blob: ae7bc1a359f09dfcf95338c98ea1ae0f0e1e3054 [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 Tarreau11f64d62013-12-16 02:32:37 +01007 2013/12/16
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>
Simon Hormand60d6912013-11-25 10:46:36 +0900780 Sometimes it is desirable to avoid sending agent and health checks to
781 servers at exact intervals, for instance when many logical servers are
782 located on the same physical server. With the help of this parameter, it
783 becomes possible to add some randomness in the check interval between 0
784 and +/- 50%. A value between 2 and 5 seems to show good results. The
785 default value remains at 0.
Willy Tarreaufe255b72007-10-14 23:09:26 +0200786
Willy Tarreau27a674e2009-08-17 07:23:33 +0200787tune.bufsize <number>
788 Sets the buffer size to this size (in bytes). Lower values allow more
789 sessions to coexist in the same amount of RAM, and higher values allow some
790 applications with very large cookies to work. The default value is 16384 and
791 can be changed at build time. It is strongly recommended not to change this
792 from the default value, as very low values will break some services such as
793 statistics, and values larger than default size will increase memory usage,
794 possibly causing the system to run out of memory. At least the global maxconn
795 parameter should be decreased by the same factor as this one is increased.
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +0400796 If HTTP request is larger than (tune.bufsize - tune.maxrewrite), haproxy will
797 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
798 than this size, haproxy will return HTTP 502 (Bad Gateway).
Willy Tarreau27a674e2009-08-17 07:23:33 +0200799
Willy Tarreau43961d52010-10-04 20:39:20 +0200800tune.chksize <number>
801 Sets the check buffer size to this size (in bytes). Higher values may help
802 find string or regex patterns in very large pages, though doing so may imply
803 more memory and CPU usage. The default value is 16384 and can be changed at
804 build time. It is not recommended to change this value, but to use better
805 checks whenever possible.
806
William Lallemandf3747832012-11-09 12:33:10 +0100807tune.comp.maxlevel <number>
808 Sets the maximum compression level. The compression level affects CPU
809 usage during compression. This value affects CPU usage during compression.
810 Each session using compression initializes the compression algorithm with
811 this value. The default value is 1.
812
Willy Tarreau193b8c62012-11-22 00:17:38 +0100813tune.http.cookielen <number>
814 Sets the maximum length of captured cookies. This is the maximum value that
815 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
816 will automatically be truncated to this one. It is important not to set too
817 high a value because all cookie captures still allocate this size whatever
818 their configured value (they share a same pool). This value is per request
819 per response, so the memory allocated is twice this value per connection.
820 When not specified, the limit is set to 63 characters. It is recommended not
821 to change this value.
822
Willy Tarreauac1932d2011-10-24 19:14:41 +0200823tune.http.maxhdr <number>
824 Sets the maximum number of headers in a request. When a request comes with a
825 number of headers greater than this value (including the first line), it is
826 rejected with a "400 Bad Request" status code. Similarly, too large responses
827 are blocked with "502 Bad Gateway". The default value is 101, which is enough
828 for all usages, considering that the widely deployed Apache server uses the
829 same limit. It can be useful to push this limit further to temporarily allow
830 a buggy application to work by the time it gets fixed. Keep in mind that each
831 new header consumes 32bits of memory for each session, so don't push this
832 limit too high.
833
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100834tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +0100835 Sets the maximum number of consecutive connections a process may accept in a
836 row before switching to other work. In single process mode, higher numbers
837 give better performance at high connection rates. However in multi-process
838 modes, keeping a bit of fairness between processes generally is better to
839 increase performance. This value applies individually to each listener, so
840 that the number of processes a listener is bound to is taken into account.
841 This value defaults to 64. In multi-process mode, it is divided by twice
842 the number of processes the listener is bound to. Setting this value to -1
843 completely disables the limitation. It should normally not be needed to tweak
844 this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100845
846tune.maxpollevents <number>
847 Sets the maximum amount of events that can be processed at once in a call to
848 the polling system. The default value is adapted to the operating system. It
849 has been noticed that reducing it below 200 tends to slightly decrease
850 latency at the expense of network bandwidth, and increasing it above 200
851 tends to trade latency for slightly increased bandwidth.
852
Willy Tarreau27a674e2009-08-17 07:23:33 +0200853tune.maxrewrite <number>
854 Sets the reserved buffer space to this size in bytes. The reserved space is
855 used for header rewriting or appending. The first reads on sockets will never
856 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
857 bufsize, though that does not make much sense since there are rarely large
858 numbers of headers to add. Setting it too high prevents processing of large
859 requests or responses. Setting it too low prevents addition of new headers
860 to already large requests or to POST requests. It is generally wise to set it
861 to about 1024. It is automatically readjusted to half of bufsize if it is
862 larger than that. This means you don't have to worry about it when changing
863 bufsize.
864
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200865tune.pipesize <number>
866 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
867 are the default size for the system. But sometimes when using TCP splicing,
868 it can improve performance to increase pipe sizes, especially if it is
869 suspected that pipes are not filled and that many calls to splice() are
870 performed. This has an impact on the kernel's memory footprint, so this must
871 not be changed if impacts are not understood.
872
Willy Tarreaue803de22010-01-21 17:43:04 +0100873tune.rcvbuf.client <number>
874tune.rcvbuf.server <number>
875 Forces the kernel socket receive buffer size on the client or the server side
876 to the specified value in bytes. This value applies to all TCP/HTTP frontends
877 and backends. It should normally never be set, and the default size (0) lets
878 the kernel autotune this value depending on the amount of available memory.
879 However it can sometimes help to set it to very low values (eg: 4096) in
880 order to save kernel memory by preventing it from buffering too large amounts
881 of received data. Lower values will significantly increase CPU usage though.
882
883tune.sndbuf.client <number>
884tune.sndbuf.server <number>
885 Forces the kernel socket send buffer size on the client or the server side to
886 the specified value in bytes. This value applies to all TCP/HTTP frontends
887 and backends. It should normally never be set, and the default size (0) lets
888 the kernel autotune this value depending on the amount of available memory.
889 However it can sometimes help to set it to very low values (eg: 4096) in
890 order to save kernel memory by preventing it from buffering too large amounts
891 of received data. Lower values will significantly increase CPU usage though.
892 Another use case is to prevent write timeouts with extremely slow clients due
893 to the kernel waiting for a large part of the buffer to be read before
894 notifying haproxy again.
895
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100896tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +0100897 Sets the size of the global SSL session cache, in a number of blocks. A block
898 is large enough to contain an encoded session without peer certificate.
899 An encoded session with peer certificate is stored in multiple blocks
900 depending on the size of the peer certificate. A block use approximatively
901 200 bytes of memory. The default value may be forced at build time, otherwise
902 defaults to 20000. When the cache is full, the most idle entries are purged
903 and reassigned. Higher values reduce the occurrence of such a purge, hence
904 the number of CPU-intensive SSL handshakes by ensuring that all users keep
905 their session as long as possible. All entries are pre-allocated upon startup
Emeric Brun22890a12012-12-28 14:41:32 +0100906 and are shared between all processes if "nbproc" is greater than 1. Setting
907 this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100908
Emeric Brun4f65bff2012-11-16 15:11:00 +0100909tune.ssl.lifetime <timeout>
910 Sets how long a cached SSL session may remain valid. This time is expressed
911 in seconds and defaults to 300 (5 mn). It is important to understand that it
912 does not guarantee that sessions will last that long, because if the cache is
913 full, the longest idle sessions will be purged despite their configured
914 lifetime. The real usefulness of this setting is to prevent sessions from
915 being used for too long.
916
Willy Tarreaubfd59462013-02-21 07:46:09 +0100917tune.ssl.maxrecord <number>
918 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
919 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
920 data only once it has received a full record. With large records, it means
921 that clients might have to download up to 16kB of data before starting to
922 process them. Limiting the value can improve page load times on browsers
923 located over high latency or low bandwidth networks. It is suggested to find
924 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
925 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
926 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
927 2859 gave good results during tests. Use "strace -e trace=write" to find the
928 best value.
929
William Lallemanda509e4c2012-11-07 16:54:34 +0100930tune.zlib.memlevel <number>
931 Sets the memLevel parameter in zlib initialization for each session. It
932 defines how much memory should be allocated for the intenal compression
933 state. A value of 1 uses minimum memory but is slow and reduces compression
934 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
935 between 1 and 9. The default value is 8.
936
937tune.zlib.windowsize <number>
938 Sets the window size (the size of the history buffer) as a parameter of the
939 zlib initialization for each session. Larger values of this parameter result
940 in better compression at the expense of memory usage. Can be a value between
941 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200942
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009433.3. Debugging
944--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +0200945
946debug
947 Enables debug mode which dumps to stdout all exchanges, and disables forking
948 into background. It is the equivalent of the command-line argument "-d". It
949 should never be used in a production configuration since it may prevent full
950 system startup.
951
952quiet
953 Do not display any message during startup. It is equivalent to the command-
954 line argument "-q".
955
Emeric Brunf099e792010-09-27 12:05:28 +0200956
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01009573.4. Userlists
958--------------
959It is possible to control access to frontend/backend/listen sections or to
960http stats by allowing only authenticated and authorized users. To do this,
961it is required to create at least one userlist and to define users.
962
963userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +0100964 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100965 used to store authentication & authorization data for independent customers.
966
967group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +0100968 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100969 attach users to this group by using a comma separated list of names
970 proceeded by "users" keyword.
971
Cyril Bontéf0c60612010-02-06 14:44:47 +0100972user <username> [password|insecure-password <password>]
973 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100974 Adds user <username> to the current userlist. Both secure (encrypted) and
975 insecure (unencrypted) passwords can be used. Encrypted passwords are
Cyril Bonté78caf842010-03-10 22:41:43 +0100976 evaluated using the crypt(3) function so depending of the system's
977 capabilities, different algorithms are supported. For example modern Glibc
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100978 based Linux system supports MD5, SHA-256, SHA-512 and of course classic,
979 DES-based method of crypting passwords.
980
981
982 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +0100983 userlist L1
984 group G1 users tiger,scott
985 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100986
Cyril Bontéf0c60612010-02-06 14:44:47 +0100987 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
988 user scott insecure-password elgato
989 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100990
Cyril Bontéf0c60612010-02-06 14:44:47 +0100991 userlist L2
992 group G1
993 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100994
Cyril Bontéf0c60612010-02-06 14:44:47 +0100995 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
996 user scott insecure-password elgato groups G1,G2
997 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100998
999 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001000
Emeric Brunf099e792010-09-27 12:05:28 +02001001
10023.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02001003----------
Emeric Brunf099e792010-09-27 12:05:28 +02001004It is possible to synchronize server entries in stick tables between several
1005haproxy instances over TCP connections in a multi-master fashion. Each instance
1006pushes its local updates and insertions to remote peers. Server IDs are used to
1007identify servers remotely, so it is important that configurations look similar
1008or at least that the same IDs are forced on each server on all participants.
1009Interrupted exchanges are automatically detected and recovered from the last
1010known point. In addition, during a soft restart, the old process connects to
1011the new one using such a TCP connection to push all its entries before the new
1012process tries to connect to other peers. That ensures very fast replication
1013during a reload, it typically takes a fraction of a second even for large
1014tables.
1015
1016peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001017 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02001018 which is referenced by one or more stick-tables.
1019
1020peer <peername> <ip>:<port>
1021 Defines a peer inside a peers section.
1022 If <peername> is set to the local peer name (by default hostname, or forced
1023 using "-L" command line option), haproxy will listen for incoming remote peer
1024 connection on <ip>:<port>. Otherwise, <ip>:<port> defines where to connect to
1025 to join the remote peer, and <peername> is used at the protocol level to
1026 identify and validate the remote peer on the server side.
1027
1028 During a soft restart, local peer <ip>:<port> is used by the old instance to
1029 connect the new one and initiate a complete replication (teaching process).
1030
1031 It is strongly recommended to have the exact same peers declaration on all
1032 peers and to only rely on the "-L" command line argument to change the local
1033 peer name. This makes it easier to maintain coherent configuration files
1034 across all peers.
1035
Willy Tarreaudad36a32013-03-11 01:20:04 +01001036 Any part of the address string may reference any number of environment
1037 variables by preceding their name with a dollar sign ('$') and optionally
1038 enclosing them with braces ('{}'), similarly to what is done in Bourne shell.
1039
Cyril Bontédc4d9032012-04-08 21:57:39 +02001040 Example:
Emeric Brunf099e792010-09-27 12:05:28 +02001041 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001042 peer haproxy1 192.168.0.1:1024
1043 peer haproxy2 192.168.0.2:1024
1044 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02001045
1046 backend mybackend
1047 mode tcp
1048 balance roundrobin
1049 stick-table type ip size 20k peers mypeers
1050 stick on src
1051
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001052 server srv1 192.168.0.30:80
1053 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02001054
1055
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010564. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02001057----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001058
Willy Tarreau6a06a402007-07-15 20:15:28 +02001059Proxy configuration can be located in a set of sections :
1060 - defaults <name>
1061 - frontend <name>
1062 - backend <name>
1063 - listen <name>
1064
1065A "defaults" section sets default parameters for all other sections following
1066its declaration. Those default parameters are reset by the next "defaults"
1067section. See below for the list of parameters which can be set in a "defaults"
Willy Tarreau0ba27502007-12-24 16:55:16 +01001068section. The name is optional but its use is encouraged for better readability.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001069
1070A "frontend" section describes a set of listening sockets accepting client
1071connections.
1072
1073A "backend" section describes a set of servers to which the proxy will connect
1074to forward incoming connections.
1075
1076A "listen" section defines a complete proxy with its frontend and backend
1077parts combined in one section. It is generally useful for TCP-only traffic.
1078
Willy Tarreau0ba27502007-12-24 16:55:16 +01001079All proxy names must be formed from upper and lower case letters, digits,
1080'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
1081case-sensitive, which means that "www" and "WWW" are two different proxies.
1082
1083Historically, all proxy names could overlap, it just caused troubles in the
1084logs. Since the introduction of content switching, it is mandatory that two
1085proxies with overlapping capabilities (frontend/backend) have different names.
1086However, it is still permitted that a frontend and a backend share the same
1087name, as this configuration seems to be commonly encountered.
1088
1089Right now, two major proxy modes are supported : "tcp", also known as layer 4,
1090and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001091bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01001092protocol, and can interact with it by allowing, blocking, switching, adding,
1093modifying, or removing arbitrary contents in requests or responses, based on
1094arbitrary criteria.
1095
Willy Tarreau0ba27502007-12-24 16:55:16 +01001096
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010974.1. Proxy keywords matrix
1098--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001099
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001100The following list of keywords is supported. Most of them may only be used in a
1101limited set of section types. Some of them are marked as "deprecated" because
1102they are inherited from an old syntax which may be confusing or functionally
1103limited, and there are new recommended keywords to replace them. Keywords
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001104marked with "(*)" can be optionally inverted using the "no" prefix, eg. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001105option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02001106and must be disabled for a specific instance. Such options may also be prefixed
1107with "default" in order to restore default settings regardless of what has been
1108specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001109
Willy Tarreau6a06a402007-07-15 20:15:28 +02001110
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001111 keyword defaults frontend listen backend
1112------------------------------------+----------+----------+---------+---------
1113acl - X X X
1114appsession - - X X
1115backlog X X X -
1116balance X - X X
1117bind - X X -
1118bind-process X X X X
1119block - X X X
1120capture cookie - X X -
1121capture request header - X X -
1122capture response header - X X -
1123clitimeout (deprecated) X X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02001124compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001125contimeout (deprecated) X - X X
1126cookie X - X X
1127default-server X - X X
1128default_backend X X X -
1129description - X X X
1130disabled X X X X
1131dispatch - - X X
1132enabled X X X X
1133errorfile X X X X
1134errorloc X X X X
1135errorloc302 X X X X
1136-- keyword -------------------------- defaults - frontend - listen -- backend -
1137errorloc303 X X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001138force-persist - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001139fullconn X - X X
1140grace X X X X
1141hash-type X - X X
1142http-check disable-on-404 X - X X
Willy Tarreaubd741542010-03-16 18:46:54 +01001143http-check expect - - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02001144http-check send-state X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001145http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02001146http-response - X X X
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02001147tcp-check expect - - X X
1148tcp-check send - - X X
1149tcp-check send-binary - - X X
Baptiste Assmann2c42ef52013-10-09 21:57:02 +02001150http-send-name-header - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001151id - X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001152ignore-persist - X X X
William Lallemand0f99e342011-10-12 17:50:54 +02001153log (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001154maxconn X X X -
1155mode X X X X
1156monitor fail - X X -
1157monitor-net X X X -
1158monitor-uri X X X -
1159option abortonclose (*) X - X X
1160option accept-invalid-http-request (*) X X X -
1161option accept-invalid-http-response (*) X - X X
1162option allbackups (*) X - X X
1163option checkcache (*) X - X X
1164option clitcpka (*) X X X -
1165option contstats (*) X X X -
1166option dontlog-normal (*) X X X -
1167option dontlognull (*) X X X -
1168option forceclose (*) X X X X
1169-- keyword -------------------------- defaults - frontend - listen -- backend -
1170option forwardfor X X X X
Willy Tarreau16bfb022010-01-16 19:48:41 +01001171option http-keep-alive (*) X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02001172option http-no-delay (*) X X X X
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02001173option http-pretend-keepalive (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001174option http-server-close (*) X X X X
1175option http-use-proxy-header (*) X X X -
1176option httpchk X - X X
1177option httpclose (*) X X X X
1178option httplog X X X X
1179option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001180option independent-streams (*) X X X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02001181option ldap-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001182option log-health-checks (*) X - X X
1183option log-separate-errors (*) X X X -
1184option logasap (*) X X X -
1185option mysql-check X - X X
Rauf Kuliyev38b41562011-01-04 15:14:13 +01001186option pgsql-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001187option nolinger (*) X X X X
1188option originalto X X X X
1189option persist (*) X - X X
1190option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02001191option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001192option smtpchk X - X X
1193option socket-stats (*) X X X -
1194option splice-auto (*) X X X X
1195option splice-request (*) X X X X
1196option splice-response (*) X X X X
1197option srvtcpka (*) X - X X
1198option ssl-hello-chk X - X X
1199-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreaued179852013-12-16 01:07:00 +01001200option tcp-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001201option tcp-smart-accept (*) X X X -
1202option tcp-smart-connect (*) X - X X
1203option tcpka X X X X
1204option tcplog X X X X
1205option transparent (*) X - X X
1206persist rdp-cookie X - X X
1207rate-limit sessions X X X -
1208redirect - X X X
1209redisp (deprecated) X - X X
1210redispatch (deprecated) X - X X
1211reqadd - X X X
1212reqallow - X X X
1213reqdel - X X X
1214reqdeny - X X X
1215reqiallow - X X X
1216reqidel - X X X
1217reqideny - X X X
1218reqipass - X X X
1219reqirep - X X X
1220reqisetbe - X X X
1221reqitarpit - X X X
1222reqpass - X X X
1223reqrep - X X X
1224-- keyword -------------------------- defaults - frontend - listen -- backend -
1225reqsetbe - X X X
1226reqtarpit - X X X
1227retries X - X X
1228rspadd - X X X
1229rspdel - X X X
1230rspdeny - X X X
1231rspidel - X X X
1232rspideny - X X X
1233rspirep - X X X
1234rsprep - X X X
1235server - - X X
1236source X - X X
1237srvtimeout (deprecated) X - X X
Cyril Bonté66c327d2010-10-12 00:14:37 +02001238stats admin - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001239stats auth X - X X
1240stats enable X - X X
1241stats hide-version X - X X
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02001242stats http-request - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001243stats realm X - X X
1244stats refresh X - X X
1245stats scope X - X X
1246stats show-desc X - X X
1247stats show-legends X - X X
1248stats show-node X - X X
1249stats uri X - X X
1250-- keyword -------------------------- defaults - frontend - listen -- backend -
1251stick match - - X X
1252stick on - - X X
1253stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02001254stick store-response - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001255stick-table - - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02001256tcp-request connection - X X -
1257tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02001258tcp-request inspect-delay - X X X
Emeric Brun0a3b67f2010-09-24 15:34:53 +02001259tcp-response content - - X X
1260tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001261timeout check X - X X
1262timeout client X X X -
1263timeout clitimeout (deprecated) X X X -
1264timeout connect X - X X
1265timeout contimeout (deprecated) X - X X
1266timeout http-keep-alive X X X X
1267timeout http-request X X X X
1268timeout queue X - X X
1269timeout server X - X X
1270timeout srvtimeout (deprecated) X - X X
1271timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02001272timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001273transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01001274unique-id-format X X X -
1275unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001276use_backend - X X -
Willy Tarreau4a5cade2012-04-05 21:09:48 +02001277use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001278------------------------------------+----------+----------+---------+---------
1279 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02001280
Willy Tarreau0ba27502007-12-24 16:55:16 +01001281
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012824.2. Alphabetically sorted keywords reference
1283---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001284
1285This section provides a description of each keyword and its usage.
1286
1287
1288acl <aclname> <criterion> [flags] [operator] <value> ...
1289 Declare or complete an access list.
1290 May be used in sections : defaults | frontend | listen | backend
1291 no | yes | yes | yes
1292 Example:
1293 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
1294 acl invalid_src src_port 0:1023
1295 acl local_dst hdr(host) -i localhost
1296
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001297 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001298
1299
Cyril Bontéb21570a2009-11-29 20:04:48 +01001300appsession <cookie> len <length> timeout <holdtime>
1301 [request-learn] [prefix] [mode <path-parameters|query-string>]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001302 Define session stickiness on an existing application cookie.
1303 May be used in sections : defaults | frontend | listen | backend
1304 no | no | yes | yes
1305 Arguments :
1306 <cookie> this is the name of the cookie used by the application and which
1307 HAProxy will have to learn for each new session.
1308
Cyril Bontéb21570a2009-11-29 20:04:48 +01001309 <length> this is the max number of characters that will be memorized and
Willy Tarreau0ba27502007-12-24 16:55:16 +01001310 checked in each cookie value.
1311
1312 <holdtime> this is the time after which the cookie will be removed from
1313 memory if unused. If no unit is specified, this time is in
1314 milliseconds.
1315
Cyril Bontébf47aeb2009-10-15 00:15:40 +02001316 request-learn
1317 If this option is specified, then haproxy will be able to learn
1318 the cookie found in the request in case the server does not
1319 specify any in response. This is typically what happens with
1320 PHPSESSID cookies, or when haproxy's session expires before
1321 the application's session and the correct server is selected.
