur, chances are high that the system
10761 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010762 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010763 system.
10764
10765 - "feconn" is the total number of concurrent connections on the frontend when
10766 the session was logged. It is useful to estimate the amount of resource
10767 required to sustain high loads, and to detect when the frontend's "maxconn"
10768 has been reached. Most often when this value increases by huge jumps, it is
10769 because there is congestion on the backend servers, but sometimes it can be
10770 caused by a denial of service attack.
10771
10772 - "beconn" is the total number of concurrent connections handled by the
10773 backend when the session was logged. It includes the total number of
10774 concurrent connections active on servers as well as the number of
10775 connections pending in queues. It is useful to estimate the amount of
10776 additional servers needed to support high loads for a given application.
10777 Most often when this value increases by huge jumps, it is because there is
10778 congestion on the backend servers, but sometimes it can be caused by a
10779 denial of service attack.
10780
10781 - "srv_conn" is the total number of concurrent connections still active on
10782 the server when the session was logged. It can never exceed the server's
10783 configured "maxconn" parameter. If this value is very often close or equal
10784 to the server's "maxconn", it means that traffic regulation is involved a
10785 lot, meaning that either the server's maxconn value is too low, or that
10786 there aren't enough servers to process the load with an optimal response
10787 time. When only one of the server's "srv_conn" is high, it usually means
10788 that this server has some trouble causing the requests to take longer to be
10789 processed than on other servers.
10790
10791 - "retries" is the number of connection retries experienced by this session
10792 when trying to connect to the server. It must normally be zero, unless a
10793 server is being stopped at the same moment the connection was attempted.
10794 Frequent retries generally indicate either a network problem between
10795 haproxy and the server, or a misconfigured system backlog on the server
10796 preventing new connections from being queued. This field may optionally be
10797 prefixed with a '+' sign, indicating that the session has experienced a
10798 redispatch after the maximal retry count has been reached on the initial
10799 server. In this case, the server name appearing in the log is the one the
10800 connection was redispatched to, and not the first one, though both may
10801 sometimes be the same in case of hashing for instance. So as a general rule
10802 of thumb, when a '+' is present in front of the retry count, this count
10803 should not be attributed to the logged server.
10804
10805 - "srv_queue" is the total number of requests which were processed before
10806 this one in the server queue. It is zero when the request has not gone
10807 through the server queue. It makes it possible to estimate the approximate
10808 server's response time by dividing the time spent in queue by the number of
10809 requests in the queue. It is worth noting that if a session experiences a
10810 redispatch and passes through two server queues, their positions will be
10811 cumulated. A request should not pass through both the server queue and the
10812 backend queue unless a redispatch occurs.
10813
10814 - "backend_queue" is the total number of requests which were processed before
10815 this one in the backend's global queue. It is zero when the request has not
10816 gone through the global queue. It makes it possible to estimate the average
10817 queue length, which easily translates into a number of missing servers when
10818 divided by a server's "maxconn" parameter. It is worth noting that if a
10819 session experiences a redispatch, it may pass twice in the backend's queue,
10820 and then both positions will be cumulated. A request should not pass
10821 through both the server queue and the backend queue unless a redispatch
10822 occurs.
10823
10824 - "captured_request_headers" is a list of headers captured in the request due
10825 to the presence of the "capture request header" statement in the frontend.
10826 Multiple headers can be captured, they will be delimited by a vertical bar
10827 ('|'). When no capture is enabled, the braces do not appear, causing a
10828 shift of remaining fields. It is important to note that this field may
10829 contain spaces, and that using it requires a smarter log parser than when
10830 it's not used. Please consult the section "Capturing HTTP headers and
10831 cookies" below for more details.
10832
10833 - "captured_response_headers" is a list of headers captured in the response
10834 due to the presence of the "capture response header" statement in the
10835 frontend. Multiple headers can be captured, they will be delimited by a
10836 vertical bar ('|'). When no capture is enabled, the braces do not appear,
10837 causing a shift of remaining fields. It is important to note that this
10838 field may contain spaces, and that using it requires a smarter log parser
10839 than when it's not used. Please consult the section "Capturing HTTP headers
10840 and cookies" below for more details.
10841
10842 - "http_request" is the complete HTTP request line, including the method,
10843 request and HTTP version string. Non-printable characters are encoded (see
10844 below the section "Non-printable characters"). This is always the last
10845 field, and it is always delimited by quotes and is the only one which can
10846 contain quotes. If new fields are added to the log format, they will be
10847 added before this field. This field might be truncated if the request is
10848 huge and does not fit in the standard syslog buffer (1024 characters). This
10849 is the reason why this field must always remain the last one.
10850
10851
Cyril Bontédc4d9032012-04-08 21:57:39 +0200108528.2.4. Custom log format
10853------------------------
William Lallemand48940402012-01-30 16:47:22 +010010854
Willy Tarreau2beef582012-12-20 17:22:52 +010010855The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010010856mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010010857
10858HAproxy understands some log format variables. % precedes log format variables.
10859Variables can take arguments using braces ('{}'), and multiple arguments are
10860separated by commas within the braces. Flags may be added or removed by
10861prefixing them with a '+' or '-' sign.
10862
10863Special variable "%o" may be used to propagate its flags to all other
10864variables on the same format string. This is particularly handy with quoted
10865string formats ("Q").
10866
Willy Tarreauc8368452012-12-21 00:09:23 +010010867If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreau74ca5042013-06-11 23:12:07 +020010868as a pattern extraction rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010010869less common information such as the client's SSL certificate's DN, or to log
10870the key that would be used to store an entry into a stick table.
10871
William Lallemand48940402012-01-30 16:47:22 +010010872Note: spaces must be escaped. A space character is considered as a separator.
10873HAproxy will automatically merge consecutive separators.
10874
10875Flags are :
10876 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010877 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
William Lallemand48940402012-01-30 16:47:22 +010010878
10879 Example:
10880
10881 log-format %T\ %t\ Some\ Text
10882 log-format %{+Q}o\ %t\ %s\ %{-Q}r
10883
10884At the moment, the default HTTP format is defined this way :
10885
Willy Tarreau2beef582012-12-20 17:22:52 +010010886 log-format %ci:%cp\ [%t]\ %ft\ %b/%s\ %Tq/%Tw/%Tc/%Tr/%Tt\ %ST\ %B\ %CC\ \
10887 %CS\ %tsc\ %ac/%fc/%bc/%sc/%rc\ %sq/%bq\ %hr\ %hs\ %{+Q}r
William Lallemand48940402012-01-30 16:47:22 +010010888
William Lallemandbddd4fd2012-02-27 11:23:10 +010010889the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010010890
Willy Tarreau2beef582012-12-20 17:22:52 +010010891 log-format %{+Q}o\ %{-Q}ci\ -\ -\ [%T]\ %r\ %ST\ %B\ \"\"\ \"\"\ %cp\ \
Willy Tarreau773d65f2012-10-12 14:56:11 +020010892 %ms\ %ft\ %b\ %s\ \%Tq\ %Tw\ %Tc\ %Tr\ %Tt\ %tsc\ %ac\ %fc\ \
Willy Tarreau2beef582012-12-20 17:22:52 +010010893 %bc\ %sc\ %rc\ %sq\ %bq\ %CC\ %CS\ \%hrl\ %hsl
William Lallemand48940402012-01-30 16:47:22 +010010894
William Lallemandbddd4fd2012-02-27 11:23:10 +010010895and the default TCP format is defined this way :
10896
Willy Tarreau2beef582012-12-20 17:22:52 +010010897 log-format %ci:%cp\ [%t]\ %ft\ %b/%s\ %Tw/%Tc/%Tt\ %B\ %ts\ \
William Lallemandbddd4fd2012-02-27 11:23:10 +010010898 %ac/%fc/%bc/%sc/%rc\ %sq/%bq
10899
William Lallemand48940402012-01-30 16:47:22 +010010900Please refer to the table below for currently defined variables :
10901
William Lallemandbddd4fd2012-02-27 11:23:10 +010010902 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020010903 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010010904 +---+------+-----------------------------------------------+-------------+
10905 | | %o | special variable, apply flags on all next var | |
10906 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010010907 | | %B | bytes_read (from server to client) | numeric |
10908 | H | %CC | captured_request_cookie | string |
10909 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020010910 | | %H | hostname | string |
William Lallemanda73203e2012-03-12 12:48:57 +010010911 | | %ID | unique-id | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010010912 | H | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020010913 | | %T | gmt_date_time | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010010914 | | %Tc | Tc | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080010915 | | %Tl | local_date_time | date |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020010916 | H | %Tq | Tq | numeric |
10917 | H | %Tr | Tr | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020010918 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010010919 | | %Tt | Tt | numeric |
10920 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010010921 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010010922 | | %ac | actconn | numeric |
10923 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010010924 | | %bc | beconn (backend concurrent connections) | numeric |
10925 | | %bi | backend_source_ip (connecting address) | IP |
10926 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010010927 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010010928 | | %ci | client_ip (accepted address) | IP |
10929 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010010930 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010010931 | | %fc | feconn (frontend concurrent connections) | numeric |
10932 | | %fi | frontend_ip (accepting address) | IP |
10933 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020010934 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020010935 | H | %hr | captured_request_headers default style | string |
10936 | H | %hrl | captured_request_headers CLF style | string list |
10937 | H | %hs | captured_response_headers default style | string |
10938 | H | %hsl | captured_response_headers CLF style | string list |
William Lallemandbddd4fd2012-02-27 11:23:10 +010010939 | | %ms | accept date milliseconds | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020010940 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020010941 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010010942 | | %rc | retries | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010010943 | H | %rt | http_request_counter | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010010944 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010010945 | | %sc | srv_conn (server concurrent connections) | numeric |
10946 | | %si | server_IP (target address) | IP |
10947 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010010948 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020010949 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
10950 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010010951 | | %t | date_time (with millisecond resolution) | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010010952 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020010953 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010010954 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010010955
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020010956 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010010957
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010010958
109598.2.5. Error log format
10960-----------------------
10961
10962When an incoming connection fails due to an SSL handshake or an invalid PROXY
10963protocol header, haproxy will log the event using a shorter, fixed line format.