1322 It is recommended to specify this option to improve reliability.
1323
Cyril Bontéb21570a2009-11-29 20:04:48 +01001324 prefix When this option is specified, haproxy will match on the cookie
1325 prefix (or URL parameter prefix). The appsession value is the
1326 data following this prefix.
1327
1328 Example :
1329 appsession ASPSESSIONID len 64 timeout 3h prefix
1330
1331 This will match the cookie ASPSESSIONIDXXXX=XXXXX,
1332 the appsession value will be XXXX=XXXXX.
1333
1334 mode This option allows to change the URL parser mode.
1335 2 modes are currently supported :
1336 - path-parameters :
1337 The parser looks for the appsession in the path parameters
1338 part (each parameter is separated by a semi-colon), which is
1339 convenient for JSESSIONID for example.
1340 This is the default mode if the option is not set.
1341 - query-string :
1342 In this mode, the parser will look for the appsession in the
1343 query string.
1344
Willy Tarreau0ba27502007-12-24 16:55:16 +01001345 When an application cookie is defined in a backend, HAProxy will check when
1346 the server sets such a cookie, and will store its value in a table, and
1347 associate it with the server's identifier. Up to <length> characters from
1348 the value will be retained. On each connection, haproxy will look for this
Cyril Bontéb21570a2009-11-29 20:04:48 +01001349 cookie both in the "Cookie:" headers, and as a URL parameter (depending on
1350 the mode used). If a known value is found, the client will be directed to the
1351 server associated with this value. Otherwise, the load balancing algorithm is
Willy Tarreau0ba27502007-12-24 16:55:16 +01001352 applied. Cookies are automatically removed from memory when they have been
1353 unused for a duration longer than <holdtime>.
1354
1355 The definition of an application cookie is limited to one per backend.
1356
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01001357 Note : Consider not using this feature in multi-process mode (nbproc > 1)
1358 unless you know what you do : memory is not shared between the
1359 processes, which can result in random behaviours.
1360
Willy Tarreau0ba27502007-12-24 16:55:16 +01001361 Example :
1362 appsession JSESSIONID len 52 timeout 3h
1363
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01001364 See also : "cookie", "capture cookie", "balance", "stick", "stick-table",
1365 "ignore-persist", "nbproc" and "bind-process".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001366
1367
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01001368backlog <conns>
1369 Give hints to the system about the approximate listen backlog desired size
1370 May be used in sections : defaults | frontend | listen | backend
1371 yes | yes | yes | no
1372 Arguments :
1373 <conns> is the number of pending connections. Depending on the operating
1374 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02001375 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01001376
1377 In order to protect against SYN flood attacks, one solution is to increase
1378 the system's SYN backlog size. Depending on the system, sometimes it is just
1379 tunable via a system parameter, sometimes it is not adjustable at all, and
1380 sometimes the system relies on hints given by the application at the time of
1381 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
1382 to the listen() syscall. On systems which can make use of this value, it can
1383 sometimes be useful to be able to specify a different value, hence this
1384 backlog parameter.
1385
1386 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
1387 used as a hint and the system accepts up to the smallest greater power of
1388 two, and never more than some limits (usually 32768).
1389
1390 See also : "maxconn" and the target operating system's tuning guide.
1391
1392
Willy Tarreau0ba27502007-12-24 16:55:16 +01001393balance <algorithm> [ <arguments> ]
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001394balance url_param <param> [check_post [<max_wait>]]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001395 Define the load balancing algorithm to be used in a backend.
1396 May be used in sections : defaults | frontend | listen | backend
1397 yes | no | yes | yes
1398 Arguments :
1399 <algorithm> is the algorithm used to select a server when doing load
1400 balancing. This only applies when no persistence information
1401 is available, or when a connection is redispatched to another
1402 server. <algorithm> may be one of the following :
1403
1404 roundrobin Each server is used in turns, according to their weights.
1405 This is the smoothest and fairest algorithm when the server's
1406 processing time remains equally distributed. This algorithm
1407 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02001408 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08001409 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02001410 large farms, when a server becomes up after having been down
1411 for a very short time, it may sometimes take a few hundreds
1412 requests for it to be re-integrated into the farm and start
1413 receiving traffic. This is normal, though very rare. It is
1414 indicated here in case you would have the chance to observe
1415 it, so that you don't worry.
1416
1417 static-rr Each server is used in turns, according to their weights.
1418 This algorithm is as similar to roundrobin except that it is
1419 static, which means that changing a server's weight on the
1420 fly will have no effect. On the other hand, it has no design
1421 limitation on the number of servers, and when a server goes
1422 up, it is always immediately reintroduced into the farm, once
1423 the full map is recomputed. It also uses slightly less CPU to
1424 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01001425
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01001426 leastconn The server with the lowest number of connections receives the
1427 connection. Round-robin is performed within groups of servers
1428 of the same load to ensure that all servers will be used. Use
1429 of this algorithm is recommended where very long sessions are
1430 expected, such as LDAP, SQL, TSE, etc... but is not very well
1431 suited for protocols using short sessions such as HTTP. This
1432 algorithm is dynamic, which means that server weights may be
1433 adjusted on the fly for slow starts for instance.
1434
Willy Tarreauf09c6602012-02-13 17:12:08 +01001435 first The first server with available connection slots receives the
1436 connection. The servers are choosen from the lowest numeric
1437 identifier to the highest (see server parameter "id"), which
1438 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02001439 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01001440 not make sense to use this algorithm without setting maxconn.
1441 The purpose of this algorithm is to always use the smallest
1442 number of servers so that extra servers can be powered off
1443 during non-intensive hours. This algorithm ignores the server
1444 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02001445 or IMAP than HTTP, though it can be useful there too. In
1446 order to use this algorithm efficiently, it is recommended
1447 that a cloud controller regularly checks server usage to turn
1448 them off when unused, and regularly checks backend queue to
1449 turn new servers on when the queue inflates. Alternatively,
1450 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01001451
Willy Tarreau0ba27502007-12-24 16:55:16 +01001452 source The source IP address is hashed and divided by the total
1453 weight of the running servers to designate which server will
1454 receive the request. This ensures that the same client IP
1455 address will always reach the same server as long as no
1456 server goes down or up. If the hash result changes due to the
1457 number of running servers changing, many clients will be
1458 directed to a different server. This algorithm is generally
1459 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001460 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01001461 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001462 static by default, which means that changing a server's
1463 weight on the fly will have no effect, but this can be
1464 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001465
Oskar Stolc8dc41842012-05-19 10:19:54 +01001466 uri This algorithm hashes either the left part of the URI (before
1467 the question mark) or the whole URI (if the "whole" parameter
1468 is present) and divides the hash value by the total weight of
1469 the running servers. The result designates which server will
1470 receive the request. This ensures that the same URI will
1471 always be directed to the same server as long as no server
1472 goes up or down. This is used with proxy caches and
1473 anti-virus proxies in order to maximize the cache hit rate.
1474 Note that this algorithm may only be used in an HTTP backend.
1475 This algorithm is static by default, which means that
1476 changing a server's weight on the fly will have no effect,
1477 but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001478
Oskar Stolc8dc41842012-05-19 10:19:54 +01001479 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02001480 "depth", both followed by a positive integer number. These
1481 options may be helpful when it is needed to balance servers
1482 based on the beginning of the URI only. The "len" parameter
1483 indicates that the algorithm should only consider that many
1484 characters at the beginning of the URI to compute the hash.
1485 Note that having "len" set to 1 rarely makes sense since most
1486 URIs start with a leading "/".
1487
1488 The "depth" parameter indicates the maximum directory depth
1489 to be used to compute the hash. One level is counted for each
1490 slash in the request. If both parameters are specified, the
1491 evaluation stops when either is reached.
1492
Willy Tarreau0ba27502007-12-24 16:55:16 +01001493 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001494 the query string of each HTTP GET request.
1495
1496 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02001497 request entity will be searched for the parameter argument,
1498 when it is not found in a query string after a question mark
1499 ('?') in the URL. Optionally, specify a number of octets to
1500 wait for before attempting to search the message body. If the
1501 entity can not be searched, then round robin is used for each
1502 request. For instance, if your clients always send the LB
1503 parameter in the first 128 bytes, then specify that. The
1504 default is 48. The entity data will not be scanned until the
1505 required number of octets have arrived at the gateway, this
1506 is the minimum of: (default/max_wait, Content-Length or first
1507 chunk length). If Content-Length is missing or zero, it does
1508 not need to wait for more data than the client promised to
1509 send. When Content-Length is present and larger than
1510 <max_wait>, then waiting is limited to <max_wait> and it is
1511 assumed that this will be enough data to search for the
1512 presence of the parameter. In the unlikely event that
1513 Transfer-Encoding: chunked is used, only the first chunk is
1514 scanned. Parameter values separated by a chunk boundary, may
1515 be randomly balanced if at all.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001516
1517 If the parameter is found followed by an equal sign ('=') and
1518 a value, then the value is hashed and divided by the total
1519 weight of the running servers. The result designates which
1520 server will receive the request.
1521
1522 This is used to track user identifiers in requests and ensure
1523 that a same user ID will always be sent to the same server as
1524 long as no server goes up or down. If no value is found or if
1525 the parameter is not found, then a round robin algorithm is
1526 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001527 backend. This algorithm is static by default, which means
1528 that changing a server's weight on the fly will have no
1529 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001530
Cyril Bontédc4d9032012-04-08 21:57:39 +02001531 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
1532 request. Just as with the equivalent ACL 'hdr()' function,
1533 the header name in parenthesis is not case sensitive. If the
1534 header is absent or if it does not contain any value, the
1535 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01001536
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001537 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01001538 reducing the hash algorithm to the main domain part with some
1539 specific headers such as 'Host'. For instance, in the Host
1540 value "haproxy.1wt.eu", only "1wt" will be considered.
1541
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001542 This algorithm is static by default, which means that
1543 changing a server's weight on the fly will have no effect,
1544 but this can be changed using "hash-type".
1545
Emeric Brun736aa232009-06-30 17:56:00 +02001546 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02001547 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02001548 The RDP cookie <name> (or "mstshash" if omitted) will be
1549 looked up and hashed for each incoming TCP request. Just as
1550 with the equivalent ACL 'req_rdp_cookie()' function, the name
1551 is not case-sensitive. This mechanism is useful as a degraded
1552 persistence mode, as it makes it possible to always send the
1553 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001554 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02001555 used instead.
1556
1557 Note that for this to work, the frontend must ensure that an
1558 RDP cookie is already present in the request buffer. For this
1559 you must use 'tcp-request content accept' rule combined with
1560 a 'req_rdp_cookie_cnt' ACL.
1561
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001562 This algorithm is static by default, which means that
1563 changing a server's weight on the fly will have no effect,
1564 but this can be changed using "hash-type".
1565
Cyril Bontédc4d9032012-04-08 21:57:39 +02001566 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09001567
Willy Tarreau0ba27502007-12-24 16:55:16 +01001568 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02001569 algorithms. Right now, only "url_param" and "uri" support an
1570 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001571
Marek Majkowski9c30fc12008-04-27 23:25:55 +02001572 balance uri [len <len>] [depth <depth>]
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001573 balance url_param <param> [check_post [<max_wait>]]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001574
Willy Tarreau3cd9af22009-03-15 14:06:41 +01001575 The load balancing algorithm of a backend is set to roundrobin when no other
1576 algorithm, mode nor option have been set. The algorithm may only be set once
1577 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001578
1579 Examples :
1580 balance roundrobin
1581 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001582 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01001583 balance hdr(User-Agent)
1584 balance hdr(host)
1585 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001586
1587 Note: the following caveats and limitations on using the "check_post"
1588 extension with "url_param" must be considered :
1589
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001590 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001591 to determine if the parameters will be found in the body or entity which
1592 may contain binary data. Therefore another method may be required to
1593 restrict consideration of POST requests that have no URL parameters in
1594 the body. (see acl reqideny http_end)
1595
1596 - using a <max_wait> value larger than the request buffer size does not
1597 make sense and is useless. The buffer size is set at build time, and
1598 defaults to 16 kB.
1599
1600 - Content-Encoding is not supported, the parameter search will probably
1601 fail; and load balancing will fall back to Round Robin.
1602
1603 - Expect: 100-continue is not supported, load balancing will fall back to
1604 Round Robin.
1605
1606 - Transfer-Encoding (RFC2616 3.6.1) is only supported in the first chunk.
1607 If the entire parameter value is not present in the first chunk, the
1608 selection of server is undefined (actually, defined by how little
1609 actually appeared in the first chunk).
1610
1611 - This feature does not support generation of a 100, 411 or 501 response.
1612
1613 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001614 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001615 white space or control characters are found, indicating the end of what
1616 might be a URL parameter list. This is probably not a concern with SGML
1617 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001618
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001619 See also : "dispatch", "cookie", "appsession", "transparent", "hash-type" and
1620 "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001621
1622
Willy Tarreaub6205fd2012-09-24 12:27:33 +02001623bind [<address>]:<port_range> [, ...] [param*]
1624bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001625 Define one or several listening addresses and/or ports in a frontend.
1626 May be used in sections : defaults | frontend | listen | backend
1627 no | yes | yes | no
1628 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01001629 <address> is optional and can be a host name, an IPv4 address, an IPv6
1630 address, or '*'. It designates the address the frontend will
1631 listen on. If unset, all IPv4 addresses of the system will be
1632 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01001633 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01001634 Optionally, an address family prefix may be used before the
1635 address to force the family regardless of the address format,
1636 which can be useful to specify a path to a unix socket with
1637 no slash ('/'). Currently supported prefixes are :
1638 - 'ipv4@' -> address is always IPv4
1639 - 'ipv6@' -> address is always IPv6
1640 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau40aa0702013-03-10 23:51:38 +01001641 - 'fd@<n>' -> use file descriptor <n> inherited from the
1642 parent. The fd must be bound and may or may not already
1643 be listening.
Willy Tarreaudad36a32013-03-11 01:20:04 +01001644 Any part of the address string may reference any number of
1645 environment variables by preceding their name with a dollar
1646 sign ('$') and optionally enclosing them with braces ('{}'),
1647 similarly to what is done in Bourne shell.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01001648
Willy Tarreauc5011ca2010-03-22 11:53:56 +01001649 <port_range> is either a unique TCP port, or a port range for which the
1650 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001651 above. The port is mandatory for TCP listeners. Note that in
1652 the case of an IPv6 address, the port is always the number
1653 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01001654 - a numerical port (ex: '80')
1655 - a dash-delimited ports range explicitly stating the lower
1656 and upper bounds (ex: '2000-2100') which are included in
1657 the range.
1658
1659 Particular care must be taken against port ranges, because
1660 every <address:port> couple consumes one socket (= a file
1661 descriptor), so it's easy to consume lots of descriptors
1662 with a simple range, and to run out of sockets. Also, each
1663 <address:port> couple must be used only once among all
1664 instances running on a same system. Please note that binding
1665 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001666 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01001667 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001668
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001669 <path> is a UNIX socket path beginning with a slash ('/'). This is
1670 alternative to the TCP listening port. Haproxy will then
1671 receive UNIX connections on the socket located at this place.
1672 The path must begin with a slash and by default is absolute.
1673 It can be relative to the prefix defined by "unix-bind" in
1674 the global section. Note that the total length of the prefix
1675 followed by the socket path cannot exceed some system limits
1676 for UNIX sockets, which commonly are set to 107 characters.
1677
Willy Tarreaub6205fd2012-09-24 12:27:33 +02001678 <param*> is a list of parameters common to all sockets declared on the
1679 same line. These numerous parameters depend on OS and build
1680 options and have a complete section dedicated to them. Please
1681 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02001682
Willy Tarreau0ba27502007-12-24 16:55:16 +01001683 It is possible to specify a list of address:port combinations delimited by
1684 commas. The frontend will then listen on all of these addresses. There is no
1685 fixed limit to the number of addresses and ports which can be listened on in
1686 a frontend, as well as there is no limit to the number of "bind" statements
1687 in a frontend.
1688
1689 Example :
1690 listen http_proxy
1691 bind :80,:443
1692 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001693 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01001694
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02001695 listen http_https_proxy
1696 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02001697 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02001698
Willy Tarreau24709282013-03-10 21:32:12 +01001699 listen http_https_proxy_explicit
1700 bind ipv6@:80
1701 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
1702 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
1703
Willy Tarreaudad36a32013-03-11 01:20:04 +01001704 listen external_bind_app1
1705 bind fd@${FD_APP1}
1706
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001707 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02001708 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001709
1710
Willy Tarreau110ecc12012-11-15 17:50:01 +01001711bind-process [ all | odd | even | <number 1-32>[-<number 1-32>] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001712 Limit visibility of an instance to a certain set of processes numbers.
1713 May be used in sections : defaults | frontend | listen | backend
1714 yes | yes | yes | yes
1715 Arguments :
1716 all All process will see this instance. This is the default. It
1717 may be used to override a default value.
1718
1719 odd This instance will be enabled on processes 1,3,5,...31. This
1720 option may be combined with other numbers.
1721
1722 even This instance will be enabled on processes 2,4,6,...32. This
1723 option may be combined with other numbers. Do not use it
1724 with less than 2 processes otherwise some instances might be
1725 missing from all processes.
1726
Willy Tarreau110ecc12012-11-15 17:50:01 +01001727 number The instance will be enabled on this process number or range,
1728 whose values must all be between 1 and 32. You must be
1729 careful not to reference a process number greater than the
1730 configured global.nbproc, otherwise some instances might be
1731 missing from all processes.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001732
1733 This keyword limits binding of certain instances to certain processes. This
1734 is useful in order not to have too many processes listening to the same
1735 ports. For instance, on a dual-core machine, it might make sense to set
1736 'nbproc 2' in the global section, then distributes the listeners among 'odd'
1737 and 'even' instances.
1738
1739 At the moment, it is not possible to reference more than 32 processes using
1740 this keyword, but this should be more than enough for most setups. Please
1741 note that 'all' really means all processes and is not limited to the first
1742 32.
1743
1744 If some backends are referenced by frontends bound to other processes, the
1745 backend automatically inherits the frontend's processes.
1746
1747 Example :
1748 listen app_ip1
1749 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02001750 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001751
1752 listen app_ip2
1753 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02001754 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001755
1756 listen management
1757 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02001758 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001759
Willy Tarreau110ecc12012-11-15 17:50:01 +01001760 listen management
1761 bind 10.0.0.4:80
1762 bind-process 1-4
1763
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001764 See also : "nbproc" in global section.
1765
1766
Willy Tarreau0ba27502007-12-24 16:55:16 +01001767block { if | unless } <condition>
1768 Block a layer 7 request if/unless a condition is matched
1769 May be used in sections : defaults | frontend | listen | backend
1770 no | yes | yes | yes
1771
1772 The HTTP request will be blocked very early in the layer 7 processing
1773 if/unless <condition> is matched. A 403 error will be returned if the request
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001774 is blocked. The condition has to reference ACLs (see section 7). This is
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02001775 typically used to deny access to certain sensitive resources if some
Willy Tarreau0ba27502007-12-24 16:55:16 +01001776 conditions are met or not met. There is no fixed limit to the number of
1777 "block" statements per instance.
1778
1779 Example:
1780 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
1781 acl invalid_src src_port 0:1023
1782 acl local_dst hdr(host) -i localhost
1783 block if invalid_src || local_dst
1784
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001785 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001786
1787
1788capture cookie <name> len <length>
1789 Capture and log a cookie in the request and in the response.
1790 May be used in sections : defaults | frontend | listen | backend
1791 no | yes | yes | no
1792 Arguments :
1793 <name> is the beginning of the name of the cookie to capture. In order
1794 to match the exact name, simply suffix the name with an equal
1795 sign ('='). The full name will appear in the logs, which is
1796 useful with application servers which adjust both the cookie name
1797 and value (eg: ASPSESSIONXXXXX).
1798
1799 <length> is the maximum number of characters to report in the logs, which
1800 include the cookie name, the equal sign and the value, all in the
1801 standard "name=value" form. The string will be truncated on the
1802 right if it exceeds <length>.
1803
1804 Only the first cookie is captured. Both the "cookie" request headers and the
1805 "set-cookie" response headers are monitored. This is particularly useful to
1806 check for application bugs causing session crossing or stealing between
1807 users, because generally the user's cookies can only change on a login page.
1808
1809 When the cookie was not presented by the client, the associated log column
1810 will report "-". When a request does not cause a cookie to be assigned by the
1811 server, a "-" is reported in the response column.
1812
1813 The capture is performed in the frontend only because it is necessary that
1814 the log format does not change for a given frontend depending on the
1815 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01001816 "capture cookie" statement in a frontend. The maximum capture length is set
1817 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
1818 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001819
1820 Example:
1821 capture cookie ASPSESSION len 32
1822
1823 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001824 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001825
1826
1827capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01001828 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001829 May be used in sections : defaults | frontend | listen | backend
1830 no | yes | yes | no
1831 Arguments :
1832 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01001833 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01001834 appear in the requests, with the first letter of each word in
1835 upper case. The header name will not appear in the logs, only the
1836 value is reported, but the position in the logs is respected.
1837
1838 <length> is the maximum number of characters to extract from the value and
1839 report in the logs. The string will be truncated on the right if
1840 it exceeds <length>.
1841
Willy Tarreau4460d032012-11-21 23:37:37 +01001842 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01001843 value will be added to the logs between braces ('{}'). If multiple headers
1844 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01001845 in the same order they were declared in the configuration. Non-existent
1846 headers will be logged just as an empty string. Common uses for request
1847 header captures include the "Host" field in virtual hosting environments, the
1848 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001849 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01001850 environments to find where the request came from.
1851
1852 Note that when capturing headers such as "User-agent", some spaces may be
1853 logged, making the log analysis more difficult. Thus be careful about what
1854 you log if you know your log parser is not smart enough to rely on the
1855 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001856
Willy Tarreau0900abb2012-11-22 00:21:46 +01001857 There is no limit to the number of captured request headers nor to their
1858 length, though it is wise to keep them low to limit memory usage per session.
1859 In order to keep log format consistent for a same frontend, header captures
1860 can only be declared in a frontend. It is not possible to specify a capture
1861 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001862
1863 Example:
1864 capture request header Host len 15
1865 capture request header X-Forwarded-For len 15
1866 capture request header Referrer len 15
1867
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001868 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01001869 about logging.
1870
1871
1872capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01001873 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001874 May be used in sections : defaults | frontend | listen | backend
1875 no | yes | yes | no
1876 Arguments :
1877 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01001878 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01001879 appear in the response, with the first letter of each word in
1880 upper case. The header name will not appear in the logs, only the
1881 value is reported, but the position in the logs is respected.
1882
1883 <length> is the maximum number of characters to extract from the value and
1884 report in the logs. The string will be truncated on the right if
1885 it exceeds <length>.
1886
Willy Tarreau4460d032012-11-21 23:37:37 +01001887 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01001888 result will be added to the logs between braces ('{}') after the captured
1889 request headers. If multiple headers are captured, they will be delimited by
1890 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01001891 the configuration. Non-existent headers will be logged just as an empty
1892 string. Common uses for response header captures include the "Content-length"
1893 header which indicates how many bytes are expected to be returned, the
1894 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001895
Willy Tarreau0900abb2012-11-22 00:21:46 +01001896 There is no limit to the number of captured response headers nor to their
1897 length, though it is wise to keep them low to limit memory usage per session.
1898 In order to keep log format consistent for a same frontend, header captures
1899 can only be declared in a frontend. It is not possible to specify a capture
1900 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001901
1902 Example:
1903 capture response header Content-length len 9
1904 capture response header Location len 15
1905
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001906 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01001907 about logging.
1908
1909
Cyril Bontéf0c60612010-02-06 14:44:47 +01001910clitimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01001911 Set the maximum inactivity time on the client side.
1912 May be used in sections : defaults | frontend | listen | backend
1913 yes | yes | yes | no
1914 Arguments :
1915 <timeout> is the timeout value is specified in milliseconds by default, but
1916 can be in any other unit if the number is suffixed by the unit,
1917 as explained at the top of this document.
1918
1919 The inactivity timeout applies when the client is expected to acknowledge or
1920 send data. In HTTP mode, this timeout is particularly important to consider
1921 during the first phase, when the client sends the request, and during the
1922 response while it is reading data sent by the server. The value is specified
1923 in milliseconds by default, but can be in any other unit if the number is
1924 suffixed by the unit, as specified at the top of this document. In TCP mode
1925 (and to a lesser extent, in HTTP mode), it is highly recommended that the
1926 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01001927 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01001928 losses by specifying timeouts that are slightly above multiples of 3 seconds
1929 (eg: 4 or 5 seconds).
1930
1931 This parameter is specific to frontends, but can be specified once for all in
1932 "defaults" sections. This is in fact one of the easiest solutions not to
1933 forget about it. An unspecified timeout results in an infinite timeout, which
1934 is not recommended. Such a usage is accepted and works but reports a warning
1935 during startup because it may results in accumulation of expired sessions in
1936 the system if the system's timeouts are not configured either.
1937
1938 This parameter is provided for compatibility but is currently deprecated.
1939 Please use "timeout client" instead.
1940
Willy Tarreau036fae02008-01-06 13:24:40 +01001941 See also : "timeout client", "timeout http-request", "timeout server", and
1942 "srvtimeout".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001943
Cyril Bonté316a8cf2012-11-11 13:38:27 +01001944compression algo <algorithm> ...
1945compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02001946compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02001947 Enable HTTP compression.
1948 May be used in sections : defaults | frontend | listen | backend
1949 yes | yes | yes | yes
1950 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01001951 algo is followed by the list of supported compression algorithms.
1952 type is followed by the list of MIME types that will be compressed.
1953 offload makes haproxy work as a compression offloader only (see notes).
1954
1955 The currently supported algorithms are :
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04001956 identity this is mostly for debugging, and it was useful for developing
Cyril Bonté316a8cf2012-11-11 13:38:27 +01001957 the compression feature. Identity does not apply any change on
1958 data.
1959
1960 gzip applies gzip compression. This setting is only available when
1961 support for zlib was built in.
1962
1963 deflate same as gzip, but with deflate algorithm and zlib format.
1964 Note that this algorithm has ambiguous support on many browsers
1965 and no support at all from recent ones. It is strongly
1966 recommended not to use it for anything else than experimentation.
1967 This setting is only available when support for zlib was built
1968 in.
1969
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04001970 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01001971 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04001972 If backend servers support HTTP compression, these directives
1973 will be no-op: haproxy will see the compressed response and will not
1974 compress again. If backend servers do not support HTTP compression and
1975 there is Accept-Encoding header in request, haproxy will compress the
1976 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02001977
1978 The "offload" setting makes haproxy remove the Accept-Encoding header to
1979 prevent backend servers from compressing responses. It is strongly
1980 recommended not to do this because this means that all the compression work
1981 will be done on the single point where haproxy is located. However in some
1982 deployment scenarios, haproxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04001983 with broken HTTP compression implementation which can't be turned off.
1984 In that case haproxy can be used to prevent that gateway from emitting
1985 invalid payloads. In this case, simply removing the header in the
1986 configuration does not work because it applies before the header is parsed,
1987 so that prevents haproxy from compressing. The "offload" setting should
1988 then be used for such scenarios.
William Lallemand82fe75c2012-10-23 10:25:10 +02001989
William Lallemand05097442012-11-20 12:14:28 +01001990 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01001991 * the request does not advertise a supported compression algorithm in the
1992 "Accept-Encoding" header
1993 * the response message is not HTTP/1.1
William Lallemandd3002612012-11-26 14:34:47 +01001994 * HTTP status code is not 200
William Lallemand8bb4e342013-12-10 17:28:48 +01001995 * response header "Transfer-Encoding" contains "chunked" (Temporary
1996 Workaround)
Baptiste Assmann650d53d2013-01-05 15:44:44 +01001997 * response contain neither a "Content-Length" header nor a
1998 "Transfer-Encoding" whose last value is "chunked"
1999 * response contains a "Content-Type" header whose first value starts with
2000 "multipart"
2001 * the response contains the "no-transform" value in the "Cache-control"
2002 header
2003 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
2004 and later
2005 * The response contains a "Content-Encoding" header, indicating that the
2006 response is already compressed (see compression offload)
William Lallemand05097442012-11-20 12:14:28 +01002007
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002008 Note: The compression does not rewrite Etag headers, and does not emit the
2009 Warning header.
William Lallemand05097442012-11-20 12:14:28 +01002010
William Lallemand82fe75c2012-10-23 10:25:10 +02002011 Examples :
2012 compression algo gzip
2013 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01002014
Cyril Bontéf0c60612010-02-06 14:44:47 +01002015contimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002016 Set the maximum time to wait for a connection attempt to a server to succeed.
2017 May be used in sections : defaults | frontend | listen | backend
2018 yes | no | yes | yes
2019 Arguments :
2020 <timeout> is the timeout value is specified in milliseconds by default, but
2021 can be in any other unit if the number is suffixed by the unit,
2022 as explained at the top of this document.
2023
2024 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002025 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01002026 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01002027 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
2028 connect timeout also presets the queue timeout to the same value if this one
2029 has not been specified. Historically, the contimeout was also used to set the
2030 tarpit timeout in a listen section, which is not possible in a pure frontend.
2031
2032 This parameter is specific to backends, but can be specified once for all in
2033 "defaults" sections. This is in fact one of the easiest solutions not to
2034 forget about it. An unspecified timeout results in an infinite timeout, which
2035 is not recommended. Such a usage is accepted and works but reports a warning
2036 during startup because it may results in accumulation of failed sessions in
2037 the system if the system's timeouts are not configured either.
2038
2039 This parameter is provided for backwards compatibility but is currently
2040 deprecated. Please use "timeout connect", "timeout queue" or "timeout tarpit"
2041 instead.
2042
2043 See also : "timeout connect", "timeout queue", "timeout tarpit",
2044 "timeout server", "contimeout".
2045
2046
Willy Tarreau55165fe2009-05-10 12:02:55 +02002047cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02002048 [ postonly ] [ preserve ] [ httponly ] [ secure ]
2049 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002050 Enable cookie-based persistence in a backend.
2051 May be used in sections : defaults | frontend | listen | backend
2052 yes | no | yes | yes
2053 Arguments :
2054 <name> is the name of the cookie which will be monitored, modified or
2055 inserted in order to bring persistence. This cookie is sent to
2056 the client via a "Set-Cookie" header in the response, and is
2057 brought back by the client in a "Cookie" header in all requests.
2058 Special care should be taken to choose a name which does not
2059 conflict with any likely application cookie. Also, if the same
2060 backends are subject to be used by the same clients (eg:
2061 HTTP/HTTPS), care should be taken to use different cookie names
2062 between all backends if persistence between them is not desired.
2063
2064 rewrite This keyword indicates that the cookie will be provided by the
2065 server and that haproxy will have to modify its value to set the
2066 server's identifier in it. This mode is handy when the management
2067 of complex combinations of "Set-cookie" and "Cache-control"
2068 headers is left to the application. The application can then
2069 decide whether or not it is appropriate to emit a persistence
2070 cookie. Since all responses should be monitored, this mode only
2071 works in HTTP close mode. Unless the application behaviour is
2072 very complex and/or broken, it is advised not to start with this
2073 mode for new deployments. This keyword is incompatible with
2074 "insert" and "prefix".
2075
2076 insert This keyword indicates that the persistence cookie will have to
Willy Tarreaua79094d2010-08-31 22:54:15 +02002077 be inserted by haproxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002078
Willy Tarreaua79094d2010-08-31 22:54:15 +02002079 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002080 server. When used without the "preserve" option, if the server
2081 emits a cookie with the same name, it will be remove before
2082 processing. For this reason, this mode can be used to upgrade
2083 existing configurations running in the "rewrite" mode. The cookie
2084 will only be a session cookie and will not be stored on the
2085 client's disk. By default, unless the "indirect" option is added,
2086 the server will see the cookies emitted by the client. Due to
2087 caching effects, it is generally wise to add the "nocache" or
2088 "postonly" keywords (see below). The "insert" keyword is not
2089 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002090
2091 prefix This keyword indicates that instead of relying on a dedicated
2092 cookie for the persistence, an existing one will be completed.
2093 This may be needed in some specific environments where the client
2094 does not support more than one single cookie and the application
2095 already needs it. In this case, whenever the server sets a cookie
2096 named <name>, it will be prefixed with the server's identifier
2097 and a delimiter. The prefix will be removed from all client
2098 requests so that the server still finds the cookie it emitted.
2099 Since all requests and responses are subject to being modified,
2100 this mode requires the HTTP close mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02002101 not compatible with "rewrite" and "insert". Note: it is highly
2102 recommended not to use "indirect" with "prefix", otherwise server
2103 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002104
Willy Tarreaua79094d2010-08-31 22:54:15 +02002105 indirect When this option is specified, no cookie will be emitted to a
2106 client which already has a valid one for the server which has
2107 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002108 it will be removed, unless the "preserve" option is also set. In
2109 "insert" mode, this will additionally remove cookies from the
2110 requests transmitted to the server, making the persistence
2111 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02002112 Note: it is highly recommended not to use "indirect" with
2113 "prefix", otherwise server cookie updates would not be sent to
2114 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002115
2116 nocache This option is recommended in conjunction with the insert mode
2117 when there is a cache between the client and HAProxy, as it
2118 ensures that a cacheable response will be tagged non-cacheable if
2119 a cookie needs to be inserted. This is important because if all
2120 persistence cookies are added on a cacheable home page for
2121 instance, then all customers will then fetch the page from an
2122 outer cache and will all share the same persistence cookie,
2123 leading to one server receiving much more traffic than others.
2124 See also the "insert" and "postonly" options.
2125
2126 postonly This option ensures that cookie insertion will only be performed
2127 on responses to POST requests. It is an alternative to the
2128 "nocache" option, because POST responses are not cacheable, so
2129 this ensures that the persistence cookie will never get cached.
2130 Since most sites do not need any sort of persistence before the
2131 first POST which generally is a login request, this is a very
2132 efficient method to optimize caching without risking to find a
2133 persistence cookie in the cache.
2134 See also the "insert" and "nocache" options.
2135
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002136 preserve This option may only be used with "insert" and/or "indirect". It
2137 allows the server to emit the persistence cookie itself. In this
2138 case, if a cookie is found in the response, haproxy will leave it
2139 untouched. This is useful in order to end persistence after a
2140 logout request for instance. For this, the server just has to
2141 emit a cookie with an invalid value (eg: empty) or with a date in
2142 the past. By combining this mechanism with the "disable-on-404"
2143 check option, it is possible to perform a completely graceful
2144 shutdown because users will definitely leave the server after
2145 they logout.
2146
Willy Tarreau4992dd22012-05-31 21:02:17 +02002147 httponly This option tells haproxy to add an "HttpOnly" cookie attribute
2148 when a cookie is inserted. This attribute is used so that a
2149 user agent doesn't share the cookie with non-HTTP components.
2150 Please check RFC6265 for more information on this attribute.
2151
2152 secure This option tells haproxy to add a "Secure" cookie attribute when
2153 a cookie is inserted. This attribute is used so that a user agent
2154 never emits this cookie over non-secure channels, which means
2155 that a cookie learned with this flag will be presented only over
2156 SSL/TLS connections. Please check RFC6265 for more information on
2157 this attribute.
2158
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002159 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002160 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01002161 name. If the domain begins with a dot, the browser is allowed to
2162 use it for any host ending with that name. It is also possible to
2163 specify several domain names by invoking this option multiple
2164 times. Some browsers might have small limits on the number of
2165 domains, so be careful when doing that. For the record, sending
2166 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002167
Willy Tarreau996a92c2010-10-13 19:30:47 +02002168 maxidle This option allows inserted cookies to be ignored after some idle
2169 time. It only works with insert-mode cookies. When a cookie is
2170 sent to the client, the date this cookie was emitted is sent too.
2171 Upon further presentations of this cookie, if the date is older
2172 than the delay indicated by the parameter (in seconds), it will
2173 be ignored. Otherwise, it will be refreshed if needed when the
2174 response is sent to the client. This is particularly useful to
2175 prevent users who never close their browsers from remaining for
2176 too long on the same server (eg: after a farm size change). When
2177 this option is set and a cookie has no date, it is always
2178 accepted, but gets refreshed in the response. This maintains the
2179 ability for admins to access their sites. Cookies that have a
2180 date in the future further than 24 hours are ignored. Doing so
2181 lets admins fix timezone issues without risking kicking users off
2182 the site.
2183
2184 maxlife This option allows inserted cookies to be ignored after some life
2185 time, whether they're in use or not. It only works with insert
2186 mode cookies. When a cookie is first sent to the client, the date
2187 this cookie was emitted is sent too. Upon further presentations
2188 of this cookie, if the date is older than the delay indicated by
2189 the parameter (in seconds), it will be ignored. If the cookie in
2190 the request has no date, it is accepted and a date will be set.
2191 Cookies that have a date in the future further than 24 hours are
2192 ignored. Doing so lets admins fix timezone issues without risking
2193 kicking users off the site. Contrary to maxidle, this value is
2194 not refreshed, only the first visit date counts. Both maxidle and
2195 maxlife may be used at the time. This is particularly useful to
2196 prevent users who never close their browsers from remaining for
2197 too long on the same server (eg: after a farm size change). This
2198 is stronger than the maxidle method in that it forces a
2199 redispatch after some absolute delay.
2200
Willy Tarreau0ba27502007-12-24 16:55:16 +01002201 There can be only one persistence cookie per HTTP backend, and it can be
2202 declared in a defaults section. The value of the cookie will be the value
2203 indicated after the "cookie" keyword in a "server" statement. If no cookie
2204 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002205
Willy Tarreau0ba27502007-12-24 16:55:16 +01002206 Examples :
2207 cookie JSESSIONID prefix
2208 cookie SRV insert indirect nocache
2209 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02002210 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01002211
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02002212 See also : "appsession", "balance source", "capture cookie", "server"
Cyril Bonté0d4bf012010-04-25 23:21:46 +02002213 and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002214
Willy Tarreau983e01e2010-01-11 18:42:06 +01002215
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002216default-server [param*]
2217 Change default options for a server in a backend
2218 May be used in sections : defaults | frontend | listen | backend
2219 yes | no | yes | yes
2220 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01002221 <param*> is a list of parameters for this server. The "default-server"
2222 keyword accepts an important number of options and has a complete
2223 section dedicated to it. Please refer to section 5 for more
2224 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002225
Willy Tarreau983e01e2010-01-11 18:42:06 +01002226 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002227 default-server inter 1000 weight 13
2228
2229 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01002230
Willy Tarreau983e01e2010-01-11 18:42:06 +01002231
Willy Tarreau0ba27502007-12-24 16:55:16 +01002232default_backend <backend>
2233 Specify the backend to use when no "use_backend" rule has been matched.
2234 May be used in sections : defaults | frontend | listen | backend
2235 yes | yes | yes | no
2236 Arguments :
2237 <backend> is the name of the backend to use.
2238
2239 When doing content-switching between frontend and backends using the
2240 "use_backend" keyword, it is often useful to indicate which backend will be
2241 used when no rule has matched. It generally is the dynamic backend which
2242 will catch all undetermined requests.
2243
Willy Tarreau0ba27502007-12-24 16:55:16 +01002244 Example :
2245
2246 use_backend dynamic if url_dyn
2247 use_backend static if url_css url_img extension_img
2248 default_backend dynamic
2249
Willy Tarreau2769aa02007-12-27 18:26:09 +01002250 See also : "use_backend", "reqsetbe", "reqisetbe"
2251
Willy Tarreau0ba27502007-12-24 16:55:16 +01002252
Baptiste Assmann27f51342013-10-09 06:51:49 +02002253description <string>
2254 Describe a listen, frontend or backend.
2255 May be used in sections : defaults | frontend | listen | backend
2256 no | yes | yes | yes
2257 Arguments : string
2258
2259 Allows to add a sentence to describe the related object in the HAProxy HTML
2260 stats page. The description will be printed on the right of the object name
2261 it describes.
2262 No need to backslash spaces in the <string> arguments.
2263
2264
Willy Tarreau0ba27502007-12-24 16:55:16 +01002265disabled
2266 Disable a proxy, frontend or backend.
2267 May be used in sections : defaults | frontend | listen | backend
2268 yes | yes | yes | yes
2269 Arguments : none
2270
2271 The "disabled" keyword is used to disable an instance, mainly in order to
2272 liberate a listening port or to temporarily disable a service. The instance
2273 will still be created and its configuration will be checked, but it will be
2274 created in the "stopped" state and will appear as such in the statistics. It
2275 will not receive any traffic nor will it send any health-checks or logs. It
2276 is possible to disable many instances at once by adding the "disabled"
2277 keyword in a "defaults" section.
2278
2279 See also : "enabled"
2280
2281
Willy Tarreau5ce94572010-06-07 14:35:41 +02002282dispatch <address>:<port>
2283 Set a default server address
2284 May be used in sections : defaults | frontend | listen | backend
2285 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02002286 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02002287
2288 <address> is the IPv4 address of the default server. Alternatively, a
2289 resolvable hostname is supported, but this name will be resolved
2290 during start-up.
2291
2292 <ports> is a mandatory port specification. All connections will be sent
2293 to this port, and it is not permitted to use port offsets as is
2294 possible with normal servers.
2295
Willy Tarreau787aed52011-04-15 06:45:37 +02002296 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02002297 server can take the connection. In the past it was used to forward non
2298 persistent connections to an auxiliary load balancer. Due to its simple
2299 syntax, it has also been used for simple TCP relays. It is recommended not to
2300 use it for more clarity, and to use the "server" directive instead.
2301
2302 See also : "server"
2303
2304
Willy Tarreau0ba27502007-12-24 16:55:16 +01002305enabled
2306 Enable a proxy, frontend or backend.
2307 May be used in sections : defaults | frontend | listen | backend
2308 yes | yes | yes | yes
2309 Arguments : none
2310
2311 The "enabled" keyword is used to explicitly enable an instance, when the
2312 defaults has been set to "disabled". This is very rarely used.
2313
2314 See also : "disabled"
2315
2316
2317errorfile <code> <file>
2318 Return a file contents instead of errors generated by HAProxy
2319 May be used in sections : defaults | frontend | listen | backend
2320 yes | yes | yes | yes
2321 Arguments :
2322 <code> is the HTTP status code. Currently, HAProxy is capable of
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002323 generating codes 200, 400, 403, 408, 500, 502, 503, and 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002324
2325 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002326 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01002327 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01002328 error pages, and to use absolute paths, since files are read
2329 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002330
2331 It is important to understand that this keyword is not meant to rewrite
2332 errors returned by the server, but errors detected and returned by HAProxy.
2333 This is why the list of supported errors is limited to a small set.
2334
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002335 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2336
Willy Tarreau0ba27502007-12-24 16:55:16 +01002337 The files are returned verbatim on the TCP socket. This allows any trick such
2338 as redirections to another URL or site, as well as tricks to clean cookies,
2339 force enable or disable caching, etc... The package provides default error
2340 files returning the same contents as default errors.
2341
Willy Tarreau59140a22009-02-22 12:02:30 +01002342 The files should not exceed the configured buffer size (BUFSIZE), which
2343 generally is 8 or 16 kB, otherwise they will be truncated. It is also wise
2344 not to put any reference to local contents (eg: images) in order to avoid
2345 loops between the client and HAProxy when all servers are down, causing an
2346 error to be returned instead of an image. For better HTTP compliance, it is
2347 recommended that all header lines end with CR-LF and not LF alone.
2348
Willy Tarreau0ba27502007-12-24 16:55:16 +01002349 The files are read at the same time as the configuration and kept in memory.
2350 For this reason, the errors continue to be returned even when the process is
2351 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01002352 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01002353 403 status code and interrogating a blocked URL.
2354
2355 See also : "errorloc", "errorloc302", "errorloc303"
2356
Willy Tarreau59140a22009-02-22 12:02:30 +01002357 Example :
2358 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
2359 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
2360 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
2361
Willy Tarreau2769aa02007-12-27 18:26:09 +01002362
2363errorloc <code> <url>
2364errorloc302 <code> <url>
2365 Return an HTTP redirection to a URL instead of errors generated by HAProxy
2366 May be used in sections : defaults | frontend | listen | backend
2367 yes | yes | yes | yes
2368 Arguments :
2369 <code> is the HTTP status code. Currently, HAProxy is capable of
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002370 generating codes 200, 400, 403, 408, 500, 502, 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002371
2372 <url> it is the exact contents of the "Location" header. It may contain
2373 either a relative URI to an error page hosted on the same site,
2374 or an absolute URI designating an error page on another site.
2375 Special care should be given to relative URIs to avoid redirect
2376 loops if the URI itself may generate the same error (eg: 500).
2377
2378 It is important to understand that this keyword is not meant to rewrite
2379 errors returned by the server, but errors detected and returned by HAProxy.