10964By default, logs are emitted at the LOG_INFO level, unless the option
10965"log-separate-errors" is set in the backend, in which case the LOG_ERR level
10966will be used. Connections on which no data are exchanged (eg: probes) are not
10967logged if the "dontlognull" option is set.
10968
10969The format looks like this :
10970
10971 >>> Dec 3 18:27:14 localhost \
10972 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
10973 Connection error during SSL handshake
10974
10975 Field Format Extract from the example above
10976 1 process_name '[' pid ']:' haproxy[6103]:
10977 2 client_ip ':' client_port 127.0.0.1:56059
10978 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
10979 4 frontend_name "/" bind_name ":" frt/f1:
10980 5 message Connection error during SSL handshake
10981
10982These fields just provide minimal information to help debugging connection
10983failures.
10984
10985
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200109868.3. Advanced logging options
10987-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010988
10989Some advanced logging options are often looked for but are not easy to find out
10990just by looking at the various options. Here is an entry point for the few
10991options which can enable better logging. Please refer to the keywords reference
10992for more information about their usage.
10993
10994
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200109958.3.1. Disabling logging of external tests
10996------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010997
10998It is quite common to have some monitoring tools perform health checks on
10999haproxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
11000commercial load-balancer, and sometimes it will simply be a more complete
11001monitoring system such as Nagios. When the tests are very frequent, users often
11002ask how to disable logging for those checks. There are three possibilities :
11003
11004 - if connections come from everywhere and are just TCP probes, it is often
11005 desired to simply disable logging of connections without data exchange, by
11006 setting "option dontlognull" in the frontend. It also disables logging of
11007 port scans, which may or may not be desired.
11008
11009 - if the connection come from a known source network, use "monitor-net" to
11010 declare this network as monitoring only. Any host in this network will then
11011 only be able to perform health checks, and their requests will not be
11012 logged. This is generally appropriate to designate a list of equipments
11013 such as other load-balancers.
11014
11015 - if the tests are performed on a known URI, use "monitor-uri" to declare
11016 this URI as dedicated to monitoring. Any host sending this request will
11017 only get the result of a health-check, and the request will not be logged.
11018
11019
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200110208.3.2. Logging before waiting for the session to terminate
11021----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011022
11023The problem with logging at end of connection is that you have no clue about
11024what is happening during very long sessions, such as remote terminal sessions
11025or large file downloads. This problem can be worked around by specifying
11026"option logasap" in the frontend. Haproxy will then log as soon as possible,
11027just before data transfer begins. This means that in case of TCP, it will still
11028log the connection status to the server, and in case of HTTP, it will log just
11029after processing the server headers. In this case, the number of bytes reported
11030is the number of header bytes sent to the client. In order to avoid confusion
11031with normal logs, the total time field and the number of bytes are prefixed
11032with a '+' sign which means that real numbers are certainly larger.
11033
11034
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200110358.3.3. Raising log level upon errors
11036------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020011037
11038Sometimes it is more convenient to separate normal traffic from errors logs,
11039for instance in order to ease error monitoring from log files. When the option
11040"log-separate-errors" is used, connections which experience errors, timeouts,
11041retries, redispatches or HTTP status codes 5xx will see their syslog level
11042raised from "info" to "err". This will help a syslog daemon store the log in
11043a separate file. It is very important to keep the errors in the normal traffic
11044file too, so that log ordering is not altered. You should also be careful if
11045you already have configured your syslog daemon to store all logs higher than
11046"notice" in an "admin" file, because the "err" level is higher than "notice".
11047
11048
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200110498.3.4. Disabling logging of successful connections
11050--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020011051
11052Although this may sound strange at first, some large sites have to deal with
11053multiple thousands of logs per second and are experiencing difficulties keeping
11054them intact for a long time or detecting errors within them. If the option
11055"dontlog-normal" is set on the frontend, all normal connections will not be
11056logged. In this regard, a normal connection is defined as one without any
11057error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
11058and a response with a status 5xx is not considered normal and will be logged
11059too. Of course, doing is is really discouraged as it will remove most of the
11060useful information from the logs. Do this only if you have no other
11061alternative.
11062
11063
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200110648.4. Timing events
11065------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011066
11067Timers provide a great help in troubleshooting network problems. All values are
11068reported in milliseconds (ms). These timers should be used in conjunction with
11069the session termination flags. In TCP mode with "option tcplog" set on the
11070frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
11071mode, 5 control points are reported under the form "Tq/Tw/Tc/Tr/Tt" :
11072
11073 - Tq: total time to get the client request (HTTP mode only). It's the time
11074 elapsed between the moment the client connection was accepted and the
11075 moment the proxy received the last HTTP header. The value "-1" indicates
11076 that the end of headers (empty line) has never been seen. This happens when
11077 the client closes prematurely or times out.
11078
11079 - Tw: total time spent in the queues waiting for a connection slot. It
11080 accounts for backend queue as well as the server queues, and depends on the
11081 queue size, and the time needed for the server to complete previous
11082 requests. The value "-1" means that the request was killed before reaching
11083 the queue, which is generally what happens with invalid or denied requests.
11084
11085 - Tc: total time to establish the TCP connection to the server. It's the time
11086 elapsed between the moment the proxy sent the connection request, and the
11087 moment it was acknowledged by the server, or between the TCP SYN packet and
11088 the matching SYN/ACK packet in return. The value "-1" means that the
11089 connection never established.
11090
11091 - Tr: server response time (HTTP mode only). It's the time elapsed between
11092 the moment the TCP connection was established to the server and the moment
11093 the server sent its complete response headers. It purely shows its request
11094 processing time, without the network overhead due to the data transmission.
11095 It is worth noting that when the client has data to send to the server, for
11096 instance during a POST request, the time already runs, and this can distort
11097 apparent response time. For this reason, it's generally wise not to trust
11098 too much this field for POST requests initiated from clients behind an
11099 untrusted network. A value of "-1" here means that the last the response
11100 header (empty line) was never seen, most likely because the server timeout
11101 stroke before the server managed to process the request.
11102
11103 - Tt: total session duration time, between the moment the proxy accepted it
11104 and the moment both ends were closed. The exception is when the "logasap"
11105 option is specified. In this case, it only equals (Tq+Tw+Tc+Tr), and is
11106 prefixed with a '+' sign. From this field, we can deduce "Td", the data
11107 transmission time, by substracting other timers when valid :
11108
11109 Td = Tt - (Tq + Tw + Tc + Tr)
11110
11111 Timers with "-1" values have to be excluded from this equation. In TCP
11112 mode, "Tq" and "Tr" have to be excluded too. Note that "Tt" can never be
11113 negative.
11114
11115These timers provide precious indications on trouble causes. Since the TCP
11116protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
11117that timers close to multiples of 3s are nearly always related to lost packets
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011118due to network problems (wires, negotiation, congestion). Moreover, if "Tt" is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011119close to a timeout value specified in the configuration, it often means that a
11120session has been aborted on timeout.
11121
11122Most common cases :
11123
11124 - If "Tq" is close to 3000, a packet has probably been lost between the
11125 client and the proxy. This is very rare on local networks but might happen
11126 when clients are on far remote networks and send large requests. It may
11127 happen that values larger than usual appear here without any network cause.
11128 Sometimes, during an attack or just after a resource starvation has ended,
11129 haproxy may accept thousands of connections in a few milliseconds. The time
11130 spent accepting these connections will inevitably slightly delay processing
11131 of other connections, and it can happen that request times in the order of
11132 a few tens of milliseconds are measured after a few thousands of new
Patrick Mezard105faca2010-06-12 17:02:46 +020011133 connections have been accepted at once. Setting "option http-server-close"
11134 may display larger request times since "Tq" also measures the time spent
11135 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011136
11137 - If "Tc" is close to 3000, a packet has probably been lost between the
11138 server and the proxy during the server connection phase. This value should
11139 always be very low, such as 1 ms on local networks and less than a few tens
11140 of ms on remote networks.