2380 This is why the list of supported errors is limited to a small set.
2381
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002382 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2383
Willy Tarreau2769aa02007-12-27 18:26:09 +01002384 Note that both keyword return the HTTP 302 status code, which tells the
2385 client to fetch the designated URL using the same HTTP method. This can be
2386 quite problematic in case of non-GET methods such as POST, because the URL
2387 sent to the client might not be allowed for something other than GET. To
2388 workaround this problem, please use "errorloc303" which send the HTTP 303
2389 status code, indicating to the client that the URL must be fetched with a GET
2390 request.
2391
2392 See also : "errorfile", "errorloc303"
2393
2394
2395errorloc303 <code> <url>
2396 Return an HTTP redirection to a URL instead of errors generated by HAProxy
2397 May be used in sections : defaults | frontend | listen | backend
2398 yes | yes | yes | yes
2399 Arguments :
2400 <code> is the HTTP status code. Currently, HAProxy is capable of
2401 generating codes 400, 403, 408, 500, 502, 503, and 504.
2402
2403 <url> it is the exact contents of the "Location" header. It may contain
2404 either a relative URI to an error page hosted on the same site,
2405 or an absolute URI designating an error page on another site.
2406 Special care should be given to relative URIs to avoid redirect
2407 loops if the URI itself may generate the same error (eg: 500).
2408
2409 It is important to understand that this keyword is not meant to rewrite
2410 errors returned by the server, but errors detected and returned by HAProxy.
2411 This is why the list of supported errors is limited to a small set.
2412
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002413 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2414
Willy Tarreau2769aa02007-12-27 18:26:09 +01002415 Note that both keyword return the HTTP 303 status code, which tells the
2416 client to fetch the designated URL using the same HTTP GET method. This
2417 solves the usual problems associated with "errorloc" and the 302 code. It is
2418 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002419 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002420
2421 See also : "errorfile", "errorloc", "errorloc302"
2422
2423
Willy Tarreau4de91492010-01-22 19:10:05 +01002424force-persist { if | unless } <condition>
2425 Declare a condition to force persistence on down servers
2426 May be used in sections: defaults | frontend | listen | backend
2427 no | yes | yes | yes
2428
2429 By default, requests are not dispatched to down servers. It is possible to
2430 force this using "option persist", but it is unconditional and redispatches
2431 to a valid server if "option redispatch" is set. That leaves with very little
2432 possibilities to force some requests to reach a server which is artificially
2433 marked down for maintenance operations.
2434
2435 The "force-persist" statement allows one to declare various ACL-based
2436 conditions which, when met, will cause a request to ignore the down status of
2437 a server and still try to connect to it. That makes it possible to start a
2438 server, still replying an error to the health checks, and run a specially
2439 configured browser to test the service. Among the handy methods, one could
2440 use a specific source IP address, or a specific cookie. The cookie also has
2441 the advantage that it can easily be added/removed on the browser from a test
2442 page. Once the service is validated, it is then possible to open the service
2443 to the world by returning a valid response to health checks.
2444
2445 The forced persistence is enabled when an "if" condition is met, or unless an
2446 "unless" condition is met. The final redispatch is always disabled when this
2447 is used.
2448
Cyril Bonté0d4bf012010-04-25 23:21:46 +02002449 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02002450 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01002451
2452
Willy Tarreau2769aa02007-12-27 18:26:09 +01002453fullconn <conns>
2454 Specify at what backend load the servers will reach their maxconn
2455 May be used in sections : defaults | frontend | listen | backend
2456 yes | no | yes | yes
2457 Arguments :
2458 <conns> is the number of connections on the backend which will make the
2459 servers use the maximal number of connections.
2460
Willy Tarreau198a7442008-01-17 12:05:32 +01002461 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01002462 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01002463 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01002464 load. The server will then always accept at least <minconn> connections,
2465 never more than <maxconn>, and the limit will be on the ramp between both
2466 values when the backend has less than <conns> concurrent connections. This
2467 makes it possible to limit the load on the servers during normal loads, but
2468 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002469 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002470
Willy Tarreaufbb78422011-06-05 15:38:35 +02002471 Since it's hard to get this value right, haproxy automatically sets it to
2472 10% of the sum of the maxconns of all frontends that may branch to this
2473 backend. That way it's safe to leave it unset.
2474
Willy Tarreau2769aa02007-12-27 18:26:09 +01002475 Example :
2476 # The servers will accept between 100 and 1000 concurrent connections each
2477 # and the maximum of 1000 will be reached when the backend reaches 10000
2478 # connections.
2479 backend dynamic
2480 fullconn 10000
2481 server srv1 dyn1:80 minconn 100 maxconn 1000
2482 server srv2 dyn2:80 minconn 100 maxconn 1000
2483
2484 See also : "maxconn", "server"
2485
2486
2487grace <time>
2488 Maintain a proxy operational for some time after a soft stop
2489 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01002490 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01002491 Arguments :
2492 <time> is the time (by default in milliseconds) for which the instance
2493 will remain operational with the frontend sockets still listening
2494 when a soft-stop is received via the SIGUSR1 signal.
2495
2496 This may be used to ensure that the services disappear in a certain order.
2497 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002498 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01002499 needed by the equipment to detect the failure.
2500
2501 Note that currently, there is very little benefit in using this parameter,
2502 and it may in fact complicate the soft-reconfiguration process more than
2503 simplify it.
2504
Willy Tarreau0ba27502007-12-24 16:55:16 +01002505
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05002506hash-type <method> <function> <modifier>
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002507 Specify a method to use for mapping hashes to servers
2508 May be used in sections : defaults | frontend | listen | backend
2509 yes | no | yes | yes
2510 Arguments :
Bhaskar98634f02013-10-29 23:30:51 -04002511 <method> is the method used to select a server from the hash computed by
2512 the <function> :
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002513
Bhaskar98634f02013-10-29 23:30:51 -04002514 map-based the hash table is a static array containing all alive servers.
2515 The hashes will be very smooth, will consider weights, but
2516 will be static in that weight changes while a server is up
2517 will be ignored. This means that there will be no slow start.
2518 Also, since a server is selected by its position in the array,
2519 most mappings are changed when the server count changes. This
2520 means that when a server goes up or down, or when a server is
2521 added to a farm, most connections will be redistributed to
2522 different servers. This can be inconvenient with caches for
2523 instance.
Willy Tarreau798a39c2010-11-24 15:04:29 +01002524
Bhaskar98634f02013-10-29 23:30:51 -04002525 consistent the hash table is a tree filled with many occurrences of each
2526 server. The hash key is looked up in the tree and the closest
2527 server is chosen. This hash is dynamic, it supports changing
2528 weights while the servers are up, so it is compatible with the
2529 slow start feature. It has the advantage that when a server
2530 goes up or down, only its associations are moved. When a
2531 server is added to the farm, only a few part of the mappings
2532 are redistributed, making it an ideal method for caches.
2533 However, due to its principle, the distribution will never be
2534 very smooth and it may sometimes be necessary to adjust a
2535 server's weight or its ID to get a more balanced distribution.
2536 In order to get the same distribution on multiple load
2537 balancers, it is important that all servers have the exact
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05002538 same IDs. Note: consistent hash uses sdbm and avalanche if no
2539 hash function is specified.
Bhaskar98634f02013-10-29 23:30:51 -04002540
2541 <function> is the hash function to be used :
2542
2543 sdbm this function was created intially for sdbm (a public-domain
2544 reimplementation of ndbm) database library. It was found to do
2545 well in scrambling bits, causing better distribution of the keys
2546 and fewer splits. It also happens to be a good general hashing
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05002547 function with good distribution, unless the total server weight
2548 is a multiple of 64, in which case applying the avalanche
2549 modifier may help.
Bhaskar98634f02013-10-29 23:30:51 -04002550
2551 djb2 this function was first proposed by Dan Bernstein many years ago
2552 on comp.lang.c. Studies have shown that for certain workload this
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05002553 function provides a better distribution than sdbm. It generally
2554 works well with text-based inputs though it can perform extremely
2555 poorly with numeric-only input or when the total server weight is
2556 a multiple of 33, unless the avalanche modifier is also used.
2557
Willy Tarreaua0f42712013-11-14 14:30:35 +01002558 wt6 this function was designed for haproxy while testing other
2559 functions in the past. It is not as smooth as the other ones, but
2560 is much less sensible to the input data set or to the number of
2561 servers. It can make sense as an alternative to sdbm+avalanche or
2562 djb2+avalanche for consistent hashing or when hashing on numeric
2563 data such as a source IP address or a visitor identifier in a URL
2564 parameter.
2565
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05002566 <modifier> indicates an optional method applied after hashing the key :
2567
2568 avalanche This directive indicates that the result from the hash
2569 function above should not be used in its raw form but that
2570 a 4-byte full avalanche hash must be applied first. The
2571 purpose of this step is to mix the resulting bits from the
2572 previous hash in order to avoid any undesired effect when
2573 the input contains some limited values or when the number of
2574 servers is a multiple of one of the hash's components (64
2575 for SDBM, 33 for DJB2). Enabling avalanche tends to make the
2576 result less predictable, but it's also not as smooth as when
2577 using the original function. Some testing might be needed
2578 with some workloads. This hash is one of the many proposed
2579 by Bob Jenkins.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002580
Bhaskar98634f02013-10-29 23:30:51 -04002581 The default hash type is "map-based" and is recommended for most usages. The
2582 default function is "sdbm", the selection of a function should be based on
2583 the range of the values being hashed.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002584
2585 See also : "balance", "server"
2586
2587
Willy Tarreau0ba27502007-12-24 16:55:16 +01002588http-check disable-on-404
2589 Enable a maintenance mode upon HTTP/404 response to health-checks
2590 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01002591 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01002592 Arguments : none
2593
2594 When this option is set, a server which returns an HTTP code 404 will be
2595 excluded from further load-balancing, but will still receive persistent
2596 connections. This provides a very convenient method for Web administrators
2597 to perform a graceful shutdown of their servers. It is also important to note
2598 that a server which is detected as failed while it was in this mode will not
2599 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
2600 will immediately be reinserted into the farm. The status on the stats page
2601 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01002602 option only works in conjunction with the "httpchk" option. If this option
2603 is used with "http-check expect", then it has precedence over it so that 404
2604 responses will still be considered as soft-stop.
2605
2606 See also : "option httpchk", "http-check expect"
2607
2608
2609http-check expect [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002610 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01002611 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02002612 yes | no | yes | yes
Willy Tarreaubd741542010-03-16 18:46:54 +01002613 Arguments :
2614 <match> is a keyword indicating how to look for a specific pattern in the
2615 response. The keyword may be one of "status", "rstatus",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002616 "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01002617 exclamation mark ("!") to negate the match. Spaces are allowed
2618 between the exclamation mark and the keyword. See below for more
2619 details on the supported keywords.
2620
2621 <pattern> is the pattern to look for. It may be a string or a regular
2622 expression. If the pattern contains spaces, they must be escaped
2623 with the usual backslash ('\').
2624
2625 By default, "option httpchk" considers that response statuses 2xx and 3xx
2626 are valid, and that others are invalid. When "http-check expect" is used,
2627 it defines what is considered valid or invalid. Only one "http-check"
2628 statement is supported in a backend. If a server fails to respond or times
2629 out, the check obviously fails. The available matches are :
2630
2631 status <string> : test the exact string match for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002632 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01002633 response's status code is exactly this string. If the
2634 "status" keyword is prefixed with "!", then the response
2635 will be considered invalid if the status code matches.
2636
2637 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002638 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01002639 response's status code matches the expression. If the
2640 "rstatus" keyword is prefixed with "!", then the response
2641 will be considered invalid if the status code matches.
2642 This is mostly used to check for multiple codes.
2643
2644 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002645 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01002646 response's body contains this exact string. If the
2647 "string" keyword is prefixed with "!", then the response
2648 will be considered invalid if the body contains this
2649 string. This can be used to look for a mandatory word at
2650 the end of a dynamic page, or to detect a failure when a
2651 specific error appears on the check page (eg: a stack
2652 trace).
2653
2654 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002655 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01002656 response's body matches this expression. If the "rstring"
2657 keyword is prefixed with "!", then the response will be
2658 considered invalid if the body matches the expression.
2659 This can be used to look for a mandatory word at the end
2660 of a dynamic page, or to detect a failure when a specific
2661 error appears on the check page (eg: a stack trace).
2662
2663 It is important to note that the responses will be limited to a certain size
2664 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
2665 Thus, too large responses may not contain the mandatory pattern when using
2666 "string" or "rstring". If a large response is absolutely required, it is
2667 possible to change the default max size by setting the global variable.
2668 However, it is worth keeping in mind that parsing very large responses can
2669 waste some CPU cycles, especially when regular expressions are used, and that
2670 it is always better to focus the checks on smaller resources.
2671
2672 Last, if "http-check expect" is combined with "http-check disable-on-404",
2673 then this last one has precedence when the server responds with 404.
2674
2675 Examples :
2676 # only accept status 200 as valid
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01002677 http-check expect status 200
Willy Tarreaubd741542010-03-16 18:46:54 +01002678
2679 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01002680 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01002681
2682 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01002683 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01002684
2685 # check that we have a correct hexadecimal tag before /html
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01002686 http-check expect rstring <!--tag:[0-9a-f]*</html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01002687
Willy Tarreaubd741542010-03-16 18:46:54 +01002688 See also : "option httpchk", "http-check disable-on-404"
Willy Tarreau2769aa02007-12-27 18:26:09 +01002689
2690
Willy Tarreauef781042010-01-27 11:53:01 +01002691http-check send-state
2692 Enable emission of a state header with HTTP health checks
2693 May be used in sections : defaults | frontend | listen | backend
2694 yes | no | yes | yes
2695 Arguments : none
2696
2697 When this option is set, haproxy will systematically send a special header
2698 "X-Haproxy-Server-State" with a list of parameters indicating to each server
2699 how they are seen by haproxy. This can be used for instance when a server is
2700 manipulated without access to haproxy and the operator needs to know whether
2701 haproxy still sees it up or not, or if the server is the last one in a farm.
2702
2703 The header is composed of fields delimited by semi-colons, the first of which
2704 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
2705 checks on the total number before transition, just as appears in the stats
2706 interface. Next headers are in the form "<variable>=<value>", indicating in
2707 no specific order some values available in the stats interface :
2708 - a variable "name", containing the name of the backend followed by a slash
2709 ("/") then the name of the server. This can be used when a server is
2710 checked in multiple backends.
2711
2712 - a variable "node" containing the name of the haproxy node, as set in the
2713 global "node" variable, otherwise the system's hostname if unspecified.
2714
2715 - a variable "weight" indicating the weight of the server, a slash ("/")
2716 and the total weight of the farm (just counting usable servers). This
2717 helps to know if other servers are available to handle the load when this
2718 one fails.
2719
2720 - a variable "scur" indicating the current number of concurrent connections
2721 on the server, followed by a slash ("/") then the total number of
2722 connections on all servers of the same backend.
2723
2724 - a variable "qcur" indicating the current number of requests in the
2725 server's queue.
2726
2727 Example of a header received by the application server :
2728 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
2729 scur=13/22; qcur=0
2730
2731 See also : "option httpchk", "http-check disable-on-404"
2732
Willy Tarreauccbcc372012-12-27 12:37:57 +01002733http-request { allow | deny | tarpit | auth [realm <realm>] | redirect <rule> |
Willy Tarreauf4c43c12013-06-11 17:01:13 +02002734 add-header <name> <fmt> | set-header <name> <fmt> |
Willy Tarreau51347ed2013-06-11 19:34:13 +02002735 set-nice <nice> | set-log-level <level> | set-tos <tos> |
2736 set-mark <mark> }
Cyril Bontéf0c60612010-02-06 14:44:47 +01002737 [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002738 Access control for Layer 7 requests
2739
2740 May be used in sections: defaults | frontend | listen | backend
2741 no | yes | yes | yes
2742
Willy Tarreau20b0de52012-12-24 15:45:22 +01002743 The http-request statement defines a set of rules which apply to layer 7
2744 processing. The rules are evaluated in their declaration order when they are
2745 met in a frontend, listen or backend section. Any rule may optionally be
2746 followed by an ACL-based condition, in which case it will only be evaluated
2747 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002748
Willy Tarreau20b0de52012-12-24 15:45:22 +01002749 The first keyword is the rule's action. Currently supported actions include :
2750 - "allow" : this stops the evaluation of the rules and lets the request
2751 pass the check. No further "http-request" rules are evaluated.
2752
2753 - "deny" : this stops the evaluation of the rules and immediately rejects
2754 the request and emits an HTTP 403 error. No further "http-request" rules
2755 are evaluated.
2756
Willy Tarreauccbcc372012-12-27 12:37:57 +01002757 - "tarpit" : this stops the evaluation of the rules and immediately blocks
2758 the request without responding for a delay specified by "timeout tarpit"
2759 or "timeout connect" if the former is not set. After that delay, if the
2760 client is still connected, an HTTP error 500 is returned so that the
2761 client does not suspect it has been tarpitted. Logs will report the flags
2762 "PT". The goal of the tarpit rule is to slow down robots during an attack
2763 when they're limited on the number of concurrent requests. It can be very
2764 efficient against very dumb robots, and will significantly reduce the
2765 load on firewalls compared to a "deny" rule. But when facing "correctly"
2766 developped robots, it can make things worse by forcing haproxy and the
2767 front firewall to support insane number of concurrent connections.
2768
Willy Tarreau20b0de52012-12-24 15:45:22 +01002769 - "auth" : this stops the evaluation of the rules and immediately responds
2770 with an HTTP 401 or 407 error code to invite the user to present a valid
2771 user name and password. No further "http-request" rules are evaluated. An
2772 optional "realm" parameter is supported, it sets the authentication realm
2773 that is returned with the response (typically the application's name).
2774
Willy Tarreau81499eb2012-12-27 12:19:02 +01002775 - "redirect" : this performs an HTTP redirection based on a redirect rule.
2776 This is exactly the same as the "redirect" statement except that it
2777 inserts a redirect rule which can be processed in the middle of other
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01002778 "http-request" rules and that these rules use the "log-format" strings.
2779 See the "redirect" keyword for the rule's syntax.
Willy Tarreau81499eb2012-12-27 12:19:02 +01002780
Willy Tarreau20b0de52012-12-24 15:45:22 +01002781 - "add-header" appends an HTTP header field whose name is specified in
2782 <name> and whose value is defined by <fmt> which follows the log-format
2783 rules (see Custom Log Format in section 8.2.4). This is particularly
2784 useful to pass connection-specific information to the server (eg: the
2785 client's SSL certificate), or to combine several headers into one. This
2786 rule is not final, so it is possible to add other similar rules. Note
2787 that header addition is performed immediately, so one rule might reuse
2788 the resulting header from a previous rule.
2789
2790 - "set-header" does the same as "add-header" except that the header name
2791 is first removed if it existed. This is useful when passing security
2792 information to the server, where the header must not be manipulated by
2793 external users.
2794
Willy Tarreauf4c43c12013-06-11 17:01:13 +02002795 - "set-nice" sets the "nice" factor of the current request being processed.
2796 It only has effect against the other requests being processed at the same
2797 time. The default value is 0, unless altered by the "nice" setting on the
2798 "bind" line. The accepted range is -1024..1024. The higher the value, the
2799 nicest the request will be. Lower values will make the request more
2800 important than other ones. This can be useful to improve the speed of
2801 some requests, or lower the priority of non-important requests. Using
2802 this setting without prior experimentation can cause some major slowdown.
2803
Willy Tarreau9a355ec2013-06-11 17:45:46 +02002804 - "set-log-level" is used to change the log level of the current request
2805 when a certain condition is met. Valid levels are the 8 syslog levels
2806 (see the "log" keyword) plus the special level "silent" which disables
2807 logging for this request. This rule is not final so the last matching
2808 rule wins. This rule can be useful to disable health checks coming from
2809 another equipment.
2810
Willy Tarreau42cf39e2013-06-11 18:51:32 +02002811 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
2812 the client to the value passed in <tos> on platforms which support this.
2813 This value represents the whole 8 bits of the IP TOS field, and can be
2814 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
2815 that only the 6 higher bits are used in DSCP or TOS, and the two lower
2816 bits are always 0. This can be used to adjust some routing behaviour on
2817 border routers based on some information from the request. See RFC 2474,
2818 2597, 3260 and 4594 for more information.
2819
Willy Tarreau51347ed2013-06-11 19:34:13 +02002820 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
2821 client to the value passed in <mark> on platforms which support it. This
2822 value is an unsigned 32 bit value which can be matched by netfilter and
2823 by the routing table. It can be expressed both in decimal or hexadecimal
2824 format (prefixed by "0x"). This can be useful to force certain packets to
2825 take a different route (for example a cheaper network path for bulk
2826 downloads). This works on Linux kernels 2.6.32 and above and requires
2827 admin privileges.
2828
Willy Tarreau20b0de52012-12-24 15:45:22 +01002829 There is no limit to the number of http-request statements per instance.
2830
2831 It is important to know that http-request rules are processed very early in
2832 the HTTP processing, just after "block" rules and before "reqdel" or "reqrep"
2833 rules. That way, headers added by "add-header"/"set-header" are visible by
2834 almost all further ACL rules.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002835
2836 Example:
Cyril Bonté78caf842010-03-10 22:41:43 +01002837 acl nagios src 192.168.129.3
2838 acl local_net src 192.168.0.0/16
2839 acl auth_ok http_auth(L1)
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002840
Cyril Bonté78caf842010-03-10 22:41:43 +01002841 http-request allow if nagios
2842 http-request allow if local_net auth_ok
2843 http-request auth realm Gimme if local_net auth_ok
2844 http-request deny
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002845
Cyril Bonté78caf842010-03-10 22:41:43 +01002846 Example:
2847 acl auth_ok http_auth_group(L1) G1
Cyril Bonté78caf842010-03-10 22:41:43 +01002848 http-request auth unless auth_ok
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002849
Willy Tarreau20b0de52012-12-24 15:45:22 +01002850 Example:
2851 http-request set-header X-Haproxy-Current-Date %T
2852 http-request set-header X-SSL %[ssl_fc]
2853 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id]
2854 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
2855 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
2856 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
2857 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
2858 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
2859 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
2860
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02002861 See also : "stats http-request", section 3.4 about userlists and section 7
2862 about ACL usage.