11141
Willy Tarreau55165fe2009-05-10 12:02:55 +020011142 - If "Tr" is nearly always lower than 3000 except some rare values which seem
11143 to be the average majored by 3000, there are probably some packets lost
11144 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011145
11146 - If "Tt" is large even for small byte counts, it generally is because
11147 neither the client nor the server decides to close the connection, for
11148 instance because both have agreed on a keep-alive connection mode. In order
11149 to solve this issue, it will be needed to specify "option httpclose" on
11150 either the frontend or the backend. If the problem persists, it means that
11151 the server ignores the "close" connection mode and expects the client to
11152 close. Then it will be required to use "option forceclose". Having the
11153 smallest possible 'Tt' is important when connection regulation is used with
11154 the "maxconn" option on the servers, since no new connection will be sent
11155 to the server until another one is released.
11156
11157Other noticeable HTTP log cases ('xx' means any value to be ignored) :
11158
11159 Tq/Tw/Tc/Tr/+Tt The "option logasap" is present on the frontend and the log
11160 was emitted before the data phase. All the timers are valid
11161 except "Tt" which is shorter than reality.
11162
11163 -1/xx/xx/xx/Tt The client was not able to send a complete request in time
11164 or it aborted too early. Check the session termination flags
11165 then "timeout http-request" and "timeout client" settings.
11166
11167 Tq/-1/xx/xx/Tt It was not possible to process the request, maybe because
11168 servers were out of order, because the request was invalid
11169 or forbidden by ACL rules. Check the session termination
11170 flags.
11171
11172 Tq/Tw/-1/xx/Tt The connection could not establish on the server. Either it
11173 actively refused it or it timed out after Tt-(Tq+Tw) ms.
11174 Check the session termination flags, then check the
11175 "timeout connect" setting. Note that the tarpit action might
11176 return similar-looking patterns, with "Tw" equal to the time
11177 the client connection was maintained open.
11178
11179 Tq/Tw/Tc/-1/Tt The server has accepted the connection but did not return
11180 a complete response in time, or it closed its connexion
11181 unexpectedly after Tt-(Tq+Tw+Tc) ms. Check the session
11182 termination flags, then check the "timeout server" setting.
11183
11184
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200111858.5. Session state at disconnection
11186-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011187
11188TCP and HTTP logs provide a session termination indicator in the
11189"termination_state" field, just before the number of active connections. It is
111902-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
11191each of which has a special meaning :
11192
11193 - On the first character, a code reporting the first event which caused the
11194 session to terminate :
11195
11196 C : the TCP session was unexpectedly aborted by the client.
11197
11198 S : the TCP session was unexpectedly aborted by the server, or the
11199 server explicitly refused it.
11200
11201 P : the session was prematurely aborted by the proxy, because of a
11202 connection limit enforcement, because a DENY filter was matched,
11203 because of a security check which detected and blocked a dangerous
11204 error in server response which might have caused information leak
Willy Tarreau570f2212013-06-10 16:42:09 +020011205 (eg: cacheable cookie).
11206
11207 L : the session was locally processed by haproxy and was not passed to
11208 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011209
11210 R : a resource on the proxy has been exhausted (memory, sockets, source
11211 ports, ...). Usually, this appears during the connection phase, and
11212 system logs should contain a copy of the precise error. If this
11213 happens, it must be considered as a very serious anomaly which
11214 should be fixed as soon as possible by any means.
11215
11216 I : an internal error was identified by the proxy during a self-check.
11217 This should NEVER happen, and you are encouraged to report any log
11218 containing this, because this would almost certainly be a bug. It
11219 would be wise to preventively restart the process after such an
11220 event too, in case it would be caused by memory corruption.
11221
Simon Horman752dc4a2011-06-21 14:34:59 +090011222 D : the session was killed by haproxy because the server was detected
11223 as down and was configured to kill all connections when going down.
11224
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070011225 U : the session was killed by haproxy on this backup server because an
11226 active server was detected as up and was configured to kill all
11227 backup connections when going up.
11228
Willy Tarreaua2a64e92011-09-07 23:01:56 +020011229 K : the session was actively killed by an admin operating on haproxy.
11230
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011231 c : the client-side timeout expired while waiting for the client to
11232 send or receive data.
11233
11234 s : the server-side timeout expired while waiting for the server to
11235 send or receive data.
11236
11237 - : normal session completion, both the client and the server closed
11238 with nothing left in the buffers.
11239
11240 - on the second character, the TCP or HTTP session state when it was closed :
11241
Willy Tarreauf7b30a92010-12-06 22:59:17 +010011242 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011243 (HTTP mode only). Nothing was sent to any server.
11244
11245 Q : the proxy was waiting in the QUEUE for a connection slot. This can
11246 only happen when servers have a 'maxconn' parameter set. It can
11247 also happen in the global queue after a redispatch consecutive to
11248 a failed attempt to connect to a dying server. If no redispatch is
11249 reported, then no connection attempt was made to any server.
11250
11251 C : the proxy was waiting for the CONNECTION to establish on the
11252 server. The server might at most have noticed a connection attempt.
11253
11254 H : the proxy was waiting for complete, valid response HEADERS from the
11255 server (HTTP only).
11256
11257 D : the session was in the DATA phase.
11258
11259 L : the proxy was still transmitting LAST data to the client while the
11260 server had already finished. This one is very rare as it can only
11261 happen when the client dies while receiving the last packets.
11262
11263 T : the request was tarpitted. It has been held open with the client
11264 during the whole "timeout tarpit" duration or until the client
11265 closed, both of which will be reported in the "Tw" timer.
11266
11267 - : normal session completion after end of data transfer.
11268
11269 - the third character tells whether the persistence cookie was provided by
11270 the client (only in HTTP mode) :
11271
11272 N : the client provided NO cookie. This is usually the case for new
11273 visitors, so counting the number of occurrences of this flag in the
11274 logs generally indicate a valid trend for the site frequentation.
11275
11276 I : the client provided an INVALID cookie matching no known server.
11277 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020011278 cookies between HTTP/HTTPS sites, persistence conditionally
11279 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011280
11281 D : the client provided a cookie designating a server which was DOWN,
11282 so either "option persist" was used and the client was sent to
11283 this server, or it was not set and the client was redispatched to
11284 another server.
11285
Willy Tarreau996a92c2010-10-13 19:30:47 +020011286 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011287 server.
11288
Willy Tarreau996a92c2010-10-13 19:30:47 +020011289 E : the client provided a valid cookie, but with a last date which was
11290 older than what is allowed by the "maxidle" cookie parameter, so
11291 the cookie is consider EXPIRED and is ignored. The request will be
11292 redispatched just as if there was no cookie.
11293
11294 O : the client provided a valid cookie, but with a first date which was
11295 older than what is allowed by the "maxlife" cookie parameter, so
11296 the cookie is consider too OLD and is ignored. The request will be
11297 redispatched just as if there was no cookie.
11298
Willy Tarreauc89ccb62012-04-05 21:18:22 +020011299 U : a cookie was present but was not used to select the server because
11300 some other server selection mechanism was used instead (typically a
11301 "use-server" rule).
11302
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011303 - : does not apply (no cookie set in configuration).
11304
11305 - the last character reports what operations were performed on the persistence
11306 cookie returned by the server (only in HTTP mode) :
11307
11308 N : NO cookie was provided by the server, and none was inserted either.
11309
11310 I : no cookie was provided by the server, and the proxy INSERTED one.
11311 Note that in "cookie insert" mode, if the server provides a cookie,
11312 it will still be overwritten and reported as "I" here.
11313
Willy Tarreau996a92c2010-10-13 19:30:47 +020011314 U : the proxy UPDATED the last date in the cookie that was presented by
11315 the client. This can only happen in insert mode with "maxidle". It
11316 happens everytime there is activity at a different date than the
11317 date indicated in the cookie. If any other change happens, such as
11318 a redispatch, then the cookie will be marked as inserted instead.
11319
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011320 P : a cookie was PROVIDED by the server and transmitted as-is.
11321
11322 R : the cookie provided by the server was REWRITTEN by the proxy, which
11323 happens in "cookie rewrite" or "cookie prefix" modes.
11324
11325 D : the cookie provided by the server was DELETED by the proxy.
11326
11327 - : does not apply (no cookie set in configuration).
11328
Willy Tarreau996a92c2010-10-13 19:30:47 +020011329The combination of the two first flags gives a lot of information about what
11330was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011331helpful to detect server saturation, network troubles, local system resource
11332starvation, attacks, etc...
11333
11334The most common termination flags combinations are indicated below. They are
11335alphabetically sorted, with the lowercase set just after the upper case for
11336easier finding and understanding.
11337
11338 Flags Reason
11339
11340 -- Normal termination.
11341
11342 CC The client aborted before the connection could be established to the
11343 server. This can happen when haproxy tries to connect to a recently
11344 dead (or unchecked) server, and the client aborts while haproxy is
11345 waiting for the server to respond or for "timeout connect" to expire.
11346
11347 CD The client unexpectedly aborted during data transfer. This can be
11348 caused by a browser crash, by an intermediate equipment between the
11349 client and haproxy which decided to actively break the connection,
11350 by network routing issues between the client and haproxy, or by a
11351 keep-alive session between the server and the client terminated first
11352 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010011353
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011354 cD The client did not send nor acknowledge any data for as long as the
11355 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020011356 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011357
11358 CH The client aborted while waiting for the server to start responding.
11359 It might be the server taking too long to respond or the client
11360 clicking the 'Stop' button too fast.
11361
11362 cH The "timeout client" stroke while waiting for client data during a
11363 POST request. This is sometimes caused by too large TCP MSS values
11364 for PPPoE networks which cannot transport full-sized packets. It can
11365 also happen when client timeout is smaller than server timeout and
11366 the server takes too long to respond.
11367
11368 CQ The client aborted while its session was queued, waiting for a server
11369 with enough empty slots to accept it. It might be that either all the
11370 servers were saturated or that the assigned server was taking too
11371 long a time to respond.
11372
11373 CR The client aborted before sending a full HTTP request. Most likely
11374 the request was typed by hand using a telnet client, and aborted
11375 too early. The HTTP status code is likely a 400 here. Sometimes this
11376 might also be caused by an IDS killing the connection between haproxy
11377 and the client.
11378
11379 cR The "timeout http-request" stroke before the client sent a full HTTP
11380 request. This is sometimes caused by too large TCP MSS values on the
11381 client side for PPPoE networks which cannot transport full-sized
11382 packets, or by clients sending requests by hand and not typing fast
11383 enough, or forgetting to enter the empty line at the end of the
11384 request. The HTTP status code is likely a 408 here.
11385
11386 CT The client aborted while its session was tarpitted. It is important to
11387 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020011388 wrong tarpit rules have been written. If a lot of them happen, it
11389 might make sense to lower the "timeout tarpit" value to something
11390 closer to the average reported "Tw" timer, in order not to consume
11391 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011392
Willy Tarreau570f2212013-06-10 16:42:09 +020011393 LR The request was intercepted and locally handled by haproxy. Generally
11394 it means that this was a redirect or a stats request.
11395
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011396 SC The server or an equipment between it and haproxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011397 the TCP connection (the proxy received a TCP RST or an ICMP message
11398 in return). Under some circumstances, it can also be the network
11399 stack telling the proxy that the server is unreachable (eg: no route,
11400 or no ARP response on local network). When this happens in HTTP mode,
11401 the status code is likely a 502 or 503 here.
11402
11403 sC The "timeout connect" stroke before a connection to the server could
11404 complete. When this happens in HTTP mode, the status code is likely a
11405 503 or 504 here.
11406
11407 SD The connection to the server died with an error during the data
11408 transfer. This usually means that haproxy has received an RST from
11409 the server or an ICMP message from an intermediate equipment while
11410 exchanging data with the server. This can be caused by a server crash
11411 or by a network issue on an intermediate equipment.
11412
11413 sD The server did not send nor acknowledge any data for as long as the
11414 "timeout server" setting during the data phase. This is often caused
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011415 by too short timeouts on L4 equipments before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011416 load-balancers, ...), as well as keep-alive sessions maintained
11417 between the client and the server expiring first on haproxy.
11418
11419 SH The server aborted before sending its full HTTP response headers, or
11420 it crashed while processing the request. Since a server aborting at
11421 this moment is very rare, it would be wise to inspect its logs to
11422 control whether it crashed and why. The logged request may indicate a
11423 small set of faulty requests, demonstrating bugs in the application.
11424 Sometimes this might also be caused by an IDS killing the connection
11425 between haproxy and the server.
11426
11427 sH The "timeout server" stroke before the server could return its
11428 response headers. This is the most common anomaly, indicating too
11429 long transactions, probably caused by server or database saturation.
11430 The immediate workaround consists in increasing the "timeout server"
11431 setting, but it is important to keep in mind that the user experience
11432 will suffer from these long response times. The only long term
11433 solution is to fix the application.
11434
11435 sQ The session spent too much time in queue and has been expired. See
11436 the "timeout queue" and "timeout connect" settings to find out how to
11437 fix this if it happens too often. If it often happens massively in
11438 short periods, it may indicate general problems on the affected
11439 servers due to I/O or database congestion, or saturation caused by
11440 external attacks.
11441
11442 PC The proxy refused to establish a connection to the server because the
11443 process' socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020011444 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011445 so that it does not happen anymore. This status is very rare and
11446 might happen when the global "ulimit-n" parameter is forced by hand.
11447
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010011448 PD The proxy blocked an incorrectly formatted chunked encoded message in
11449 a request or a response, after the server has emitted its headers. In
11450 most cases, this will indicate an invalid message from the server to
11451 the client.
11452
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011453 PH The proxy blocked the server's response, because it was invalid,
11454 incomplete, dangerous (cache control), or matched a security filter.
11455 In any case, an HTTP 502 error is sent to the client. One possible
11456 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010011457 containing unauthorized characters. It is also possible but quite
11458 rare, that the proxy blocked a chunked-encoding request from the
11459 client due to an invalid syntax, before the server responded. In this
11460 case, an HTTP 400 error is sent to the client and reported in the
11461 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011462
11463 PR The proxy blocked the client's HTTP request, either because of an
11464 invalid HTTP syntax, in which case it returned an HTTP 400 error to
11465 the client, or because a deny filter matched, in which case it
11466 returned an HTTP 403 error.
11467
11468 PT The proxy blocked the client's request and has tarpitted its
11469 connection before returning it a 500 server error. Nothing was sent
11470 to the server. The connection was maintained open for as long as
11471 reported by the "Tw" timer field.
11472
11473 RC A local resource has been exhausted (memory, sockets, source ports)
11474 preventing the connection to the server from establishing. The error
11475 logs will tell precisely what was missing. This is very rare and can
11476 only be solved by proper system tuning.
11477
Willy Tarreau996a92c2010-10-13 19:30:47 +020011478The combination of the two last flags gives a lot of information about how
11479persistence was handled by the client, the server and by haproxy. This is very
11480important to troubleshoot disconnections, when users complain they have to
11481re-authenticate. The commonly encountered flags are :
11482
11483 -- Persistence cookie is not enabled.
11484
11485 NN No cookie was provided by the client, none was inserted in the
11486 response. For instance, this can be in insert mode with "postonly"
11487 set on a GET request.
11488
11489 II A cookie designating an invalid server was provided by the client,
11490 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040011491 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020011492 value can be presented by a client when no other server knows it.
11493
11494 NI No cookie was provided by the client, one was inserted in the
11495 response. This typically happens for first requests from every user
11496 in "insert" mode, which makes it an easy way to count real users.
11497
11498 VN A cookie was provided by the client, none was inserted in the
11499 response. This happens for most responses for which the client has
11500 already got a cookie.
11501
11502 VU A cookie was provided by the client, with a last visit date which is
11503 not completely up-to-date, so an updated cookie was provided in
11504 response. This can also happen if there was no date at all, or if
11505 there was a date but the "maxidle" parameter was not set, so that the
11506 cookie can be switched to unlimited time.
11507
11508 EI A cookie was provided by the client, with a last visit date which is
11509 too old for the "maxidle" parameter, so the cookie was ignored and a
11510 new cookie was inserted in the response.
11511
11512 OI A cookie was provided by the client, with a first visit date which is
11513 too old for the "maxlife" parameter, so the cookie was ignored and a
11514 new cookie was inserted in the response.
11515
11516 DI The server designated by the cookie was down, a new server was
11517 selected and a new cookie was emitted in the response.
11518
11519 VI The server designated by the cookie was not marked dead but could not
11520 be reached. A redispatch happened and selected another one, which was
11521 then advertised in the response.
11522
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011523
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200115248.6. Non-printable characters
11525-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011526
11527In order not to cause trouble to log analysis tools or terminals during log
11528consulting, non-printable characters are not sent as-is into log files, but are
11529converted to the two-digits hexadecimal representation of their ASCII code,
11530prefixed by the character '#'. The only characters that can be logged without
11531being escaped are comprised between 32 and 126 (inclusive). Obviously, the
11532escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
11533is the same for the character '"' which becomes "#22", as well as '{', '|' and
11534'}' when logging headers.
11535
11536Note that the space character (' ') is not encoded in headers, which can cause
11537issues for tools relying on space count to locate fields. A typical header
11538containing spaces is "User-Agent".