Willy Tarreauef781042010-01-27 11:53:01 +01002863
Willy Tarreauf4c43c12013-06-11 17:01:13 +02002864http-response { allow | deny | add-header <name> <fmt> | set-nice <nice> |
Willy Tarreau51347ed2013-06-11 19:34:13 +02002865 set-header <name> <fmt> | set-log-level <level> |
Lukas Tribus2dd1d1a2013-06-19 23:34:41 +02002866 set-mark <mark> | set-tos <tos> }
2867 [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02002868 Access control for Layer 7 responses
2869
2870 May be used in sections: defaults | frontend | listen | backend
2871 no | yes | yes | yes
2872
2873 The http-response statement defines a set of rules which apply to layer 7
2874 processing. The rules are evaluated in their declaration order when they are
2875 met in a frontend, listen or backend section. Any rule may optionally be
2876 followed by an ACL-based condition, in which case it will only be evaluated
2877 if the condition is true. Since these rules apply on responses, the backend
2878 rules are applied first, followed by the frontend's rules.
2879
2880 The first keyword is the rule's action. Currently supported actions include :
2881 - "allow" : this stops the evaluation of the rules and lets the response
2882 pass the check. No further "http-response" rules are evaluated for the
2883 current section.
2884
2885 - "deny" : this stops the evaluation of the rules and immediately rejects
2886 the response and emits an HTTP 502 error. No further "http-response"
2887 rules are evaluated.
2888
2889 - "add-header" appends an HTTP header field whose name is specified in
2890 <name> and whose value is defined by <fmt> which follows the log-format
2891 rules (see Custom Log Format in section 8.2.4). This may be used to send
2892 a cookie to a client for example, or to pass some internal information.
2893 This rule is not final, so it is possible to add other similar rules.
2894 Note that header addition is performed immediately, so one rule might
2895 reuse the resulting header from a previous rule.
2896
2897 - "set-header" does the same as "add-header" except that the header name
2898 is first removed if it existed. This is useful when passing security
2899 information to the server, where the header must not be manipulated by
2900 external users.
2901
Willy Tarreauf4c43c12013-06-11 17:01:13 +02002902 - "set-nice" sets the "nice" factor of the current request being processed.
2903 It only has effect against the other requests being processed at the same
2904 time. The default value is 0, unless altered by the "nice" setting on the
2905 "bind" line. The accepted range is -1024..1024. The higher the value, the
2906 nicest the request will be. Lower values will make the request more
2907 important than other ones. This can be useful to improve the speed of
2908 some requests, or lower the priority of non-important requests. Using
2909 this setting without prior experimentation can cause some major slowdown.
2910
Willy Tarreau9a355ec2013-06-11 17:45:46 +02002911 - "set-log-level" is used to change the log level of the current request
2912 when a certain condition is met. Valid levels are the 8 syslog levels
2913 (see the "log" keyword) plus the special level "silent" which disables
2914 logging for this request. This rule is not final so the last matching
2915 rule wins. This rule can be useful to disable health checks coming from
2916 another equipment.
2917
Willy Tarreau42cf39e2013-06-11 18:51:32 +02002918 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
2919 the client to the value passed in <tos> on platforms which support this.
2920 This value represents the whole 8 bits of the IP TOS field, and can be
2921 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
2922 that only the 6 higher bits are used in DSCP or TOS, and the two lower
2923 bits are always 0. This can be used to adjust some routing behaviour on
2924 border routers based on some information from the request. See RFC 2474,
2925 2597, 3260 and 4594 for more information.
2926
Willy Tarreau51347ed2013-06-11 19:34:13 +02002927 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
2928 client to the value passed in <mark> on platforms which support it. This
2929 value is an unsigned 32 bit value which can be matched by netfilter and
2930 by the routing table. It can be expressed both in decimal or hexadecimal
2931 format (prefixed by "0x"). This can be useful to force certain packets to
2932 take a different route (for example a cheaper network path for bulk
2933 downloads). This works on Linux kernels 2.6.32 and above and requires
2934 admin privileges.
2935
Willy Tarreaue365c0b2013-06-11 16:06:12 +02002936 There is no limit to the number of http-response statements per instance.
2937
Godbach09250262013-07-02 01:19:15 +08002938 It is important to know that http-response rules are processed very early in
Willy Tarreaue365c0b2013-06-11 16:06:12 +02002939 the HTTP processing, before "reqdel" or "reqrep" rules. That way, headers
2940 added by "add-header"/"set-header" are visible by almost all further ACL
2941 rules.
2942
2943 See also : "http-request", section 3.4 about userlists and section 7 about
2944 ACL usage.
2945
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02002946
2947tcp-check expect [!] <match> <pattern>
2948 Specify data to be collected and analysed during a generic health check
2949 May be used in sections: defaults | frontend | listen | backend
2950 no | no | yes | yes
2951
2952 Arguments :
2953 <match> is a keyword indicating how to look for a specific pattern in the
2954 response. The keyword may be one of "string", "rstring" or
2955 binary.
2956 The keyword may be preceded by an exclamation mark ("!") to negate
2957 the match. Spaces are allowed between the exclamation mark and the
2958 keyword. See below for more details on the supported keywords.
2959
2960 <pattern> is the pattern to look for. It may be a string or a regular
2961 expression. If the pattern contains spaces, they must be escaped
2962 with the usual backslash ('\').
2963 If the match is set to binary, then the pattern must be passed as
2964 a serie of hexadecimal digits in an even number. Each sequence of
2965 two digits will represent a byte. The hexadecimal digits may be
2966 used upper or lower case.
2967
2968
2969 The available matches are intentionally similar to their http-check cousins :
2970
2971 string <string> : test the exact string matches in the response buffer.
2972 A health check response will be considered valid if the
2973 response's buffer contains this exact string. If the
2974 "string" keyword is prefixed with "!", then the response
2975 will be considered invalid if the body contains this
2976 string. This can be used to look for a mandatory pattern
2977 in a protocol response, or to detect a failure when a
2978 specific error appears in a protocol banner.
2979
2980 rstring <regex> : test a regular expression on the response buffer.
2981 A health check response will be considered valid if the
2982 response's buffer matches this expression. If the
2983 "rstring" keyword is prefixed with "!", then the response
2984 will be considered invalid if the body matches the
2985 expression.
2986
2987 binary <hexstring> : test the exact string in its hexadecimal form matches
2988 in the response buffer. A health check response will
2989 be considered valid if the response's buffer contains
2990 this exact hexadecimal string.
2991 Purpose is to match data on binary protocols.
2992
2993 It is important to note that the responses will be limited to a certain size
2994 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
2995 Thus, too large responses may not contain the mandatory pattern when using
2996 "string", "rstring" or binary. If a large response is absolutely required, it
2997 is possible to change the default max size by setting the global variable.
2998 However, it is worth keeping in mind that parsing very large responses can
2999 waste some CPU cycles, especially when regular expressions are used, and that
3000 it is always better to focus the checks on smaller resources. Also, in its
3001 current state, the check will not find any string nor regex past a null
3002 character in the response. Similarly it is not possible to request matching
3003 the null character.
3004
3005 Examples :
3006 # perform a POP check
3007 option tcp-check
3008 tcp-check expect string +OK\ POP3\ ready
3009
3010 # perform an IMAP check
3011 option tcp-check
3012 tcp-check expect string *\ OK\ IMAP4\ ready
3013
3014 # look for the redis master server
3015 option tcp-check
3016 tcp-check send PING\r\n
3017 tcp-check expect +PONG
3018 tcp-check send info\ replication\r\n
3019 tcp-check expect string role:master
3020 tcp-check send QUIT\r\n
3021 tcp-check expect string +OK
3022
3023
3024 See also : "option tcp-check", "tcp-check send", "http-check expect",
3025 tune.chksize
3026
3027
3028tcp-check send <data>
3029 Specify a string to be sent as a question during a generic health check
3030 May be used in sections: defaults | frontend | listen | backend
3031 no | no | yes | yes
3032
3033 <data> : the data to be sent as a question during a generic health check
3034 session. For now, <data> must be a string.
3035
3036 Examples :
3037 # look for the redis master server
3038 option tcp-check
3039 tcp-check send info\ replication\r\n
3040 tcp-check expect string role:master
3041
3042 See also : "option tcp-check", "tcp-check expect", "tcp-check send-binary",
3043 tune.chksize
3044
3045
3046tcp-check send-binary <hexastring>
3047 Specify an hexa digits string to be sent as a binary question during a raw
3048 tcp health check
3049 May be used in sections: defaults | frontend | listen | backend
3050 no | no | yes | yes
3051
3052 <data> : the data to be sent as a question during a generic health check
3053 session. For now, <data> must be a string.
3054 <hexastring> : test the exact string in its hexadecimal form matches in the
3055 response buffer. A health check response will be considered
3056 valid if the response's buffer contains this exact
3057 hexadecimal string.
3058 Purpose is to send binary data to ask on binary protocols.
3059
3060 Examples :
3061 # redis check in binary
3062 option tcp-check
3063 tcp-check send-binary 50494e470d0a # PING\r\n
3064 tcp-check expect binary 2b504F4e47 # +PONG
3065
3066
3067 See also : "option tcp-check", "tcp-check expect", "tcp-check send",
3068 tune.chksize
3069
3070
3071
Mark Lamourinec2247f02012-01-04 13:02:01 -05003072http-send-name-header [<header>]
3073 Add the server name to a request. Use the header string given by <header>
3074
3075 May be used in sections: defaults | frontend | listen | backend
3076 yes | no | yes | yes
3077
3078 Arguments :
3079
3080 <header> The header string to use to send the server name
3081
3082 The "http-send-name-header" statement causes the name of the target
3083 server to be added to the headers of an HTTP request. The name
3084 is added with the header string proved.
3085
3086 See also : "server"
3087
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01003088id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02003089 Set a persistent ID to a proxy.
3090 May be used in sections : defaults | frontend | listen | backend
3091 no | yes | yes | yes
3092 Arguments : none
3093
3094 Set a persistent ID for the proxy. This ID must be unique and positive.
3095 An unused ID will automatically be assigned if unset. The first assigned
3096 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01003097
3098
Cyril Bonté0d4bf012010-04-25 23:21:46 +02003099ignore-persist { if | unless } <condition>
3100 Declare a condition to ignore persistence
3101 May be used in sections: defaults | frontend | listen | backend
3102 no | yes | yes | yes
3103
3104 By default, when cookie persistence is enabled, every requests containing
3105 the cookie are unconditionally persistent (assuming the target server is up
3106 and running).
3107
3108 The "ignore-persist" statement allows one to declare various ACL-based
3109 conditions which, when met, will cause a request to ignore persistence.
3110 This is sometimes useful to load balance requests for static files, which
3111 oftenly don't require persistence. This can also be used to fully disable
3112 persistence for a specific User-Agent (for example, some web crawler bots).
3113
3114 Combined with "appsession", it can also help reduce HAProxy memory usage, as
3115 the appsession table won't grow if persistence is ignored.
3116
3117 The persistence is ignored when an "if" condition is met, or unless an
3118 "unless" condition is met.
3119
3120 See also : "force-persist", "cookie", and section 7 about ACL usage.
3121
3122
Willy Tarreau2769aa02007-12-27 18:26:09 +01003123log global
Willy Tarreauf7edefa2009-05-10 17:20:05 +02003124log <address> <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02003125no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01003126 Enable per-instance logging of events and traffic.
3127 May be used in sections : defaults | frontend | listen | backend
3128 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02003129
3130 Prefix :
3131 no should be used when the logger list must be flushed. For example,
3132 if you don't want to inherit from the default logger list. This
3133 prefix does not allow arguments.
3134
Willy Tarreau2769aa02007-12-27 18:26:09 +01003135 Arguments :
3136 global should be used when the instance's logging parameters are the
3137 same as the global ones. This is the most common usage. "global"
3138 replaces <address>, <facility> and <level> with those of the log
3139 entries found in the "global" section. Only one "log global"
3140 statement may be used per instance, and this form takes no other
3141 parameter.
3142
3143 <address> indicates where to send the logs. It takes the same format as
3144 for the "global" section's logs, and can be one of :
3145
3146 - An IPv4 address optionally followed by a colon (':') and a UDP
3147 port. If no port is specified, 514 is used by default (the
3148 standard syslog port).
3149
David du Colombier24bb5f52011-03-17 10:40:23 +01003150 - An IPv6 address followed by a colon (':') and optionally a UDP
3151 port. If no port is specified, 514 is used by default (the
3152 standard syslog port).
3153
Willy Tarreau2769aa02007-12-27 18:26:09 +01003154 - A filesystem path to a UNIX domain socket, keeping in mind
3155 considerations for chroot (be sure the path is accessible
3156 inside the chroot) and uid/gid (be sure the path is
3157 appropriately writeable).
3158
Willy Tarreaudad36a32013-03-11 01:20:04 +01003159 Any part of the address string may reference any number of
3160 environment variables by preceding their name with a dollar
3161 sign ('$') and optionally enclosing them with braces ('{}'),
3162 similarly to what is done in Bourne shell.
3163
Willy Tarreau2769aa02007-12-27 18:26:09 +01003164 <facility> must be one of the 24 standard syslog facilities :
3165
3166 kern user mail daemon auth syslog lpr news
3167 uucp cron auth2 ftp ntp audit alert cron2
3168 local0 local1 local2 local3 local4 local5 local6 local7
3169
3170 <level> is optional and can be specified to filter outgoing messages. By
3171 default, all messages are sent. If a level is specified, only
3172 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02003173 will be sent. An optional minimum level can be specified. If it
3174 is set, logs emitted with a more severe level than this one will
3175 be capped to this level. This is used to avoid sending "emerg"
3176 messages on all terminals on some default syslog configurations.
3177 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01003178
3179 emerg alert crit err warning notice info debug
3180
William Lallemand0f99e342011-10-12 17:50:54 +02003181 It is important to keep in mind that it is the frontend which decides what to
3182 log from a connection, and that in case of content switching, the log entries
3183 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003184
3185 However, backend log declaration define how and where servers status changes
3186 will be logged. Level "notice" will be used to indicate a server going up,
3187 "warning" will be used for termination signals and definitive service
3188 termination, and "alert" will be used for when a server goes down.
3189
3190 Note : According to RFC3164, messages are truncated to 1024 bytes before
3191 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003192
3193 Example :
3194 log global
Willy Tarreauf7edefa2009-05-10 17:20:05 +02003195 log 127.0.0.1:514 local0 notice # only send important events
3196 log 127.0.0.1:514 local0 notice notice # same but limit output level
Willy Tarreaudad36a32013-03-11 01:20:04 +01003197 log ${LOCAL_SYSLOG}:514 local0 notice # send to local server
3198
Willy Tarreau2769aa02007-12-27 18:26:09 +01003199
William Lallemand48940402012-01-30 16:47:22 +01003200log-format <string>
3201 Allows you to custom a log line.
3202
3203 See also : Custom Log Format (8.2.4)
3204
Willy Tarreau2769aa02007-12-27 18:26:09 +01003205
3206maxconn <conns>
3207 Fix the maximum number of concurrent connections on a frontend
3208 May be used in sections : defaults | frontend | listen | backend
3209 yes | yes | yes | no
3210 Arguments :
3211 <conns> is the maximum number of concurrent connections the frontend will
3212 accept to serve. Excess connections will be queued by the system
3213 in the socket's listen queue and will be served once a connection
3214 closes.
3215
3216 If the system supports it, it can be useful on big sites to raise this limit
3217 very high so that haproxy manages connection queues, instead of leaving the
3218 clients with unanswered connection attempts. This value should not exceed the
3219 global maxconn. Also, keep in mind that a connection contains two buffers
3220 of 8kB each, as well as some other data resulting in about 17 kB of RAM being
3221 consumed per established connection. That means that a medium system equipped
3222 with 1GB of RAM can withstand around 40000-50000 concurrent connections if
3223 properly tuned.
3224
3225 Also, when <conns> is set to large values, it is possible that the servers
3226 are not sized to accept such loads, and for this reason it is generally wise
3227 to assign them some reasonable connection limits.
3228
Vincent Bernat6341be52012-06-27 17:18:30 +02003229 By default, this value is set to 2000.
3230
Willy Tarreau2769aa02007-12-27 18:26:09 +01003231 See also : "server", global section's "maxconn", "fullconn"
3232
3233
3234mode { tcp|http|health }
3235 Set the running mode or protocol of the instance
3236 May be used in sections : defaults | frontend | listen | backend
3237 yes | yes | yes | yes
3238 Arguments :
3239 tcp The instance will work in pure TCP mode. A full-duplex connection
3240 will be established between clients and servers, and no layer 7
3241 examination will be performed. This is the default mode. It
3242 should be used for SSL, SSH, SMTP, ...
3243
3244 http The instance will work in HTTP mode. The client request will be
3245 analyzed in depth before connecting to any server. Any request
3246 which is not RFC-compliant will be rejected. Layer 7 filtering,
3247 processing and switching will be possible. This is the mode which
3248 brings HAProxy most of its value.
3249
3250 health The instance will work in "health" mode. It will just reply "OK"
Willy Tarreau82569f92012-09-27 23:48:56 +02003251 to incoming connections and close the connection. Alternatively,
3252 If the "httpchk" option is set, "HTTP/1.0 200 OK" will be sent
3253 instead. Nothing will be logged in either case. This mode is used
3254 to reply to external components health checks. This mode is
3255 deprecated and should not be used anymore as it is possible to do
3256 the same and even better by combining TCP or HTTP modes with the
3257 "monitor" keyword.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003258
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003259 When doing content switching, it is mandatory that the frontend and the
3260 backend are in the same mode (generally HTTP), otherwise the configuration
3261 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003262
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003263 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01003264 defaults http_instances
3265 mode http
3266
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003267 See also : "monitor", "monitor-net"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003268
Willy Tarreau0ba27502007-12-24 16:55:16 +01003269
Cyril Bontéf0c60612010-02-06 14:44:47 +01003270monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01003271 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003272 May be used in sections : defaults | frontend | listen | backend
3273 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01003274 Arguments :
3275 if <cond> the monitor request will fail if the condition is satisfied,
3276 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003277 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01003278 are met, for instance a low number of servers both in a
3279 backend and its backup.
3280
3281 unless <cond> the monitor request will succeed only if the condition is
3282 satisfied, and will fail otherwise. Such a condition may be
3283 based on a test on the presence of a minimum number of active
3284 servers in a list of backends.
3285
3286 This statement adds a condition which can force the response to a monitor
3287 request to report a failure. By default, when an external component queries
3288 the URI dedicated to monitoring, a 200 response is returned. When one of the
3289 conditions above is met, haproxy will return 503 instead of 200. This is
3290 very useful to report a site failure to an external component which may base
3291 routing advertisements between multiple sites on the availability reported by
3292 haproxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003293 criterion. Note that "monitor fail" only works in HTTP mode. Both status
3294 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003295
3296 Example:
3297 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01003298 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01003299 acl site_dead nbsrv(dynamic) lt 2
3300 acl site_dead nbsrv(static) lt 2
3301 monitor-uri /site_alive
3302 monitor fail if site_dead
3303
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003304 See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003305
3306
3307monitor-net <source>
3308 Declare a source network which is limited to monitor requests
3309 May be used in sections : defaults | frontend | listen | backend
3310 yes | yes | yes | no
3311 Arguments :
3312 <source> is the source IPv4 address or network which will only be able to
3313 get monitor responses to any request. It can be either an IPv4
3314 address, a host name, or an address followed by a slash ('/')
3315 followed by a mask.
3316
3317 In TCP mode, any connection coming from a source matching <source> will cause
3318 the connection to be immediately closed without any log. This allows another
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003319 equipment to probe the port and verify that it is still listening, without
Willy Tarreau2769aa02007-12-27 18:26:09 +01003320 forwarding the connection to a remote server.
3321
3322 In HTTP mode, a connection coming from a source matching <source> will be
3323 accepted, the following response will be sent without waiting for a request,
3324 then the connection will be closed : "HTTP/1.0 200 OK". This is normally
3325 enough for any front-end HTTP probe to detect that the service is UP and
Willy Tarreau82569f92012-09-27 23:48:56 +02003326 running without forwarding the request to a backend server. Note that this
3327 response is sent in raw format, without any transformation. This is important
3328 as it means that it will not be SSL-encrypted on SSL listeners.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003329
Willy Tarreau82569f92012-09-27 23:48:56 +02003330 Monitor requests are processed very early, just after tcp-request connection
3331 ACLs which are the only ones able to block them. These connections are short
3332 lived and never wait for any data from the client. They cannot be logged, and
3333 it is the intended purpose. They are only used to report HAProxy's health to
3334 an upper component, nothing more. Please note that "monitor fail" rules do
3335 not apply to connections intercepted by "monitor-net".
Willy Tarreau2769aa02007-12-27 18:26:09 +01003336
Willy Tarreau95cd2832010-03-04 23:36:33 +01003337 Last, please note that only one "monitor-net" statement can be specified in
3338 a frontend. If more than one is found, only the last one will be considered.
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003339
Willy Tarreau2769aa02007-12-27 18:26:09 +01003340 Example :
3341 # addresses .252 and .253 are just probing us.
3342 frontend www
3343 monitor-net 192.168.0.252/31
3344
3345 See also : "monitor fail", "monitor-uri"
3346
3347
3348monitor-uri <uri>
3349 Intercept a URI used by external components' monitor requests
3350 May be used in sections : defaults | frontend | listen | backend
3351 yes | yes | yes | no
3352 Arguments :
3353 <uri> is the exact URI which we want to intercept to return HAProxy's
3354 health status instead of forwarding the request.
3355
3356 When an HTTP request referencing <uri> will be received on a frontend,
3357 HAProxy will not forward it nor log it, but instead will return either
3358 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
3359 conditions defined with "monitor fail". This is normally enough for any
3360 front-end HTTP probe to detect that the service is UP and running without
3361 forwarding the request to a backend server. Note that the HTTP method, the
3362 version and all headers are ignored, but the request must at least be valid
3363 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
3364
3365 Monitor requests are processed very early. It is not possible to block nor
3366 divert them using ACLs. They cannot be logged either, and it is the intended
3367 purpose. They are only used to report HAProxy's health to an upper component,
3368 nothing more. However, it is possible to add any number of conditions using
3369 "monitor fail" and ACLs so that the result can be adjusted to whatever check
3370 can be imagined (most often the number of available servers in a backend).
3371
3372 Example :
3373 # Use /haproxy_test to report haproxy's status
3374 frontend www
3375 mode http
3376 monitor-uri /haproxy_test
3377
3378 See also : "monitor fail", "monitor-net"
3379
Willy Tarreau0ba27502007-12-24 16:55:16 +01003380
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003381option abortonclose
3382no option abortonclose
3383 Enable or disable early dropping of aborted requests pending in queues.
3384 May be used in sections : defaults | frontend | listen | backend
3385 yes | no | yes | yes
3386 Arguments : none
3387
3388 In presence of very high loads, the servers will take some time to respond.
3389 The per-instance connection queue will inflate, and the response time will
3390 increase respective to the size of the queue times the average per-session
3391 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01003392 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003393 the queue, and slowing down other users, and the servers as well, because the
3394 request will eventually be served, then aborted at the first error
3395 encountered while delivering the response.