11539
11540Last, it has been observed that some syslog daemons such as syslog-ng escape
11541the quote ('"') with a backslash ('\'). The reverse operation can safely be
11542performed since no quote may appear anywhere else in the logs.
11543
11544
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200115458.7. Capturing HTTP cookies
11546---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011547
11548Cookie capture simplifies the tracking a complete user session. This can be
11549achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011550section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011551cookie will simultaneously be checked in the request ("Cookie:" header) and in
11552the response ("Set-Cookie:" header). The respective values will be reported in
11553the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011554locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011555not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
11556user switches to a new session for example, because the server will reassign it
11557a new cookie. It is also possible to detect if a server unexpectedly sets a
11558wrong cookie to a client, leading to session crossing.
11559
11560 Examples :
11561 # capture the first cookie whose name starts with "ASPSESSION"
11562 capture cookie ASPSESSION len 32
11563
11564 # capture the first cookie whose name is exactly "vgnvisitor"
11565 capture cookie vgnvisitor= len 32
11566
11567
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200115688.8. Capturing HTTP headers
11569---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011570
11571Header captures are useful to track unique request identifiers set by an upper
11572proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
11573the response, one can search for information about the response length, how the
11574server asked the cache to behave, or an object location during a redirection.
11575
11576Header captures are performed using the "capture request header" and "capture
11577response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011578section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011579
11580It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011581time. Non-existent headers are logged as empty strings, and if one header
11582appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011583are grouped within braces '{' and '}' in the same order as they were declared,
11584and delimited with a vertical bar '|' without any space. Response headers
11585follow the same representation, but are displayed after a space following the
11586request headers block. These blocks are displayed just before the HTTP request
11587in the logs.
11588
11589 Example :
11590 # This instance chains to the outgoing proxy
11591 listen proxy-out
11592 mode http
11593 option httplog
11594 option logasap
11595 log global
11596 server cache1 192.168.1.1:3128
11597
11598 # log the name of the virtual server
11599 capture request header Host len 20
11600
11601 # log the amount of data uploaded during a POST
11602 capture request header Content-Length len 10
11603
11604 # log the beginning of the referrer
11605 capture request header Referer len 20
11606
11607 # server name (useful for outgoing proxies only)
11608 capture response header Server len 20
11609
11610 # logging the content-length is useful with "option logasap"
11611 capture response header Content-Length len 10
11612
11613 # log the expected cache behaviour on the response
11614 capture response header Cache-Control len 8
11615
11616 # the Via header will report the next proxy's name
11617 capture response header Via len 20
11618
11619 # log the URL location during a redirection
11620 capture response header Location len 20
11621
11622 >>> Aug 9 20:26:09 localhost \
11623 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
11624 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
11625 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
11626 "GET http://fr.adserver.yahoo.com/"
11627
11628 >>> Aug 9 20:30:46 localhost \
11629 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
11630 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
11631 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010011632 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011633
11634 >>> Aug 9 20:30:46 localhost \
11635 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
11636 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
11637 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
11638 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010011639 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011640
11641
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200116428.9. Examples of logs
11643---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011644
11645These are real-world examples of logs accompanied with an explanation. Some of
11646them have been made up by hand. The syslog part has been removed for better
11647reading. Their sole purpose is to explain how to decipher them.
11648
11649 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
11650 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
11651 "HEAD / HTTP/1.0"
11652
11653 => long request (6.5s) entered by hand through 'telnet'. The server replied
11654 in 147 ms, and the session ended normally ('----')
11655
11656 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
11657 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
11658 0/9 "HEAD / HTTP/1.0"
11659
11660 => Idem, but the request was queued in the global queue behind 9 other
11661 requests, and waited there for 1230 ms.
11662
11663 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
11664 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
11665 "GET /image.iso HTTP/1.0"
11666
11667 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011668 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011669 14 ms, 243 bytes of headers were sent to the client, and total time from
11670 accept to first data byte is 30 ms.
11671
11672 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
11673 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
11674 "GET /cgi-bin/bug.cgi? HTTP/1.0"
11675
11676 => the proxy blocked a server response either because of an "rspdeny" or
11677 "rspideny" filter, or because the response was improperly formatted and
Willy Tarreau3c92c5f2011-08-28 09:45:47 +020011678 not HTTP-compliant, or because it blocked sensitive information which
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011679 risked being cached. In this case, the response is replaced with a "502
11680 bad gateway". The flags ("PH--") tell us that it was haproxy who decided
11681 to return the 502 and not the server.
11682
11683 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010011684 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 +010011685
11686 => the client never completed its request and aborted itself ("C---") after
11687 8.5s, while the proxy was waiting for the request headers ("-R--").
11688 Nothing was sent to any server.
11689
11690 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
11691 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
11692
11693 => The client never completed its request, which was aborted by the
11694 time-out ("c---") after 50s, while the proxy was waiting for the request
11695 headers ("-R--"). Nothing was sent to any server, but the proxy could
11696 send a 408 return code to the client.
11697
11698 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
11699 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
11700
11701 => This log was produced with "option tcplog". The client timed out after
11702 5 seconds ("c----").
11703
11704 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
11705 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010011706 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011707
11708 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011709 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011710 (config says 'retries 3'), and no redispatch (otherwise we would have
11711 seen "/+3"). Status code 503 was returned to the client. There were 115
11712 connections on this server, 202 connections on this proxy, and 205 on
11713 the global process. It is possible that the server refused the
11714 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011715
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010011716
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200117179. Statistics and monitoring
11718----------------------------
11719
11720It is possible to query HAProxy about its status. The most commonly used
11721mechanism is the HTTP statistics page. This page also exposes an alternative
11722CSV output format for monitoring tools. The same format is provided on the
11723Unix socket.
11724
11725
117269.1. CSV format
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010011727---------------
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +010011728
Willy Tarreau7f062c42009-03-05 18:43:00 +010011729The statistics may be consulted either from the unix socket or from the HTTP
11730page. Both means provide a CSV format whose fields follow.
11731
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +010011732 0. pxname: proxy name
11733 1. svname: service name (FRONTEND for frontend, BACKEND for backend, any name
11734 for server)
11735 2. qcur: current queued requests
11736 3. qmax: max queued requests
11737 4. scur: current sessions
11738 5. smax: max sessions
11739 6. slim: sessions limit
11740 7. stot: total sessions
11741 8. bin: bytes in
11742 9. bout: bytes out
11743 10. dreq: denied requests
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010011744 11. dresp: denied responses
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +010011745 12. ereq: request errors
11746 13. econ: connection errors
Willy Tarreauae526782010-03-04 20:34:23 +010011747 14. eresp: response errors (among which srv_abrt)
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +010011748 15. wretr: retries (warning)
11749 16. wredis: redispatches (warning)
Cyril Bonté0dae5852010-02-03 00:26:28 +010011750 17. status: status (UP/DOWN/NOLB/MAINT/MAINT(via)...)
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +010011751 18. weight: server weight (server), total weight (backend)
11752 19. act: server is active (server), number of active servers (backend)
11753 20. bck: server is backup (server), number of backup servers (backend)
11754 21. chkfail: number of failed checks
11755 22. chkdown: number of UP->DOWN transitions
11756 23. lastchg: last status change (in seconds)
11757 24. downtime: total downtime (in seconds)
11758 25. qlimit: queue limit
11759 26. pid: process id (0 for first instance, 1 for second, ...)
11760 27. iid: unique proxy id
11761 28. sid: service id (unique inside a proxy)
11762 29. throttle: warm up status
11763 30. lbtot: total number of times a server was selected
11764 31. tracked: id of proxy/server if tracking is enabled
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +020011765 32. type (0=frontend, 1=backend, 2=server, 3=socket)
Krzysztof Piotr Oledzkidb57c6b2009-08-31 21:23:27 +020011766 33. rate: number of sessions per second over last elapsed second
11767 34. rate_lim: limit on new sessions per second
11768 35. rate_max: max number of new sessions per second
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +020011769 36. check_status: status of last health check, one of:
Cyril Bontéf0c60612010-02-06 14:44:47 +010011770 UNK -> unknown
11771 INI -> initializing
11772 SOCKERR -> socket error
11773 L4OK -> check passed on layer 4, no upper layers testing enabled
11774 L4TMOUT -> layer 1-4 timeout
11775 L4CON -> layer 1-4 connection problem, for example
11776 "Connection refused" (tcp rst) or "No route to host" (icmp)
11777 L6OK -> check passed on layer 6
11778 L6TOUT -> layer 6 (SSL) timeout
11779 L6RSP -> layer 6 invalid response - protocol error
11780 L7OK -> check passed on layer 7
11781 L7OKC -> check conditionally passed on layer 7, for example 404 with
11782 disable-on-404
11783 L7TOUT -> layer 7 (HTTP/SMTP) timeout
11784 L7RSP -> layer 7 invalid response - protocol error
11785 L7STS -> layer 7 response error, for example HTTP 5xx
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +020011786 37. check_code: layer5-7 code, if available
11787 38. check_duration: time in ms took to finish last health check
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011788 39. hrsp_1xx: http responses with 1xx code
11789 40. hrsp_2xx: http responses with 2xx code
11790 41. hrsp_3xx: http responses with 3xx code
11791 42. hrsp_4xx: http responses with 4xx code
11792 43. hrsp_5xx: http responses with 5xx code
11793 44. hrsp_other: http responses with other codes (protocol error)
Willy Tarreaud63335a2010-02-26 12:56:52 +010011794 45. hanafail: failed health checks details
11795 46. req_rate: HTTP requests per second over last elapsed second
11796 47. req_rate_max: max number of HTTP requests per second observed
11797 48. req_tot: total number of HTTP requests received
Willy Tarreauae526782010-03-04 20:34:23 +010011798 49. cli_abrt: number of data transfers aborted by the client
11799 50. srv_abrt: number of data transfers aborted by the server (inc. in eresp)
Willy Tarreau55058a72012-11-21 08:27:21 +010011800 51. comp_in: number of HTTP response bytes fed to the compressor
11801 52. comp_out: number of HTTP response bytes emitted by the compressor
11802 53. comp_byp: number of bytes that bypassed the HTTP compressor (CPU/BW limit)
Willy Tarreau11d4ec82012-11-26 00:49:03 +010011803 54. comp_rsp: number of HTTP responses that were compressed
Willy Tarreau844e3c52008-01-11 16:28:18 +010011804
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010011805
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200118069.2. Unix Socket commands
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010011807-------------------------
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010011808
Willy Tarreau468f4932013-08-01 16:50:16 +020011809The stats socket is not enabled by default. In order to enable it, it is
11810necessary to add one line in the global section of the haproxy configuration.
11811A second line is recommended to set a larger timeout, always appreciated when
11812issuing commands by hand :
11813
11814 global
11815 stats socket /var/run/haproxy.sock mode 600 level admin
11816 stats timeout 2m
11817
11818It is also possible to add multiple instances of the stats socket by repeating
11819the line, and make them listen to a TCP port instead of a UNIX socket. This is
11820never done by default because this is dangerous, but can be handy in some
11821situations :
11822
11823 global
11824 stats socket /var/run/haproxy.sock mode 600 level admin
11825 stats socket ipv4@192.168.0.1:9999 level admin
11826 stats timeout 2m
11827
11828To access the socket, an external utility such as "socat" is required. Socat is a
11829swiss-army knife to connect anything to anything. We use it to connect terminals
11830to the socket, or a couple of stdin/stdout pipes to it for scripts. The two main
11831syntaxes we'll use are the following :
11832
11833 # socat /var/run/haproxy.sock stdio
11834 # socat /var/run/haproxy.sock readline
11835
11836The first one is used with scripts. It is possible to send the output of a
11837script to haproxy, and pass haproxy's output to another script. That's useful
11838for retrieving counters or attack traces for example.
11839
11840The second one is only useful for issuing commands by hand. It has the benefit
11841that the terminal is handled by the readline library which supports line
11842editing and history, which is very convenient when issuing repeated commands
11843(eg: watch a counter).
11844
11845The socket supports two operation modes :
11846 - interactive
11847 - non-interactive
11848
11849The non-interactive mode is the default when socat connects to the socket. In
11850this mode, a single line may be sent. It is processed as a whole, responses are
11851sent back, and the connection closes after the end of the response. This is the
11852mode that scripts and monitoring tools use. It is possible to send multiple
11853commands in this mode, they need to be delimited by a semi-colon (';'). For
11854example :
11855
11856 # echo "show info;show stat;show table" | socat /var/run/haproxy stdio
11857
11858The interactive mode displays a prompt ('>') and waits for commands to be
11859entered on the line, then processes them, and displays the prompt again to wait
11860for a new command. This mode is entered via the "prompt" command which must be
11861sent on the first line in non-interactive mode. The mode is a flip switch, if
11862"prompt" is sent in interactive mode, it is disabled and the connection closes
11863after processing the last command of the same line.
11864
11865For this reason, when debugging by hand, it's quite common to start with the
11866"prompt" command :
11867
11868 # socat /var/run/haproxy readline
11869 prompt
11870 > show info
11871 ...
11872 >
11873
11874Since multiple commands may be issued at once, haproxy uses the empty line as a
11875delimiter to mark an end of output for each command, and takes care of ensuring
11876that no command can emit an empty line on output. A script can thus easily
11877parse the output even when multiple commands were pipelined on a single line.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010011878
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020011879It is important to understand that when multiple haproxy processes are started
11880on the same sockets, any process may pick up the request and will output its
11881own stats.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010011882
Willy Tarreau468f4932013-08-01 16:50:16 +020011883The list of commands currently supported on the stats socket is provided below.
11884If an unknown command is sent, haproxy displays the usage message which reminds
11885all supported commands. Some commands support a more complex syntax, generally
11886it will explain what part of the command is invalid when this happens.
11887
Willy Tarreaud63335a2010-02-26 12:56:52 +010011888clear counters
11889 Clear the max values of the statistics counters in each proxy (frontend &
11890 backend) and in each server. The cumulated counters are not affected. This
11891 can be used to get clean counters after an incident, without having to
11892 restart nor to clear traffic counters. This command is restricted and can
11893 only be issued on sockets configured for levels "operator" or "admin".
11894
11895clear counters all
11896 Clear all statistics counters in each proxy (frontend & backend) and in each
11897 server. This has the same effect as restarting. This command is restricted
11898 and can only be issued on sockets configured for level "admin".
11899
Simon Hormanc88b8872011-06-15 15:18:49 +090011900clear table <table> [ data.<type> <operator> <value> ] | [ key <key> ]
11901 Remove entries from the stick-table <table>.
11902
11903 This is typically used to unblock some users complaining they have been
11904 abusively denied access to a service, but this can also be used to clear some
11905 stickiness entries matching a server that is going to be replaced (see "show
11906 table" below for details). Note that sometimes, removal of an entry will be
11907 refused because it is currently tracked by a session. Retrying a few seconds
11908 later after the session ends is usual enough.
11909
11910 In the case where no options arguments are given all entries will be removed.
11911
11912 When the "data." form is used entries matching a filter applied using the
11913 stored data (see "stick-table" in section 4.2) are removed. A stored data
11914 type must be specified in <type>, and this data type must be stored in the
11915 table otherwise an error is reported. The data is compared according to
11916 <operator> with the 64-bit integer <value>. Operators are the same as with
11917 the ACLs :
11918
11919 - eq : match entries whose data is equal to this value
11920 - ne : match entries whose data is not equal to this value
11921 - le : match entries whose data is less than or equal to this value
11922 - ge : match entries whose data is greater than or equal to this value
11923 - lt : match entries whose data is less than this value
11924 - gt : match entries whose data is greater than this value
11925
11926 When the key form is used the entry <key> is removed. The key must be of the
Simon Horman619e3cc2011-06-15 15:18:52 +090011927 same type as the table, which currently is limited to IPv4, IPv6, integer and
11928 string.
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020011929
11930 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020011931 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Emeric Brun7c6b82e2010-09-24 16:34:28 +020011932 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020011933 >>> 0x80e6a4c: key=127.0.0.1 use=0 exp=3594729 gpc0=0 conn_rate(30000)=1 \
11934 bytes_out_rate(60000)=187
11935 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
11936 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020011937
11938 $ echo "clear table http_proxy key 127.0.0.1" | socat stdio /tmp/sock1
11939
11940 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Emeric Brun7c6b82e2010-09-24 16:34:28 +020011941 >>> # table: http_proxy, type: ip, size:204800, used:1
Willy Tarreau62a36c42010-08-17 15:53:10 +020011942 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
11943 bytes_out_rate(60000)=191
Simon Hormanc88b8872011-06-15 15:18:49 +090011944 $ echo "clear table http_proxy data.gpc0 eq 1" | socat stdio /tmp/sock1
11945 $ echo "show table http_proxy" | socat stdio /tmp/sock1
11946 >>> # table: http_proxy, type: ip, size:204800, used:1
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020011947
Willy Tarreau532a4502011-09-07 22:37:44 +020011948disable frontend <frontend>
11949 Mark the frontend as temporarily stopped. This corresponds to the mode which
11950 is used during a soft restart : the frontend releases the port but can be
11951 enabled again if needed. This should be used with care as some non-Linux OSes
11952 are unable to enable it back. This is intended to be used in environments
11953 where stopping a proxy is not even imaginable but a misconfigured proxy must
11954 be fixed. That way it's possible to release the port and bind it into another
11955 process to restore operations. The frontend will appear with status "STOP"
11956 on the stats page.
11957
11958 The frontend may be specified either by its name or by its numeric ID,
11959 prefixed with a sharp ('#').