3396
3397 As there is no way to distinguish between a full STOP and a simple output
3398 close on the client side, HTTP agents should be conservative and consider
3399 that the client might only have closed its output channel while waiting for
3400 the response. However, this introduces risks of congestion when lots of users
3401 do the same, and is completely useless nowadays because probably no client at
3402 all will close the session while waiting for the response. Some HTTP agents
Willy Tarreaud72758d2010-01-12 10:42:19 +01003403 support this behaviour (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003404 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01003405 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003406 of being the single component to break rare but valid traffic is extremely
3407 low, which adds to the temptation to be able to abort a session early while
3408 still not served and not pollute the servers.
3409
3410 In HAProxy, the user can choose the desired behaviour using the option
3411 "abortonclose". By default (without the option) the behaviour is HTTP
3412 compliant and aborted requests will be served. But when the option is
3413 specified, a session with an incoming channel closed will be aborted while
3414 it is still possible, either pending in the queue for a connection slot, or
3415 during the connection establishment if the server has not yet acknowledged
3416 the connection request. This considerably reduces the queue size and the load
3417 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01003418 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003419
3420 If this option has been enabled in a "defaults" section, it can be disabled
3421 in a specific instance by prepending the "no" keyword before it.
3422
3423 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
3424
3425
Willy Tarreau4076a152009-04-02 15:18:36 +02003426option accept-invalid-http-request
3427no option accept-invalid-http-request
3428 Enable or disable relaxing of HTTP request parsing
3429 May be used in sections : defaults | frontend | listen | backend
3430 yes | yes | yes | no
3431 Arguments : none
3432
3433 By default, HAProxy complies with RFC2616 in terms of message parsing. This
3434 means that invalid characters in header names are not permitted and cause an
3435 error to be returned to the client. This is the desired behaviour as such
3436 forbidden characters are essentially used to build attacks exploiting server
3437 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
3438 server will emit invalid header names for whatever reason (configuration,
3439 implementation) and the issue will not be immediately fixed. In such a case,
3440 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01003441 even if that does not make sense, by specifying this option. Similarly, the
3442 list of characters allowed to appear in a URI is well defined by RFC3986, and
3443 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
3444 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
3445 not allowed at all. Haproxy always blocks a number of them (0..32, 127). The
3446 remaining ones are blocked by default unless this option is enabled.
Willy Tarreau4076a152009-04-02 15:18:36 +02003447
3448 This option should never be enabled by default as it hides application bugs
3449 and open security breaches. It should only be deployed after a problem has
3450 been confirmed.
3451
3452 When this option is enabled, erroneous header names will still be accepted in
3453 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01003454 analysis using the "show errors" request on the UNIX stats socket. Similarly,
3455 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02003456 also helps confirming that the issue has been solved.
3457
3458 If this option has been enabled in a "defaults" section, it can be disabled
3459 in a specific instance by prepending the "no" keyword before it.
3460
3461 See also : "option accept-invalid-http-response" and "show errors" on the
3462 stats socket.
3463
3464
3465option accept-invalid-http-response
3466no option accept-invalid-http-response
3467 Enable or disable relaxing of HTTP response parsing
3468 May be used in sections : defaults | frontend | listen | backend
3469 yes | no | yes | yes
3470 Arguments : none
3471
3472 By default, HAProxy complies with RFC2616 in terms of message parsing. This
3473 means that invalid characters in header names are not permitted and cause an
3474 error to be returned to the client. This is the desired behaviour as such
3475 forbidden characters are essentially used to build attacks exploiting server
3476 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
3477 server will emit invalid header names for whatever reason (configuration,
3478 implementation) and the issue will not be immediately fixed. In such a case,
3479 it is possible to relax HAProxy's header name parser to accept any character
3480 even if that does not make sense, by specifying this option.
3481
3482 This option should never be enabled by default as it hides application bugs
3483 and open security breaches. It should only be deployed after a problem has
3484 been confirmed.
3485
3486 When this option is enabled, erroneous header names will still be accepted in
3487 responses, but the complete response will be captured in order to permit
3488 later analysis using the "show errors" request on the UNIX stats socket.
3489 Doing this also helps confirming that the issue has been solved.
3490
3491 If this option has been enabled in a "defaults" section, it can be disabled
3492 in a specific instance by prepending the "no" keyword before it.
3493
3494 See also : "option accept-invalid-http-request" and "show errors" on the
3495 stats socket.
3496
3497
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003498option allbackups
3499no option allbackups
3500 Use either all backup servers at a time or only the first one
3501 May be used in sections : defaults | frontend | listen | backend
3502 yes | no | yes | yes
3503 Arguments : none
3504
3505 By default, the first operational backup server gets all traffic when normal
3506 servers are all down. Sometimes, it may be preferred to use multiple backups
3507 at once, because one will not be enough. When "option allbackups" is enabled,
3508 the load balancing will be performed among all backup servers when all normal
3509 ones are unavailable. The same load balancing algorithm will be used and the
3510 servers' weights will be respected. Thus, there will not be any priority
3511 order between the backup servers anymore.
3512
3513 This option is mostly used with static server farms dedicated to return a
3514 "sorry" page when an application is completely offline.
3515
3516 If this option has been enabled in a "defaults" section, it can be disabled
3517 in a specific instance by prepending the "no" keyword before it.
3518
3519
3520option checkcache
3521no option checkcache
Godbach7056a352013-12-11 20:01:07 +08003522 Analyze all server responses and block responses with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003523 May be used in sections : defaults | frontend | listen | backend
3524 yes | no | yes | yes
3525 Arguments : none
3526
3527 Some high-level frameworks set application cookies everywhere and do not
3528 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003529 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003530 high risk of session crossing or stealing between users traversing the same
3531 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02003532 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003533
3534 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003535 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01003536 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003537 response to check if there's a risk of caching a cookie on a client-side
3538 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01003539 to the client are :
3540 - all those without "Set-Cookie" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003541 - all those with a return code other than 200, 203, 206, 300, 301, 410,
Willy Tarreau198a7442008-01-17 12:05:32 +01003542 provided that the server has not set a "Cache-control: public" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003543 - all those that come from a POST request, provided that the server has not
3544 set a 'Cache-Control: public' header ;
3545 - those with a 'Pragma: no-cache' header
3546 - those with a 'Cache-control: private' header
3547 - those with a 'Cache-control: no-store' header
3548 - those with a 'Cache-control: max-age=0' header
3549 - those with a 'Cache-control: s-maxage=0' header
3550 - those with a 'Cache-control: no-cache' header
3551 - those with a 'Cache-control: no-cache="set-cookie"' header
3552 - those with a 'Cache-control: no-cache="set-cookie,' header
3553 (allowing other fields after set-cookie)
3554
3555 If a response doesn't respect these requirements, then it will be blocked
Willy Tarreau198a7442008-01-17 12:05:32 +01003556 just as if it was from an "rspdeny" filter, with an "HTTP 502 bad gateway".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003557 The session state shows "PH--" meaning that the proxy blocked the response
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003558 during headers processing. Additionally, an alert will be sent in the logs so
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003559 that admins are informed that there's something to be fixed.
3560
3561 Due to the high impact on the application, the application should be tested
3562 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003563 good practice to always activate it during tests, even if it is not used in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003564 production, as it will report potentially dangerous application behaviours.
3565
3566 If this option has been enabled in a "defaults" section, it can be disabled
3567 in a specific instance by prepending the "no" keyword before it.
3568
3569
3570option clitcpka
3571no option clitcpka
3572 Enable or disable the sending of TCP keepalive packets on the client side
3573 May be used in sections : defaults | frontend | listen | backend
3574 yes | yes | yes | no
3575 Arguments : none
3576
3577 When there is a firewall or any session-aware component between a client and
3578 a server, and when the protocol involves very long sessions with long idle
3579 periods (eg: remote desktops), there is a risk that one of the intermediate
3580 components decides to expire a session which has remained idle for too long.
3581
3582 Enabling socket-level TCP keep-alives makes the system regularly send packets
3583 to the other end of the connection, leaving it active. The delay between
3584 keep-alive probes is controlled by the system only and depends both on the
3585 operating system and its tuning parameters.
3586
3587 It is important to understand that keep-alive packets are neither emitted nor
3588 received at the application level. It is only the network stacks which sees
3589 them. For this reason, even if one side of the proxy already uses keep-alives
3590 to maintain its connection alive, those keep-alive packets will not be
3591 forwarded to the other side of the proxy.
3592
3593 Please note that this has nothing to do with HTTP keep-alive.
3594
3595 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
3596 client side of a connection, which should help when session expirations are
3597 noticed between HAProxy and a client.
3598
3599 If this option has been enabled in a "defaults" section, it can be disabled
3600 in a specific instance by prepending the "no" keyword before it.
3601
3602 See also : "option srvtcpka", "option tcpka"
3603
3604
Willy Tarreau0ba27502007-12-24 16:55:16 +01003605option contstats
3606 Enable continuous traffic statistics updates
3607 May be used in sections : defaults | frontend | listen | backend
3608 yes | yes | yes | no
3609 Arguments : none
3610
3611 By default, counters used for statistics calculation are incremented
3612 only when a session finishes. It works quite well when serving small
3613 objects, but with big ones (for example large images or archives) or
3614 with A/V streaming, a graph generated from haproxy counters looks like
3615 a hedgehog. With this option enabled counters get incremented continuously,
3616 during a whole session. Recounting touches a hotpath directly so
3617 it is not enabled by default, as it has small performance impact (~0.5%).
3618
3619
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02003620option dontlog-normal
3621no option dontlog-normal
3622 Enable or disable logging of normal, successful connections
3623 May be used in sections : defaults | frontend | listen | backend
3624 yes | yes | yes | no
3625 Arguments : none
3626
3627 There are large sites dealing with several thousand connections per second
3628 and for which logging is a major pain. Some of them are even forced to turn
3629 logs off and cannot debug production issues. Setting this option ensures that
3630 normal connections, those which experience no error, no timeout, no retry nor
3631 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
3632 mode, the response status code is checked and return codes 5xx will still be
3633 logged.
3634
3635 It is strongly discouraged to use this option as most of the time, the key to
3636 complex issues is in the normal logs which will not be logged here. If you
3637 need to separate logs, see the "log-separate-errors" option instead.
3638
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003639 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02003640 logging.
3641
3642
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003643option dontlognull
3644no option dontlognull
3645 Enable or disable logging of null connections
3646 May be used in sections : defaults | frontend | listen | backend
3647 yes | yes | yes | no
3648 Arguments : none
3649
3650 In certain environments, there are components which will regularly connect to
3651 various systems to ensure that they are still alive. It can be the case from
3652 another load balancer as well as from monitoring systems. By default, even a
3653 simple port probe or scan will produce a log. If those connections pollute
3654 the logs too much, it is possible to enable option "dontlognull" to indicate
3655 that a connection on which no data has been transferred will not be logged,
3656 which typically corresponds to those probes.
3657
3658 It is generally recommended not to use this option in uncontrolled
3659 environments (eg: internet), otherwise scans and other malicious activities
3660 would not be logged.
3661
3662 If this option has been enabled in a "defaults" section, it can be disabled
3663 in a specific instance by prepending the "no" keyword before it.
3664
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003665 See also : "log", "monitor-net", "monitor-uri" and section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003666
3667
3668option forceclose
3669no option forceclose
3670 Enable or disable active connection closing after response is transferred.
3671 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaua31e5df2009-12-30 01:10:35 +01003672 yes | yes | yes | yes
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003673 Arguments : none
3674
3675 Some HTTP servers do not necessarily close the connections when they receive
3676 the "Connection: close" set by "option httpclose", and if the client does not
3677 close either, then the connection remains open till the timeout expires. This
3678 causes high number of simultaneous connections on the servers and shows high
3679 global session times in the logs.
3680
3681 When this happens, it is possible to use "option forceclose". It will
Willy Tarreau82eeaf22009-12-29 12:09:05 +01003682 actively close the outgoing server channel as soon as the server has finished
Willy Tarreau0dfdf192010-01-05 11:33:11 +01003683 to respond. This option implicitly enables the "httpclose" option. Note that
3684 this option also enables the parsing of the full request and response, which
3685 means we can close the connection to the server very quickly, releasing some
3686 resources earlier than with httpclose.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003687
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02003688 This option may also be combined with "option http-pretend-keepalive", which
3689 will disable sending of the "Connection: close" header, but will still cause
3690 the connection to be closed once the whole response is received.
3691
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003692 If this option has been enabled in a "defaults" section, it can be disabled
3693 in a specific instance by prepending the "no" keyword before it.
3694
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02003695 See also : "option httpclose" and "option http-pretend-keepalive"
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003696
3697
Willy Tarreau87cf5142011-08-19 22:57:24 +02003698option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01003699 Enable insertion of the X-Forwarded-For header to requests sent to servers
3700 May be used in sections : defaults | frontend | listen | backend
3701 yes | yes | yes | yes
3702 Arguments :
3703 <network> is an optional argument used to disable this option for sources
3704 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02003705 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01003706 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01003707
3708 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
3709 their client address. This is sometimes annoying when the client's IP address
3710 is expected in server logs. To solve this problem, the well-known HTTP header
3711 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
3712 This header contains a value representing the client's IP address. Since this
3713 header is always appended at the end of the existing header list, the server
3714 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02003715 the server's manual to find how to enable use of this standard header. Note
3716 that only the last occurrence of the header must be used, since it is really
3717 possible that the client has already brought one.
3718
Willy Tarreaud72758d2010-01-12 10:42:19 +01003719 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02003720 the default "X-Forwarded-For". This can be useful where you might already
Willy Tarreaud72758d2010-01-12 10:42:19 +01003721 have a "X-Forwarded-For" header from a different application (eg: stunnel),
3722 and you need preserve it. Also if your backend server doesn't use the
Ross Westaf72a1d2008-08-03 10:51:45 +02003723 "X-Forwarded-For" header and requires different one (eg: Zeus Web Servers
3724 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01003725
3726 Sometimes, a same HAProxy instance may be shared between a direct client
3727 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
3728 used to decrypt HTTPS traffic). It is possible to disable the addition of the
3729 header for a known source address or network by adding the "except" keyword
3730 followed by the network address. In this case, any source IP matching the
3731 network will not cause an addition of this header. Most common uses are with
3732 private networks or 127.0.0.1.
3733
Willy Tarreau87cf5142011-08-19 22:57:24 +02003734 Alternatively, the keyword "if-none" states that the header will only be
3735 added if it is not present. This should only be used in perfectly trusted
3736 environment, as this might cause a security issue if headers reaching haproxy
3737 are under the control of the end-user.
3738
Willy Tarreauc27debf2008-01-06 08:57:02 +01003739 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02003740 least one of them uses it, the header will be added. Note that the backend's
3741 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02003742 both are defined. In the case of the "if-none" argument, if at least one of
3743 the frontend or the backend does not specify it, it wants the addition to be
3744 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01003745
Willy Tarreau87cf5142011-08-19 22:57:24 +02003746 It is important to note that by default, HAProxy works in tunnel mode and
3747 only inspects the first request of a connection, meaning that only the first
3748 request will have the header appended, which is certainly not what you want.
3749 In order to fix this, ensure that any of the "httpclose", "forceclose" or
3750 "http-server-close" options is set when using this option.
Willy Tarreauc27debf2008-01-06 08:57:02 +01003751
Ross Westaf72a1d2008-08-03 10:51:45 +02003752 Examples :
Willy Tarreauc27debf2008-01-06 08:57:02 +01003753 # Public HTTP address also used by stunnel on the same machine
3754 frontend www
3755 mode http
3756 option forwardfor except 127.0.0.1 # stunnel already adds the header
3757
Ross Westaf72a1d2008-08-03 10:51:45 +02003758 # Those servers want the IP Address in X-Client
3759 backend www
3760 mode http
3761 option forwardfor header X-Client
3762
Willy Tarreau87cf5142011-08-19 22:57:24 +02003763 See also : "option httpclose", "option http-server-close",
3764 "option forceclose"
Willy Tarreauc27debf2008-01-06 08:57:02 +01003765
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02003766
Willy Tarreau16bfb022010-01-16 19:48:41 +01003767option http-keep-alive
3768no option http-keep-alive
3769 Enable or disable HTTP keep-alive from client to server
3770 May be used in sections : defaults | frontend | listen | backend
3771 yes | yes | yes | yes
3772 Arguments : none
3773
3774 By default, when a client communicates with a server, HAProxy will only
3775 analyze, log, and process the first request of each connection. Setting
3776 "option http-keep-alive" enables HTTP keep-alive mode on the client- and
3777 server- sides. This provides the lowest latency on the client side (slow
3778 network) and the fastest session reuse on the server side at the expense
3779 of maintaining idle connections to the servers. In general, it is possible
3780 with this option to achieve approximately twice the request rate that the
3781 "http-server-close" option achieves on small objects. There are mainly two
3782 situations where this option may be useful :
3783
3784 - when the server is non-HTTP compliant and authenticates the connection
3785 instead of requests (eg: NTLM authentication)
3786
3787 - when the cost of establishing the connection to the server is significant
3788 compared to the cost of retrieving the associated object from the server.
3789
3790 This last case can happen when the server is a fast static server of cache.
3791 In this case, the server will need to be properly tuned to support high enough
3792 connection counts because connections will last until the client sends another
3793 request.
3794
3795 If the client request has to go to another backend or another server due to
3796 content switching or the load balancing algorithm, the idle connection will
Willy Tarreau9420b122013-12-15 18:58:25 +01003797 immediately be closed and a new one re-opened. Option "prefer-last-server" is
3798 available to try optimize server selection so that if the server currently
3799 attached to an idle connection is usable, it will be used.
Willy Tarreau16bfb022010-01-16 19:48:41 +01003800
3801 In general it is preferred to use "option http-server-close" with application
3802 servers, and some static servers might benefit from "option http-keep-alive".
3803
3804 At the moment, logs will not indicate whether requests came from the same
3805 session or not. The accept date reported in the logs corresponds to the end
3806 of the previous request, and the request time corresponds to the time spent
3807 waiting for a new request. The keep-alive request time is still bound to the
3808 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
3809 not set.
3810
3811 This option may be set both in a frontend and in a backend. It is enabled if
3812 at least one of the frontend or backend holding a connection has it enabled.
3813 It is worth noting that "option forceclose" and "option http-server-close"
3814 have precedence over "option http-keep-alive".
3815
3816 If this option has been enabled in a "defaults" section, it can be disabled
3817 in a specific instance by prepending the "no" keyword before it.
3818
3819 See also : "option forceclose", "option http-server-close",
Willy Tarreau9420b122013-12-15 18:58:25 +01003820 "option prefer-last-server", "option http-pretend-keepalive",
3821 "option httpclose", and "1.1. The HTTP transaction model".
Willy Tarreau16bfb022010-01-16 19:48:41 +01003822
3823
Willy Tarreau96e31212011-05-30 18:10:30 +02003824option http-no-delay
3825no option http-no-delay
3826 Instruct the system to favor low interactive delays over performance in HTTP
3827 May be used in sections : defaults | frontend | listen | backend
3828 yes | yes | yes | yes
3829 Arguments : none
3830
3831 In HTTP, each payload is unidirectional and has no notion of interactivity.
3832 Any agent is expected to queue data somewhat for a reasonably low delay.
3833 There are some very rare server-to-server applications that abuse the HTTP
3834 protocol and expect the payload phase to be highly interactive, with many
3835 interleaved data chunks in both directions within a single request. This is
3836 absolutely not supported by the HTTP specification and will not work across
3837 most proxies or servers. When such applications attempt to do this through
3838 haproxy, it works but they will experience high delays due to the network
3839 optimizations which favor performance by instructing the system to wait for
3840 enough data to be available in order to only send full packets. Typical
3841 delays are around 200 ms per round trip. Note that this only happens with
3842 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
3843 affected.
3844
3845 When "option http-no-delay" is present in either the frontend or the backend
3846 used by a connection, all such optimizations will be disabled in order to
3847 make the exchanges as fast as possible. Of course this offers no guarantee on
3848 the functionality, as it may break at any other place. But if it works via
3849 HAProxy, it will work as fast as possible. This option should never be used
3850 by default, and should never be used at all unless such a buggy application
3851 is discovered. The impact of using this option is an increase of bandwidth
3852 usage and CPU usage, which may significantly lower performance in high
3853 latency environments.
3854
3855
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02003856option http-pretend-keepalive
3857no option http-pretend-keepalive
3858 Define whether haproxy will announce keepalive to the server or not
3859 May be used in sections : defaults | frontend | listen | backend
3860 yes | yes | yes | yes
3861 Arguments : none
3862
3863 When running with "option http-server-close" or "option forceclose", haproxy
3864 adds a "Connection: close" header to the request forwarded to the server.
3865 Unfortunately, when some servers see this header, they automatically refrain
3866 from using the chunked encoding for responses of unknown length, while this
3867 is totally unrelated. The immediate effect is that this prevents haproxy from
3868 maintaining the client connection alive. A second effect is that a client or
3869 a cache could receive an incomplete response without being aware of it, and
3870 consider the response complete.
3871
3872 By setting "option http-pretend-keepalive", haproxy will make the server
3873 believe it will keep the connection alive. The server will then not fall back
3874 to the abnormal undesired above. When haproxy gets the whole response, it
3875 will close the connection with the server just as it would do with the
3876 "forceclose" option. That way the client gets a normal response and the
3877 connection is correctly closed on the server side.
3878
3879 It is recommended not to enable this option by default, because most servers
3880 will more efficiently close the connection themselves after the last packet,
3881 and release its buffers slightly earlier. Also, the added packet on the
3882 network could slightly reduce the overall peak performance. However it is
3883 worth noting that when this option is enabled, haproxy will have slightly
3884 less work to do. So if haproxy is the bottleneck on the whole architecture,
3885 enabling this option might save a few CPU cycles.
3886
3887 This option may be set both in a frontend and in a backend. It is enabled if
3888 at least one of the frontend or backend holding a connection has it enabled.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003889 This option may be combined with "option httpclose", which will cause
Willy Tarreau22a95342010-09-29 14:31:41 +02003890 keepalive to be announced to the server and close to be announced to the
3891 client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02003892
3893 If this option has been enabled in a "defaults" section, it can be disabled
3894 in a specific instance by prepending the "no" keyword before it.