11960
11961 This command is restricted and can only be issued on sockets configured for
11962 level "admin".
11963
Willy Tarreaud63335a2010-02-26 12:56:52 +010011964disable server <backend>/<server>
11965 Mark the server DOWN for maintenance. In this mode, no more checks will be
11966 performed on the server until it leaves maintenance.
11967 If the server is tracked by other servers, those servers will be set to DOWN
11968 during the maintenance.
11969
11970 In the statistics page, a server DOWN for maintenance will appear with a
11971 "MAINT" status, its tracking servers with the "MAINT(via)" one.
11972
11973 Both the backend and the server may be specified either by their name or by
Willy Tarreauf5f31922011-08-02 11:32:07 +020011974 their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010011975
11976 This command is restricted and can only be issued on sockets configured for
11977 level "admin".
11978
Willy Tarreau532a4502011-09-07 22:37:44 +020011979enable frontend <frontend>
11980 Resume a frontend which was temporarily stopped. It is possible that some of
11981 the listening ports won't be able to bind anymore (eg: if another process
11982 took them since the 'disable frontend' operation). If this happens, an error
11983 is displayed. Some operating systems might not be able to resume a frontend
11984 which was disabled.
11985
11986 The frontend may be specified either by its name or by its numeric ID,
11987 prefixed with a sharp ('#').
11988
11989 This command is restricted and can only be issued on sockets configured for
11990 level "admin".
11991
Willy Tarreaud63335a2010-02-26 12:56:52 +010011992enable server <backend>/<server>
11993 If the server was previously marked as DOWN for maintenance, this marks the
11994 server UP and checks are re-enabled.
11995
11996 Both the backend and the server may be specified either by their name or by
Willy Tarreauf5f31922011-08-02 11:32:07 +020011997 their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010011998
11999 This command is restricted and can only be issued on sockets configured for
12000 level "admin".
12001
12002get weight <backend>/<server>
12003 Report the current weight and the initial weight of server <server> in
12004 backend <backend> or an error if either doesn't exist. The initial weight is
12005 the one that appears in the configuration file. Both are normally equal
12006 unless the current weight has been changed. Both the backend and the server
12007 may be specified either by their name or by their numeric ID, prefixed with a
Willy Tarreauf5f31922011-08-02 11:32:07 +020012008 sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010012009
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020012010help
12011 Print the list of known keywords and their basic usage. The same help screen
12012 is also displayed for unknown commands.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010012013
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020012014prompt
12015 Toggle the prompt at the beginning of the line and enter or leave interactive
12016 mode. In interactive mode, the connection is not closed after a command
12017 completes. Instead, the prompt will appear again, indicating the user that
12018 the interpreter is waiting for a new command. The prompt consists in a right
12019 angle bracket followed by a space "> ". This mode is particularly convenient
12020 when one wants to periodically check information such as stats or errors.
12021 It is also a good idea to enter interactive mode before issuing a "help"
12022 command.
12023
12024quit
12025 Close the connection when in interactive mode.
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010012026
Willy Tarreau2a0f4d22011-08-02 11:49:05 +020012027set maxconn frontend <frontend> <value>
Willy Tarreau3c7a79d2012-09-26 21:07:15 +020012028 Dynamically change the specified frontend's maxconn setting. Any positive
12029 value is allowed including zero, but setting values larger than the global
12030 maxconn does not make much sense. If the limit is increased and connections
12031 were pending, they will immediately be accepted. If it is lowered to a value
12032 below the current number of connections, new connections acceptation will be
Willy Tarreau2a0f4d22011-08-02 11:49:05 +020012033 delayed until the threshold is reached. The frontend might be specified by
12034 either its name or its numeric ID prefixed with a sharp ('#').
12035
Willy Tarreau91886b62011-09-07 14:38:31 +020012036set maxconn global <maxconn>
12037 Dynamically change the global maxconn setting within the range defined by the
12038 initial global maxconn setting. If it is increased and connections were
12039 pending, they will immediately be accepted. If it is lowered to a value below
12040 the current number of connections, new connections acceptation will be
12041 delayed until the threshold is reached. A value of zero restores the initial
12042 setting.
12043
Willy Tarreauf5b22872011-09-07 16:13:44 +020012044set rate-limit connections global <value>
12045 Change the process-wide connection rate limit, which is set by the global
12046 'maxconnrate' setting. A value of zero disables the limitation. This limit
12047 applies to all frontends and the change has an immediate effect. The value
12048 is passed in number of connections per second.
12049
William Lallemandd85f9172012-11-09 17:05:39 +010012050set rate-limit http-compression global <value>
12051 Change the maximum input compression rate, which is set by the global
12052 'maxcomprate' setting. A value of zero disables the limitation. The value is
William Lallemand096f5542012-11-19 17:26:05 +010012053 passed in number of kilobytes per second. The value is available in the "show
12054 info" on the line "CompressBpsRateLim" in bytes.
William Lallemandd85f9172012-11-09 17:05:39 +010012055
Willy Tarreau47060b62013-08-01 21:11:42 +020012056set table <table> key <key> [data.<data_type> <value>]*
Willy Tarreau654694e2012-06-07 01:03:16 +020012057 Create or update a stick-table entry in the table. If the key is not present,
12058 an entry is inserted. See stick-table in section 4.2 to find all possible
12059 values for <data_type>. The most likely use consists in dynamically entering
12060 entries for source IP addresses, with a flag in gpc0 to dynamically block an
Willy Tarreau47060b62013-08-01 21:11:42 +020012061 IP address or affect its quality of service. It is possible to pass multiple
12062 data_types in a single call.
Willy Tarreau654694e2012-06-07 01:03:16 +020012063
Willy Tarreaud63335a2010-02-26 12:56:52 +010012064set timeout cli <delay>
12065 Change the CLI interface timeout for current connection. This can be useful
12066 during long debugging sessions where the user needs to constantly inspect
12067 some indicators without being disconnected. The delay is passed in seconds.
12068
12069set weight <backend>/<server> <weight>[%]
12070 Change a server's weight to the value passed in argument. If the value ends
12071 with the '%' sign, then the new weight will be relative to the initially
Simon Horman58b5d292013-02-12 10:45:52 +090012072 configured weight. Absolute weights are permitted between 0 and 256.
12073 Relative weights must be positive with the resulting absolute weight is
12074 capped at 256. Servers which are part of a farm running a static
12075 load-balancing algorithm have stricter limitations because the weight
12076 cannot change once set. Thus for these servers, the only accepted values
12077 are 0 and 100% (or 0 and the initial weight). Changes take effect
12078 immediately, though certain LB algorithms require a certain amount of
12079 requests to consider changes. A typical usage of this command is to
12080 disable a server during an update by setting its weight to zero, then to
12081 enable it again after the update by setting it back to 100%. This command
12082 is restricted and can only be issued on sockets configured for level
12083 "admin". Both the backend and the server may be specified either by their
12084 name or by their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010012085
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010012086show errors [<iid>]
12087 Dump last known request and response errors collected by frontends and
12088 backends. If <iid> is specified, the limit the dump to errors concerning
Willy Tarreau6162db22009-10-10 17:13:00 +020012089 either frontend or backend whose ID is <iid>. This command is restricted
12090 and can only be issued on sockets configured for levels "operator" or
12091 "admin".
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010012092
12093 The errors which may be collected are the last request and response errors
12094 caused by protocol violations, often due to invalid characters in header
12095 names. The report precisely indicates what exact character violated the
12096 protocol. Other important information such as the exact date the error was
12097 detected, frontend and backend names, the server name (when known), the
12098 internal session ID and the source address which has initiated the session
12099 are reported too.
12100
12101 All characters are returned, and non-printable characters are encoded. The
12102 most common ones (\t = 9, \n = 10, \r = 13 and \e = 27) are encoded as one
12103 letter following a backslash. The backslash itself is encoded as '\\' to
12104 avoid confusion. Other non-printable characters are encoded '\xNN' where
12105 NN is the two-digits hexadecimal representation of the character's ASCII
12106 code.
12107
12108 Lines are prefixed with the position of their first character, starting at 0
12109 for the beginning of the buffer. At most one input line is printed per line,
12110 and large lines will be broken into multiple consecutive output lines so that
12111 the output never goes beyond 79 characters wide. It is easy to detect if a
12112 line was broken, because it will not end with '\n' and the next line's offset
12113 will be followed by a '+' sign, indicating it is a continuation of previous
12114 line.