3895
Willy Tarreau16bfb022010-01-16 19:48:41 +01003896 See also : "option forceclose", "option http-server-close", and
3897 "option http-keep-alive"
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02003898
Willy Tarreauc27debf2008-01-06 08:57:02 +01003899
Willy Tarreaub608feb2010-01-02 22:47:18 +01003900option http-server-close
3901no option http-server-close
3902 Enable or disable HTTP connection closing on the server side
3903 May be used in sections : defaults | frontend | listen | backend
3904 yes | yes | yes | yes
3905 Arguments : none
3906
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02003907 By default, when a client communicates with a server, HAProxy will only
3908 analyze, log, and process the first request of each connection. Setting
3909 "option http-server-close" enables HTTP connection-close mode on the server
3910 side while keeping the ability to support HTTP keep-alive and pipelining on
3911 the client side. This provides the lowest latency on the client side (slow
3912 network) and the fastest session reuse on the server side to save server
3913 resources, similarly to "option forceclose". It also permits non-keepalive
3914 capable servers to be served in keep-alive mode to the clients if they
3915 conform to the requirements of RFC2616. Please note that some servers do not
3916 always conform to those requirements when they see "Connection: close" in the
3917 request. The effect will be that keep-alive will never be used. A workaround
3918 consists in enabling "option http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01003919
3920 At the moment, logs will not indicate whether requests came from the same
3921 session or not. The accept date reported in the logs corresponds to the end
3922 of the previous request, and the request time corresponds to the time spent
3923 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01003924 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
3925 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01003926
3927 This option may be set both in a frontend and in a backend. It is enabled if
3928 at least one of the frontend or backend holding a connection has it enabled.
Willy Tarreau0dfdf192010-01-05 11:33:11 +01003929 It is worth noting that "option forceclose" has precedence over "option
3930 http-server-close" and that combining "http-server-close" with "httpclose"
3931 basically achieve the same result as "forceclose".
Willy Tarreaub608feb2010-01-02 22:47:18 +01003932
3933 If this option has been enabled in a "defaults" section, it can be disabled
3934 in a specific instance by prepending the "no" keyword before it.
3935
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02003936 See also : "option forceclose", "option http-pretend-keepalive",
Willy Tarreau16bfb022010-01-16 19:48:41 +01003937 "option httpclose", "option http-keep-alive", and
3938 "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01003939
3940
Willy Tarreau88d349d2010-01-25 12:15:43 +01003941option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01003942no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01003943 Make use of non-standard Proxy-Connection header instead of Connection
3944 May be used in sections : defaults | frontend | listen | backend
3945 yes | yes | yes | no
3946 Arguments : none
3947
3948 While RFC2616 explicitly states that HTTP/1.1 agents must use the
3949 Connection header to indicate their wish of persistent or non-persistent
3950 connections, both browsers and proxies ignore this header for proxied
3951 connections and make use of the undocumented, non-standard Proxy-Connection
3952 header instead. The issue begins when trying to put a load balancer between
3953 browsers and such proxies, because there will be a difference between what
3954 haproxy understands and what the client and the proxy agree on.
3955
3956 By setting this option in a frontend, haproxy can automatically switch to use
3957 that non-standard header if it sees proxied requests. A proxied request is
3958 defined here as one where the URI begins with neither a '/' nor a '*'. The
3959 choice of header only affects requests passing through proxies making use of
3960 one of the "httpclose", "forceclose" and "http-server-close" options. Note
3961 that this option can only be specified in a frontend and will affect the
3962 request along its whole life.
3963
Willy Tarreau844a7e72010-01-31 21:46:18 +01003964 Also, when this option is set, a request which requires authentication will
3965 automatically switch to use proxy authentication headers if it is itself a
3966 proxied request. That makes it possible to check or enforce authentication in
3967 front of an existing proxy.
3968
Willy Tarreau88d349d2010-01-25 12:15:43 +01003969 This option should normally never be used, except in front of a proxy.
3970
3971 See also : "option httpclose", "option forceclose" and "option
3972 http-server-close".
3973
3974
Willy Tarreaud63335a2010-02-26 12:56:52 +01003975option httpchk
3976option httpchk <uri>
3977option httpchk <method> <uri>
3978option httpchk <method> <uri> <version>
3979 Enable HTTP protocol to check on the servers health
3980 May be used in sections : defaults | frontend | listen | backend
3981 yes | no | yes | yes
3982 Arguments :
3983 <method> is the optional HTTP method used with the requests. When not set,
3984 the "OPTIONS" method is used, as it generally requires low server
3985 processing and is easy to filter out from the logs. Any method
3986 may be used, though it is not recommended to invent non-standard
3987 ones.
3988
3989 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
3990 which is accessible by default on almost any server, but may be
3991 changed to any other URI. Query strings are permitted.
3992
3993 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
3994 but some servers might behave incorrectly in HTTP 1.0, so turning
3995 it to HTTP/1.1 may sometimes help. Note that the Host field is
3996 mandatory in HTTP/1.1, and as a trick, it is possible to pass it
3997 after "\r\n" following the version string.
3998
3999 By default, server health checks only consist in trying to establish a TCP
4000 connection. When "option httpchk" is specified, a complete HTTP request is
4001 sent once the TCP connection is established, and responses 2xx and 3xx are
4002 considered valid, while all other ones indicate a server failure, including
4003 the lack of any response.
4004
4005 The port and interval are specified in the server configuration.
4006
4007 This option does not necessarily require an HTTP backend, it also works with
4008 plain TCP backends. This is particularly useful to check simple scripts bound
4009 to some dedicated ports using the inetd daemon.
4010
4011 Examples :
4012 # Relay HTTPS traffic to Apache instance and check service availability
4013 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
4014 backend https_relay
4015 mode tcp
4016 option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www
4017 server apache1 192.168.1.1:443 check port 80
4018
Simon Hormanafc47ee2013-11-25 10:46:35 +09004019 See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",
4020 "option pgsql-check", "http-check" and the "check", "port" and
4021 "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01004022
4023
Willy Tarreauc27debf2008-01-06 08:57:02 +01004024option httpclose
4025no option httpclose
4026 Enable or disable passive HTTP connection closing
4027 May be used in sections : defaults | frontend | listen | backend
4028 yes | yes | yes | yes
4029 Arguments : none
4030
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02004031 By default, when a client communicates with a server, HAProxy will only
4032 analyze, log, and process the first request of each connection. If "option
4033 httpclose" is set, it will check if a "Connection: close" header is already
4034 set in each direction, and will add one if missing. Each end should react to
4035 this by actively closing the TCP connection after each transfer, thus
4036 resulting in a switch to the HTTP close mode. Any "Connection" header
4037 different from "close" will also be removed.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004038
4039 It seldom happens that some servers incorrectly ignore this header and do not
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004040 close the connection even though they reply "Connection: close". For this
Willy Tarreau0dfdf192010-01-05 11:33:11 +01004041 reason, they are not compatible with older HTTP 1.0 browsers. If this happens
4042 it is possible to use the "option forceclose" which actively closes the
4043 request connection once the server responds. Option "forceclose" also
4044 releases the server connection earlier because it does not have to wait for
4045 the client to acknowledge it.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004046
4047 This option may be set both in a frontend and in a backend. It is enabled if
4048 at least one of the frontend or backend holding a connection has it enabled.
4049 If "option forceclose" is specified too, it has precedence over "httpclose".
Willy Tarreau0dfdf192010-01-05 11:33:11 +01004050 If "option http-server-close" is enabled at the same time as "httpclose", it
4051 basically achieves the same result as "option forceclose".
Willy Tarreauc27debf2008-01-06 08:57:02 +01004052
4053 If this option has been enabled in a "defaults" section, it can be disabled
4054 in a specific instance by prepending the "no" keyword before it.
4055
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02004056 See also : "option forceclose", "option http-server-close" and
4057 "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01004058
4059
Emeric Brun3a058f32009-06-30 18:26:00 +02004060option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01004061 Enable logging of HTTP request, session state and timers
4062 May be used in sections : defaults | frontend | listen | backend
4063 yes | yes | yes | yes
Emeric Brun3a058f32009-06-30 18:26:00 +02004064 Arguments :
4065 clf if the "clf" argument is added, then the output format will be
4066 the CLF format instead of HAProxy's default HTTP format. You can
4067 use this when you need to feed HAProxy's logs through a specific
4068 log analyser which only support the CLF format and which is not
4069 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004070
4071 By default, the log output format is very poor, as it only contains the
4072 source and destination addresses, and the instance name. By specifying
4073 "option httplog", each log line turns into a much richer format including,
4074 but not limited to, the HTTP request, the connection timers, the session
4075 status, the connections numbers, the captured headers and cookies, the
4076 frontend, backend and server name, and of course the source address and
4077 ports.
4078
4079 This option may be set either in the frontend or the backend.
4080
Emeric Brun3a058f32009-06-30 18:26:00 +02004081 If this option has been enabled in a "defaults" section, it can be disabled
4082 in a specific instance by prepending the "no" keyword before it. Specifying
4083 only "option httplog" will automatically clear the 'clf' mode if it was set
4084 by default.
4085
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004086 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004087
Willy Tarreau55165fe2009-05-10 12:02:55 +02004088
4089option http_proxy
4090no option http_proxy
4091 Enable or disable plain HTTP proxy mode
4092 May be used in sections : defaults | frontend | listen | backend
4093 yes | yes | yes | yes
4094 Arguments : none
4095
4096 It sometimes happens that people need a pure HTTP proxy which understands
4097 basic proxy requests without caching nor any fancy feature. In this case,
4098 it may be worth setting up an HAProxy instance with the "option http_proxy"
4099 set. In this mode, no server is declared, and the connection is forwarded to
4100 the IP address and port found in the URL after the "http://" scheme.
4101
4102 No host address resolution is performed, so this only works when pure IP
4103 addresses are passed. Since this option's usage perimeter is rather limited,
4104 it will probably be used only by experts who know they need exactly it. Last,
4105 if the clients are susceptible of sending keep-alive requests, it will be
Cyril Bonté2409e682010-12-14 22:47:51 +01004106 needed to add "option httpclose" to ensure that all requests will correctly
Willy Tarreau55165fe2009-05-10 12:02:55 +02004107 be analyzed.
4108
4109 If this option has been enabled in a "defaults" section, it can be disabled
4110 in a specific instance by prepending the "no" keyword before it.
4111
4112 Example :
4113 # this backend understands HTTP proxy requests and forwards them directly.
4114 backend direct_forward
4115 option httpclose
4116 option http_proxy
4117
4118 See also : "option httpclose"
4119
Willy Tarreau211ad242009-10-03 21:45:07 +02004120
Jamie Gloudon801a0a32012-08-25 00:18:33 -04004121option independent-streams
4122no option independent-streams
4123 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02004124 May be used in sections : defaults | frontend | listen | backend
4125 yes | yes | yes | yes
4126 Arguments : none
4127
4128 By default, when data is sent over a socket, both the write timeout and the
4129 read timeout for that socket are refreshed, because we consider that there is
4130 activity on that socket, and we have no other means of guessing if we should
4131 receive data or not.
4132
4133 While this default behaviour is desirable for almost all applications, there
4134 exists a situation where it is desirable to disable it, and only refresh the
4135 read timeout if there are incoming data. This happens on sessions with large
4136 timeouts and low amounts of exchanged data such as telnet session. If the
4137 server suddenly disappears, the output data accumulates in the system's
4138 socket buffers, both timeouts are correctly refreshed, and there is no way
4139 to know the server does not receive them, so we don't timeout. However, when
4140 the underlying protocol always echoes sent data, it would be enough by itself
4141 to detect the issue using the read timeout. Note that this problem does not
4142 happen with more verbose protocols because data won't accumulate long in the
4143 socket buffers.
4144
4145 When this option is set on the frontend, it will disable read timeout updates
4146 on data sent to the client. There probably is little use of this case. When
4147 the option is set on the backend, it will disable read timeout updates on
4148 data sent to the server. Doing so will typically break large HTTP posts from
4149 slow lines, so use it with caution.
4150
Jamie Gloudon801a0a32012-08-25 00:18:33 -04004151 Note: older versions used to call this setting "option independant-streams"
4152 with a spelling mistake. This spelling is still supported but
4153 deprecated.
4154
Willy Tarreauce887fd2012-05-12 12:50:00 +02004155 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02004156
4157
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02004158option ldap-check
4159 Use LDAPv3 health checks for server testing
4160 May be used in sections : defaults | frontend | listen | backend
4161 yes | no | yes | yes
4162 Arguments : none
4163
4164 It is possible to test that the server correctly talks LDAPv3 instead of just
4165 testing that it accepts the TCP connection. When this option is set, an
4166 LDAPv3 anonymous simple bind message is sent to the server, and the response
4167 is analyzed to find an LDAPv3 bind response message.
4168
4169 The server is considered valid only when the LDAP response contains success
4170 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
4171
4172 Logging of bind requests is server dependent see your documentation how to
4173 configure it.
4174
4175 Example :
4176 option ldap-check
4177
4178 See also : "option httpchk"
4179
4180
Willy Tarreau211ad242009-10-03 21:45:07 +02004181option log-health-checks
4182no option log-health-checks
4183 Enable or disable logging of health checks
4184 May be used in sections : defaults | frontend | listen | backend
4185 yes | no | yes | yes
4186 Arguments : none
4187
4188 Enable health checks logging so it possible to check for example what
4189 was happening before a server crash. Failed health check are logged if
4190 server is UP and succeeded health checks if server is DOWN, so the amount
4191 of additional information is limited.
4192
4193 If health check logging is enabled no health check status is printed
4194 when servers is set up UP/DOWN/ENABLED/DISABLED.
4195
4196 See also: "log" and section 8 about logging.
4197
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02004198
4199option log-separate-errors
4200no option log-separate-errors
4201 Change log level for non-completely successful connections
4202 May be used in sections : defaults | frontend | listen | backend
4203 yes | yes | yes | no
4204 Arguments : none
4205
4206 Sometimes looking for errors in logs is not easy. This option makes haproxy
4207 raise the level of logs containing potentially interesting information such
4208 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
4209 level changes from "info" to "err". This makes it possible to log them
4210 separately to a different file with most syslog daemons. Be careful not to
4211 remove them from the original file, otherwise you would lose ordering which
4212 provides very important information.
4213
4214 Using this option, large sites dealing with several thousand connections per
4215 second may log normal traffic to a rotating buffer and only archive smaller
4216 error logs.
4217
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004218 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02004219 logging.
4220
Willy Tarreauc27debf2008-01-06 08:57:02 +01004221
4222option logasap
4223no option logasap
4224 Enable or disable early logging of HTTP requests
4225 May be used in sections : defaults | frontend | listen | backend
4226 yes | yes | yes | no
4227 Arguments : none
4228
4229 By default, HTTP requests are logged upon termination so that the total
4230 transfer time and the number of bytes appear in the logs. When large objects
4231 are being transferred, it may take a while before the request appears in the
4232 logs. Using "option logasap", the request gets logged as soon as the server
4233 sends the complete headers. The only missing information in the logs will be
4234 the total number of bytes which will indicate everything except the amount
4235 of data transferred, and the total time which will not take the transfer
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01004236 time into account. In such a situation, it's a good practice to capture the
Willy Tarreauc27debf2008-01-06 08:57:02 +01004237 "Content-Length" response header so that the logs at least indicate how many
4238 bytes are expected to be transferred.
4239
Willy Tarreaucc6c8912009-02-22 10:53:55 +01004240 Examples :
4241 listen http_proxy 0.0.0.0:80
4242 mode http
4243 option httplog
4244 option logasap
4245 log 192.168.2.200 local3
4246
4247 >>> Feb 6 12:14:14 localhost \
4248 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
4249 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
4250 "GET /image.iso HTTP/1.0"
4251
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004252 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01004253 logging.
4254
4255
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02004256option mysql-check [ user <username> ]
4257 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01004258 May be used in sections : defaults | frontend | listen | backend
4259 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02004260 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004261 <username> This is the username which will be used when connecting to MySQL
4262 server.
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02004263
4264 If you specify a username, the check consists of sending two MySQL packet,
4265 one Client Authentication packet, and one QUIT packet, to correctly close
4266 MySQL session. We then parse the MySQL Handshake Initialisation packet and/or
4267 Error packet. It is a basic but useful test which does not produce error nor
4268 aborted connect on the server. However, it requires adding an authorization
4269 in the MySQL table, like this :
4270
4271 USE mysql;
4272 INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');
4273 FLUSH PRIVILEGES;
4274
4275 If you don't specify a username (it is deprecated and not recommended), the
4276 check only consists in parsing the Mysql Handshake Initialisation packet or
4277 Error packet, we don't send anything in this mode. It was reported that it
4278 can generate lockout if check is too frequent and/or if there is not enough
4279 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
4280 value as if a connection is established successfully within fewer than MySQL
4281 "max_connect_errors" attempts after a previous connection was interrupted,
4282 the error count for the host is cleared to zero. If HAProxy's server get
4283 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
4284
4285 Remember that this does not check database presence nor database consistency.
4286 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01004287
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02004288 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01004289
4290 Most often, an incoming MySQL server needs to see the client's IP address for
4291 various purposes, including IP privilege matching and connection logging.
4292 When possible, it is often wise to masquerade the client's IP address when
4293 connecting to the server using the "usesrc" argument of the "source" keyword,
4294 which requires the cttproxy feature to be compiled in, and the MySQL server
4295 to route the client via the machine hosting haproxy.
4296
4297 See also: "option httpchk"
4298
4299
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004300option nolinger
4301no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004302 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004303 May be used in sections: defaults | frontend | listen | backend
4304 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01004305 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004306
4307 When clients or servers abort connections in a dirty way (eg: they are
4308 physically disconnected), the session timeouts triggers and the session is
4309 closed. But it will remain in FIN_WAIT1 state for some time in the system,
4310 using some resources and possibly limiting the ability to establish newer
4311 connections.
4312
4313 When this happens, it is possible to activate "option nolinger" which forces
4314 the system to immediately remove any socket's pending data on close. Thus,
4315 the session is instantly purged from the system's tables. This usually has
4316 side effects such as increased number of TCP resets due to old retransmits
4317 getting immediately rejected. Some firewalls may sometimes complain about
4318 this too.
4319
4320 For this reason, it is not recommended to use this option when not absolutely
4321 needed. You know that you need it when you have thousands of FIN_WAIT1
4322 sessions on your system (TIME_WAIT ones do not count).
4323
4324 This option may be used both on frontends and backends, depending on the side
4325 where it is required. Use it on the frontend for clients, and on the backend
4326 for servers.
4327
4328 If this option has been enabled in a "defaults" section, it can be disabled
4329 in a specific instance by prepending the "no" keyword before it.
4330
4331
Willy Tarreau55165fe2009-05-10 12:02:55 +02004332option originalto [ except <network> ] [ header <name> ]
4333 Enable insertion of the X-Original-To header to requests sent to servers
4334 May be used in sections : defaults | frontend | listen | backend
4335 yes | yes | yes | yes
4336 Arguments :
4337 <network> is an optional argument used to disable this option for sources
4338 matching <network>
4339 <name> an optional argument to specify a different "X-Original-To"
4340 header name.
4341
4342 Since HAProxy can work in transparent mode, every request from a client can
4343 be redirected to the proxy and HAProxy itself can proxy every request to a
4344 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
4345 be lost. This is annoying when you want access rules based on destination ip
4346 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
4347 added by HAProxy to all requests sent to the server. This header contains a
4348 value representing the original destination IP address. Since this must be
4349 configured to always use the last occurrence of this header only. Note that
4350 only the last occurrence of the header must be used, since it is really
4351 possible that the client has already brought one.
4352
4353 The keyword "header" may be used to supply a different header name to replace
4354 the default "X-Original-To". This can be useful where you might already
4355 have a "X-Original-To" header from a different application, and you need
4356 preserve it. Also if your backend server doesn't use the "X-Original-To"
4357 header and requires different one.
4358
4359 Sometimes, a same HAProxy instance may be shared between a direct client
4360 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
4361 used to decrypt HTTPS traffic). It is possible to disable the addition of the
4362 header for a known source address or network by adding the "except" keyword
4363 followed by the network address. In this case, any source IP matching the
4364 network will not cause an addition of this header. Most common uses are with
4365 private networks or 127.0.0.1.
4366
4367 This option may be specified either in the frontend or in the backend. If at
4368 least one of them uses it, the header will be added. Note that the backend's
4369 setting of the header subargument takes precedence over the frontend's if
4370 both are defined.
4371
Willy Tarreau87cf5142011-08-19 22:57:24 +02004372 It is important to note that by default, HAProxy works in tunnel mode and
4373 only inspects the first request of a connection, meaning that only the first
4374 request will have the header appended, which is certainly not what you want.
4375 In order to fix this, ensure that any of the "httpclose", "forceclose" or
4376 "http-server-close" options is set when using this option.
Willy Tarreau55165fe2009-05-10 12:02:55 +02004377
4378 Examples :
4379 # Original Destination address
4380 frontend www
4381 mode http
4382 option originalto except 127.0.0.1
4383
4384 # Those servers want the IP Address in X-Client-Dst
4385 backend www
4386 mode http
4387 option originalto header X-Client-Dst
4388
Willy Tarreau87cf5142011-08-19 22:57:24 +02004389 See also : "option httpclose", "option http-server-close",
4390 "option forceclose"
Willy Tarreau55165fe2009-05-10 12:02:55 +02004391
4392
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004393option persist
4394no option persist
4395 Enable or disable forced persistence on down servers
4396 May be used in sections: defaults | frontend | listen | backend
4397 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01004398 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004399
4400 When an HTTP request reaches a backend with a cookie which references a dead
4401 server, by default it is redispatched to another server. It is possible to
4402 force the request to be sent to the dead server first using "option persist"
4403 if absolutely needed. A common use case is when servers are under extreme
4404 load and spend their time flapping. In this case, the users would still be
4405 directed to the server they opened the session on, in the hope they would be
4406 correctly served. It is recommended to use "option redispatch" in conjunction
4407 with this option so that in the event it would not be possible to connect to
4408 the server at all (server definitely dead), the client would finally be
4409 redirected to another valid server.
4410
4411 If this option has been enabled in a "defaults" section, it can be disabled
4412 in a specific instance by prepending the "no" keyword before it.