12115
12116 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020012117 $ echo "show errors" | socat stdio /tmp/sock1
12118 >>> [04/Mar/2009:15:46:56.081] backend http-in (#2) : invalid response
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010012119 src 127.0.0.1, session #54, frontend fe-eth0 (#1), server s2 (#1)
12120 response length 213 bytes, error at position 23:
12121
12122 00000 HTTP/1.0 200 OK\r\n
12123 00017 header/bizarre:blah\r\n
12124 00038 Location: blah\r\n
12125 00054 Long-line: this is a very long line which should b
12126 00104+ e broken into multiple lines on the output buffer,
12127 00154+ otherwise it would be too large to print in a ter
12128 00204+ minal\r\n
12129 00211 \r\n
12130
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012131 In the example above, we see that the backend "http-in" which has internal
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010012132 ID 2 has blocked an invalid response from its server s2 which has internal
12133 ID 1. The request was on session 54 initiated by source 127.0.0.1 and
12134 received by frontend fe-eth0 whose ID is 1. The total response length was
12135 213 bytes when the error was detected, and the error was at byte 23. This
12136 is the slash ('/') in header name "header/bizarre", which is not a valid
12137 HTTP character for a header name.
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010012138
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020012139show info
12140 Dump info about haproxy status on current process.
12141
12142show sess
12143 Dump all known sessions. Avoid doing this on slow connections as this can
Willy Tarreau6162db22009-10-10 17:13:00 +020012144 be huge. This command is restricted and can only be issued on sockets
12145 configured for levels "operator" or "admin".
12146
Willy Tarreau66dc20a2010-03-05 17:53:32 +010012147show sess <id>
12148 Display a lot of internal information about the specified session identifier.
12149 This identifier is the first field at the beginning of the lines in the dumps
12150 of "show sess" (it corresponds to the session pointer). Those information are
12151 useless to most users but may be used by haproxy developers to troubleshoot a
12152 complex bug. The output format is intentionally not documented so that it can
Willy Tarreau76153662012-11-26 01:16:39 +010012153 freely evolve depending on demands. The special id "all" dumps the states of
12154 all sessions, which can be avoided as much as possible as it is highly CPU
12155 intensive and can take a lot of time.
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020012156
12157show stat [<iid> <type> <sid>]
12158 Dump statistics in the CSV format. By passing <id>, <type> and <sid>, it is
12159 possible to dump only selected items :
12160 - <iid> is a proxy ID, -1 to dump everything
12161 - <type> selects the type of dumpable objects : 1 for frontends, 2 for
12162 backends, 4 for servers, -1 for everything. These values can be ORed,
12163 for example:
12164 1 + 2 = 3 -> frontend + backend.
12165 1 + 2 + 4 = 7 -> frontend + backend + server.
12166 - <sid> is a server ID, -1 to dump everything from the selected proxy.
12167
12168 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020012169 $ echo "show info;show stat" | socat stdio unix-connect:/tmp/sock1
12170 >>> Name: HAProxy
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020012171 Version: 1.4-dev2-49
12172 Release_date: 2009/09/23
12173 Nbproc: 1
12174 Process_num: 1
12175 (...)
12176
12177 # pxname,svname,qcur,qmax,scur,smax,slim,stot,bin,bout,dreq, (...)
12178 stats,FRONTEND,,,0,0,1000,0,0,0,0,0,0,,,,,OPEN,,,,,,,,,1,1,0, (...)
12179 stats,BACKEND,0,0,0,0,1000,0,0,0,0,0,,0,0,0,0,UP,0,0,0,,0,250,(...)
12180 (...)
12181 www1,BACKEND,0,0,0,0,1000,0,0,0,0,0,,0,0,0,0,UP,1,1,0,,0,250, (...)
12182
12183 $
12184
12185 Here, two commands have been issued at once. That way it's easy to find
12186 which process the stats apply to in multi-process mode. Notice the empty
12187 line after the information output which marks the end of the first block.
12188 A similar empty line appears at the end of the second block (stats) so that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012189 the reader knows the output has not been truncated.
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020012190
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012191show table
12192 Dump general information on all known stick-tables. Their name is returned
12193 (the name of the proxy which holds them), their type (currently zero, always
12194 IP), their size in maximum possible number of entries, and the number of
12195 entries currently in use.
12196
12197 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020012198 $ echo "show table" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090012199 >>> # table: front_pub, type: ip, size:204800, used:171454
12200 >>> # table: back_rdp, type: ip, size:204800, used:0
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012201
Simon Horman17bce342011-06-15 15:18:47 +090012202show table <name> [ data.<type> <operator> <value> ] | [ key <key> ]
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012203 Dump contents of stick-table <name>. In this mode, a first line of generic
12204 information about the table is reported as with "show table", then all
12205 entries are dumped. Since this can be quite heavy, it is possible to specify
Simon Horman17bce342011-06-15 15:18:47 +090012206 a filter in order to specify what entries to display.
12207
12208 When the "data." form is used the filter applies to the stored data (see
12209 "stick-table" in section 4.2). A stored data type must be specified
12210 in <type>, and this data type must be stored in the table otherwise an
12211 error is reported. The data is compared according to <operator> with the
12212 64-bit integer <value>. Operators are the same as with the ACLs :
12213
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012214 - eq : match entries whose data is equal to this value
12215 - ne : match entries whose data is not equal to this value
12216 - le : match entries whose data is less than or equal to this value
12217 - ge : match entries whose data is greater than or equal to this value
12218 - lt : match entries whose data is less than this value
12219 - gt : match entries whose data is greater than this value
12220
Simon Hormanc88b8872011-06-15 15:18:49 +090012221
12222 When the key form is used the entry <key> is shown. The key must be of the
Simon Horman619e3cc2011-06-15 15:18:52 +090012223 same type as the table, which currently is limited to IPv4, IPv6, integer,
12224 and string.
Simon Horman17bce342011-06-15 15:18:47 +090012225
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012226 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020012227 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090012228 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020012229 >>> 0x80e6a4c: key=127.0.0.1 use=0 exp=3594729 gpc0=0 conn_rate(30000)=1 \
12230 bytes_out_rate(60000)=187
12231 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
12232 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012233
Willy Tarreau62a36c42010-08-17 15:53:10 +020012234 $ echo "show table http_proxy data.gpc0 gt 0" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090012235 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020012236 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
12237 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012238
Willy Tarreau62a36c42010-08-17 15:53:10 +020012239 $ echo "show table http_proxy data.conn_rate gt 5" | \
12240 socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090012241 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020012242 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
12243 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012244
Simon Horman17bce342011-06-15 15:18:47 +090012245 $ echo "show table http_proxy key 127.0.0.2" | \
12246 socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090012247 >>> # table: http_proxy, type: ip, size:204800, used:2
Simon Horman17bce342011-06-15 15:18:47 +090012248 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
12249 bytes_out_rate(60000)=191
12250
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012251 When the data criterion applies to a dynamic value dependent on time such as
12252 a bytes rate, the value is dynamically computed during the evaluation of the
12253 entry in order to decide whether it has to be dumped or not. This means that
12254 such a filter could match for some time then not match anymore because as
12255 time goes, the average event rate drops.
12256
12257 It is possible to use this to extract lists of IP addresses abusing the
12258 service, in order to monitor them or even blacklist them in a firewall.
12259 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020012260 $ echo "show table http_proxy data.gpc0 gt 0" \
12261 | socat stdio /tmp/sock1 \
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012262 | fgrep 'key=' | cut -d' ' -f2 | cut -d= -f2 > abusers-ip.txt
12263 ( or | awk '/key/{ print a[split($2,a,"=")]; }' )
Krzysztof Piotr Oledzki719e7262009-10-04 15:02:46 +020012264
Willy Tarreau532a4502011-09-07 22:37:44 +020012265shutdown frontend <frontend>
12266 Completely delete the specified frontend. All the ports it was bound to will
12267 be released. It will not be possible to enable the frontend anymore after
12268 this operation. This is intended to be used in environments where stopping a
12269 proxy is not even imaginable but a misconfigured proxy must be fixed. That
12270 way it's possible to release the port and bind it into another process to
12271 restore operations. The frontend will not appear at all on the stats page
12272 once it is terminated.
12273
12274 The frontend may be specified either by its name or by its numeric ID,
12275 prefixed with a sharp ('#').
12276
12277 This command is restricted and can only be issued on sockets configured for
12278 level "admin".
12279
Willy Tarreaua295edc2011-09-07 23:21:03 +020012280shutdown session <id>
12281 Immediately terminate the session matching the specified session identifier.
12282 This identifier is the first field at the beginning of the lines in the dumps
12283 of "show sess" (it corresponds to the session pointer). This can be used to
12284 terminate a long-running session without waiting for a timeout or when an
12285 endless transfer is ongoing. Such terminated sessions are reported with a 'K'
12286 flag in the logs.
12287
Willy Tarreau52b2d222011-09-07 23:48:48 +020012288shutdown sessions <backend>/<server>
12289 Immediately terminate all the sessions attached to the specified server. This
12290 can be used to terminate long-running sessions after a server is put into
12291 maintenance mode, for instance. Such terminated sessions are reported with a
12292 'K' flag in the logs.
12293
Willy Tarreau0ba27502007-12-24 16:55:16 +010012294/*
12295 * Local variables:
12296 * fill-column: 79
12297 * End:
12298 */