4413
Willy Tarreau4de91492010-01-22 19:10:05 +01004414 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004415
4416
Willy Tarreau0c122822013-12-15 18:49:01 +01004417option pgsql-check [ user <username> ]
4418 Use PostgreSQL health checks for server testing
4419 May be used in sections : defaults | frontend | listen | backend
4420 yes | no | yes | yes
4421 Arguments :
4422 <username> This is the username which will be used when connecting to
4423 PostgreSQL server.
4424
4425 The check sends a PostgreSQL StartupMessage and waits for either
4426 Authentication request or ErrorResponse message. It is a basic but useful
4427 test which does not produce error nor aborted connect on the server.
4428 This check is identical with the "mysql-check".
4429
4430 See also: "option httpchk"
4431
4432
Willy Tarreau9420b122013-12-15 18:58:25 +01004433option prefer-last-server
4434no option prefer-last-server
4435 Allow multiple load balanced requests to remain on the same server
4436 May be used in sections: defaults | frontend | listen | backend
4437 yes | no | yes | yes
4438 Arguments : none
4439
4440 When the load balancing algorithm in use is not deterministic, and a previous
4441 request was sent to a server to which haproxy still holds a connection, it is
4442 sometimes desirable that subsequent requests on a same session go to the same
4443 server as much as possible. Note that this is different from persistence, as
4444 we only indicate a preference which haproxy tries to apply without any form
4445 of warranty. The real use is for keep-alive connections sent to servers. When
4446 this option is used, haproxy will try to reuse the same connection that is
4447 attached to the server instead of rebalancing to another server, causing a
4448 close of the connection. This can make sense for static file servers. It does
4449 not make much sense to use this in combination with hashing algorithms.
4450
4451 If this option has been enabled in a "defaults" section, it can be disabled
4452 in a specific instance by prepending the "no" keyword before it.
4453
4454 See also: "option http-keep-alive"
4455
4456
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004457option redispatch
4458no option redispatch
4459 Enable or disable session redistribution in case of connection failure
4460 May be used in sections: defaults | frontend | listen | backend
4461 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01004462 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004463
4464 In HTTP mode, if a server designated by a cookie is down, clients may
4465 definitely stick to it because they cannot flush the cookie, so they will not
4466 be able to access the service anymore.
4467
4468 Specifying "option redispatch" will allow the proxy to break their
4469 persistence and redistribute them to a working server.
4470
4471 It also allows to retry last connection to another server in case of multiple
4472 connection failures. Of course, it requires having "retries" set to a nonzero
4473 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01004474
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004475 This form is the preferred form, which replaces both the "redispatch" and
4476 "redisp" keywords.
4477
4478 If this option has been enabled in a "defaults" section, it can be disabled
4479 in a specific instance by prepending the "no" keyword before it.
4480
Willy Tarreau4de91492010-01-22 19:10:05 +01004481 See also : "redispatch", "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004482
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004483
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02004484option redis-check
4485 Use redis health checks for server testing
4486 May be used in sections : defaults | frontend | listen | backend
4487 yes | no | yes | yes
4488 Arguments : none
4489
4490 It is possible to test that the server correctly talks REDIS protocol instead
4491 of just testing that it accepts the TCP connection. When this option is set,
4492 a PING redis command is sent to the server, and the response is analyzed to
4493 find the "+PONG" response message.
4494
4495 Example :
4496 option redis-check
4497
4498 See also : "option httpchk"
4499
4500
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004501option smtpchk
4502option smtpchk <hello> <domain>
4503 Use SMTP health checks for server testing
4504 May be used in sections : defaults | frontend | listen | backend
4505 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01004506 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004507 <hello> is an optional argument. It is the "hello" command to use. It can
4508 be either "HELO" (for SMTP) or "EHLO" (for ESTMP). All other
4509 values will be turned into the default command ("HELO").
4510
4511 <domain> is the domain name to present to the server. It may only be
4512 specified (and is mandatory) if the hello command has been
4513 specified. By default, "localhost" is used.
4514
4515 When "option smtpchk" is set, the health checks will consist in TCP
4516 connections followed by an SMTP command. By default, this command is
4517 "HELO localhost". The server's return code is analyzed and only return codes
4518 starting with a "2" will be considered as valid. All other responses,
4519 including a lack of response will constitute an error and will indicate a
4520 dead server.
4521
4522 This test is meant to be used with SMTP servers or relays. Depending on the
4523 request, it is possible that some servers do not log each connection attempt,
4524 so you may want to experiment to improve the behaviour. Using telnet on port
4525 25 is often easier than adjusting the configuration.
4526
4527 Most often, an incoming SMTP server needs to see the client's IP address for
4528 various purposes, including spam filtering, anti-spoofing and logging. When
4529 possible, it is often wise to masquerade the client's IP address when
4530 connecting to the server using the "usesrc" argument of the "source" keyword,
4531 which requires the cttproxy feature to be compiled in.
4532
4533 Example :
4534 option smtpchk HELO mydomain.org
4535
4536 See also : "option httpchk", "source"
4537
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004538
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02004539option socket-stats
4540no option socket-stats
4541
4542 Enable or disable collecting & providing separate statistics for each socket.
4543 May be used in sections : defaults | frontend | listen | backend
4544 yes | yes | yes | no
4545
4546 Arguments : none
4547
4548
Willy Tarreauff4f82d2009-02-06 11:28:13 +01004549option splice-auto
4550no option splice-auto
4551 Enable or disable automatic kernel acceleration on sockets in both directions
4552 May be used in sections : defaults | frontend | listen | backend
4553 yes | yes | yes | yes
4554 Arguments : none
4555
4556 When this option is enabled either on a frontend or on a backend, haproxy
4557 will automatically evaluate the opportunity to use kernel tcp splicing to
4558 forward data between the client and the server, in either direction. Haproxy
4559 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004560 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01004561 are not much aggressive in order to limit excessive use of splicing. This
4562 option requires splicing to be enabled at compile time, and may be globally
4563 disabled with the global option "nosplice". Since splice uses pipes, using it
4564 requires that there are enough spare pipes.
4565
4566 Important note: kernel-based TCP splicing is a Linux-specific feature which
4567 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
4568 transfer data between sockets without copying these data to user-space, thus
4569 providing noticeable performance gains and CPU cycles savings. Since many
4570 early implementations are buggy, corrupt data and/or are inefficient, this
4571 feature is not enabled by default, and it should be used with extreme care.
4572 While it is not possible to detect the correctness of an implementation,
4573 2.6.29 is the first version offering a properly working implementation. In
4574 case of doubt, splicing may be globally disabled using the global "nosplice"
4575 keyword.
4576
4577 Example :
4578 option splice-auto
4579
4580 If this option has been enabled in a "defaults" section, it can be disabled
4581 in a specific instance by prepending the "no" keyword before it.
4582
4583 See also : "option splice-request", "option splice-response", and global
4584 options "nosplice" and "maxpipes"
4585
4586
4587option splice-request
4588no option splice-request
4589 Enable or disable automatic kernel acceleration on sockets for requests
4590 May be used in sections : defaults | frontend | listen | backend
4591 yes | yes | yes | yes
4592 Arguments : none
4593
4594 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004595 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01004596 the client to the server. It might still use the recv/send scheme if there
4597 are no spare pipes left. This option requires splicing to be enabled at
4598 compile time, and may be globally disabled with the global option "nosplice".
4599 Since splice uses pipes, using it requires that there are enough spare pipes.
4600
4601 Important note: see "option splice-auto" for usage limitations.
4602
4603 Example :
4604 option splice-request
4605
4606 If this option has been enabled in a "defaults" section, it can be disabled
4607 in a specific instance by prepending the "no" keyword before it.
4608
4609 See also : "option splice-auto", "option splice-response", and global options
4610 "nosplice" and "maxpipes"
4611
4612
4613option splice-response
4614no option splice-response
4615 Enable or disable automatic kernel acceleration on sockets for responses
4616 May be used in sections : defaults | frontend | listen | backend
4617 yes | yes | yes | yes
4618 Arguments : none
4619
4620 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004621 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01004622 the server to the client. It might still use the recv/send scheme if there
4623 are no spare pipes left. This option requires splicing to be enabled at
4624 compile time, and may be globally disabled with the global option "nosplice".
4625 Since splice uses pipes, using it requires that there are enough spare pipes.
4626
4627 Important note: see "option splice-auto" for usage limitations.
4628
4629 Example :
4630 option splice-response
4631
4632 If this option has been enabled in a "defaults" section, it can be disabled
4633 in a specific instance by prepending the "no" keyword before it.
4634
4635 See also : "option splice-auto", "option splice-request", and global options
4636 "nosplice" and "maxpipes"
4637
4638
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004639option srvtcpka
4640no option srvtcpka
4641 Enable or disable the sending of TCP keepalive packets on the server side
4642 May be used in sections : defaults | frontend | listen | backend
4643 yes | no | yes | yes
4644 Arguments : none
4645
4646 When there is a firewall or any session-aware component between a client and
4647 a server, and when the protocol involves very long sessions with long idle
4648 periods (eg: remote desktops), there is a risk that one of the intermediate
4649 components decides to expire a session which has remained idle for too long.
4650
4651 Enabling socket-level TCP keep-alives makes the system regularly send packets
4652 to the other end of the connection, leaving it active. The delay between
4653 keep-alive probes is controlled by the system only and depends both on the
4654 operating system and its tuning parameters.
4655
4656 It is important to understand that keep-alive packets are neither emitted nor
4657 received at the application level. It is only the network stacks which sees
4658 them. For this reason, even if one side of the proxy already uses keep-alives
4659 to maintain its connection alive, those keep-alive packets will not be
4660 forwarded to the other side of the proxy.
4661
4662 Please note that this has nothing to do with HTTP keep-alive.
4663
4664 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
4665 server side of a connection, which should help when session expirations are
4666 noticed between HAProxy and a server.
4667
4668 If this option has been enabled in a "defaults" section, it can be disabled
4669 in a specific instance by prepending the "no" keyword before it.
4670
4671 See also : "option clitcpka", "option tcpka"
4672
4673
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004674option ssl-hello-chk
4675 Use SSLv3 client hello health checks for server testing
4676 May be used in sections : defaults | frontend | listen | backend
4677 yes | no | yes | yes
4678 Arguments : none
4679
4680 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
4681 possible to test that the server correctly talks SSL instead of just testing
4682 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
4683 SSLv3 client hello messages are sent once the connection is established to
4684 the server, and the response is analyzed to find an SSL server hello message.
4685 The server is considered valid only when the response contains this server
4686 hello message.
4687
4688 All servers tested till there correctly reply to SSLv3 client hello messages,
4689 and most servers tested do not even log the requests containing only hello
4690 messages, which is appreciable.
4691
Willy Tarreau763a95b2012-10-04 23:15:39 +02004692 Note that this check works even when SSL support was not built into haproxy
4693 because it forges the SSL message. When SSL support is available, it is best
4694 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004695
Willy Tarreau763a95b2012-10-04 23:15:39 +02004696 See also: "option httpchk", "check-ssl"
4697
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004698
Willy Tarreaued179852013-12-16 01:07:00 +01004699option tcp-check
4700 Perform health checks using tcp-check send/expect sequences
4701 May be used in sections: defaults | frontend | listen | backend
4702 yes | no | yes | yes
4703
4704 This health check method is intended to be combined with "tcp-check" command
4705 lists in order to support send/expect types of health check sequences.
4706
4707 TCP checks currently support 4 modes of operations :
4708 - no "tcp-check" directive : the health check only consists in a connection
4709 attempt, which remains the default mode.
4710
4711 - "tcp-check send" or "tcp-check send-binary" only is mentionned : this is
4712 used to send a string along with a connection opening. With some
4713 protocols, it helps sending a "QUIT" message for example that prevents
4714 the server from logging a connection error for each health check. The
4715 check result will still be based on the ability to open the connection
4716 only.
4717
4718 - "tcp-check expect" only is mentionned : this is used to test a banner.
4719 The connection is opened and haproxy waits for the server to present some
4720 contents which must validate some rules. The check result will be based
4721 on the matching between the contents and the rules. This is suited for
4722 POP, IMAP, SMTP, FTP, SSH, TELNET.
4723
4724 - both "tcp-check send" and "tcp-check expect" are mentionned : this is
4725 used to test a hello-type protocol. Haproxy sends a message, the server
4726 responds and its response is analysed. the check result will be based on
4727 the maching between the response contents and the rules. This is often
4728 suited for protocols which require a binding or a request/response model.
4729 LDAP, MySQL, Redis and SSL are example of such protocols, though they
4730 already all have their dedicated checks with a deeper understanding of
4731 the respective protocols.
4732 In this mode, many questions may be sent and many answers may be
4733 analysed.
4734
4735 Examples :
4736 # perform a POP check (analyse only server's banner)
4737 option tcp-check
4738 tcp-check expect string +OK\ POP3\ ready
4739
4740 # perform an IMAP check (analyse only server's banner)
4741 option tcp-check
4742 tcp-check expect string *\ OK\ IMAP4\ ready
4743
4744 # look for the redis master server after ensuring it speaks well
4745 # redis protocol, then it exits properly.
4746 # (send a command then analyse the response 3 tims)
4747 option tcp-check
4748 tcp-check send PING\r\n
4749 tcp-check expect +PONG
4750 tcp-check send info\ replication\r\n
4751 tcp-check expect string role:master
4752 tcp-check send QUIT\r\n
4753 tcp-check expect string +OK
4754
4755 forge a HTTP request, then analyse the response
4756 (send many headers before analyzing)
4757 option tcp-check
4758 tcp-check send HEAD\ /\ HTTP/1.1\r\n
4759 tcp-check send Host:\ www.mydomain.com\r\n
4760 tcp-check send User-Agent:\ HAProxy\ tcpcheck\r\n
4761 tcp-check send \r\n
4762 tcp-check expect rstring HTTP/1\..\ (2..|3..)
4763
4764
4765 See also : "tcp-check expect", "tcp-check send"
4766
4767
Willy Tarreau9ea05a72009-06-14 12:07:01 +02004768option tcp-smart-accept
4769no option tcp-smart-accept
4770 Enable or disable the saving of one ACK packet during the accept sequence
4771 May be used in sections : defaults | frontend | listen | backend
4772 yes | yes | yes | no
4773 Arguments : none
4774
4775 When an HTTP connection request comes in, the system acknowledges it on
4776 behalf of HAProxy, then the client immediately sends its request, and the
4777 system acknowledges it too while it is notifying HAProxy about the new
4778 connection. HAProxy then reads the request and responds. This means that we
4779 have one TCP ACK sent by the system for nothing, because the request could
4780 very well be acknowledged by HAProxy when it sends its response.
4781
4782 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
4783 sending this useless ACK on platforms which support it (currently at least
4784 Linux). It must not cause any problem, because the system will send it anyway
4785 after 40 ms if the response takes more time than expected to come.
4786
4787 During complex network debugging sessions, it may be desirable to disable
4788 this optimization because delayed ACKs can make troubleshooting more complex
4789 when trying to identify where packets are delayed. It is then possible to
4790 fall back to normal behaviour by specifying "no option tcp-smart-accept".
4791
4792 It is also possible to force it for non-HTTP proxies by simply specifying
4793 "option tcp-smart-accept". For instance, it can make sense with some services
4794 such as SMTP where the server speaks first.
4795
4796 It is recommended to avoid forcing this option in a defaults section. In case
4797 of doubt, consider setting it back to automatic values by prepending the
4798 "default" keyword before it, or disabling it using the "no" keyword.
4799
Willy Tarreaud88edf22009-06-14 15:48:17 +02004800 See also : "option tcp-smart-connect"
4801
4802
4803option tcp-smart-connect
4804no option tcp-smart-connect
4805 Enable or disable the saving of one ACK packet during the connect sequence
4806 May be used in sections : defaults | frontend | listen | backend
4807 yes | no | yes | yes
4808 Arguments : none
4809
4810 On certain systems (at least Linux), HAProxy can ask the kernel not to
4811 immediately send an empty ACK upon a connection request, but to directly
4812 send the buffer request instead. This saves one packet on the network and
4813 thus boosts performance. It can also be useful for some servers, because they
4814 immediately get the request along with the incoming connection.
4815
4816 This feature is enabled when "option tcp-smart-connect" is set in a backend.
4817 It is not enabled by default because it makes network troubleshooting more
4818 complex.
4819
4820 It only makes sense to enable it with protocols where the client speaks first
4821 such as HTTP. In other situations, if there is no data to send in place of
4822 the ACK, a normal ACK is sent.
4823
4824 If this option has been enabled in a "defaults" section, it can be disabled
4825 in a specific instance by prepending the "no" keyword before it.
4826
4827 See also : "option tcp-smart-accept"
4828
Willy Tarreau9ea05a72009-06-14 12:07:01 +02004829
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004830option tcpka
4831 Enable or disable the sending of TCP keepalive packets on both sides
4832 May be used in sections : defaults | frontend | listen | backend
4833 yes | yes | yes | yes
4834 Arguments : none
4835
4836 When there is a firewall or any session-aware component between a client and
4837 a server, and when the protocol involves very long sessions with long idle
4838 periods (eg: remote desktops), there is a risk that one of the intermediate
4839 components decides to expire a session which has remained idle for too long.
4840
4841 Enabling socket-level TCP keep-alives makes the system regularly send packets
4842 to the other end of the connection, leaving it active. The delay between
4843 keep-alive probes is controlled by the system only and depends both on the
4844 operating system and its tuning parameters.
4845
4846 It is important to understand that keep-alive packets are neither emitted nor
4847 received at the application level. It is only the network stacks which sees
4848 them. For this reason, even if one side of the proxy already uses keep-alives
4849 to maintain its connection alive, those keep-alive packets will not be
4850 forwarded to the other side of the proxy.
4851
4852 Please note that this has nothing to do with HTTP keep-alive.
4853
4854 Using option "tcpka" enables the emission of TCP keep-alive probes on both
4855 the client and server sides of a connection. Note that this is meaningful
4856 only in "defaults" or "listen" sections. If this option is used in a
4857 frontend, only the client side will get keep-alives, and if this option is
4858 used in a backend, only the server side will get keep-alives. For this
4859 reason, it is strongly recommended to explicitly use "option clitcpka" and
4860 "option srvtcpka" when the configuration is split between frontends and
4861 backends.
4862
4863 See also : "option clitcpka", "option srvtcpka"
4864
Willy Tarreau844e3c52008-01-11 16:28:18 +01004865
4866option tcplog
4867 Enable advanced logging of TCP connections with session state and timers
4868 May be used in sections : defaults | frontend | listen | backend
4869 yes | yes | yes | yes
4870 Arguments : none
4871
4872 By default, the log output format is very poor, as it only contains the
4873 source and destination addresses, and the instance name. By specifying
4874 "option tcplog", each log line turns into a much richer format including, but
4875 not limited to, the connection timers, the session status, the connections
4876 numbers, the frontend, backend and server name, and of course the source
4877 address and ports. This option is useful for pure TCP proxies in order to
4878 find which of the client or server disconnects or times out. For normal HTTP
4879 proxies, it's better to use "option httplog" which is even more complete.
4880
4881 This option may be set either in the frontend or the backend.
4882
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004883 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01004884
4885
Willy Tarreau844e3c52008-01-11 16:28:18 +01004886option transparent
4887no option transparent
4888 Enable client-side transparent proxying
4889 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01004890 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01004891 Arguments : none
4892
4893 This option was introduced in order to provide layer 7 persistence to layer 3
4894 load balancers. The idea is to use the OS's ability to redirect an incoming
4895 connection for a remote address to a local process (here HAProxy), and let
4896 this process know what address was initially requested. When this option is
4897 used, sessions without cookies will be forwarded to the original destination
4898 IP address of the incoming request (which should match that of another
4899 equipment), while requests with cookies will still be forwarded to the
4900 appropriate server.
4901
4902 Note that contrary to a common belief, this option does NOT make HAProxy
4903 present the client's IP to the server when establishing the connection.
4904
Willy Tarreaua1146052011-03-01 09:51:54 +01004905 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01004906 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01004907
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004908
Emeric Brun647caf12009-06-30 17:57:00 +02004909persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02004910persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02004911 Enable RDP cookie-based persistence
4912 May be used in sections : defaults | frontend | listen | backend
4913 yes | no | yes | yes
4914 Arguments :
4915 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02004916 default cookie name "msts" will be used. There currently is no
4917 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02004918
4919 This statement enables persistence based on an RDP cookie. The RDP cookie
4920 contains all information required to find the server in the list of known
4921 servers. So when this option is set in the backend, the request is analysed
4922 and if an RDP cookie is found, it is decoded. If it matches a known server
4923 which is still UP (or if "option persist" is set), then the connection is
4924 forwarded to this server.
4925
4926 Note that this only makes sense in a TCP backend, but for this to work, the
4927 frontend must have waited long enough to ensure that an RDP cookie is present
4928 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004929 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02004930 a single "listen" section.
4931
Willy Tarreau61e28f22010-05-16 22:31:05 +02004932 Also, it is important to understand that the terminal server will emit this
4933 RDP cookie only if it is configured for "token redirection mode", which means
4934 that the "IP address redirection" option is disabled.
4935
Emeric Brun647caf12009-06-30 17:57:00 +02004936 Example :
4937 listen tse-farm
4938 bind :3389
4939 # wait up to 5s for an RDP cookie in the request
4940 tcp-request inspect-delay 5s
4941 tcp-request content accept if RDP_COOKIE
4942 # apply RDP cookie persistence
4943 persist rdp-cookie
4944 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02004945 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02004946 balance rdp-cookie
4947 server srv1 1.1.1.1:3389
4948 server srv2 1.1.1.2:3389
4949
Simon Hormanab814e02011-06-24 14:50:20 +09004950 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
4951 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02004952
4953
Willy Tarreau3a7d2072009-03-05 23:48:25 +01004954rate-limit sessions <rate>
4955 Set a limit on the number of new sessions accepted per second on a frontend
4956 May be used in sections : defaults | frontend | listen | backend
4957 yes | yes | yes | no
4958 Arguments :
4959 <rate> The <rate> parameter is an integer designating the maximum number
4960 of new sessions per second to accept on the frontend.
4961
4962 When the frontend reaches the specified number of new sessions per second, it
4963 stops accepting new connections until the rate drops below the limit again.
4964 During this time, the pending sessions will be kept in the socket's backlog
4965 (in system buffers) and haproxy will not even be aware that sessions are
4966 pending. When applying very low limit on a highly loaded service, it may make
4967 sense to increase the socket's backlog using the "backlog" keyword.
4968
4969 This feature is particularly efficient at blocking connection-based attacks
4970 or service abuse on fragile servers. Since the session rate is measured every
4971 millisecond, it is extremely accurate. Also, the limit applies immediately,
4972 no delay is needed at all to detect the threshold.
4973
4974 Example : limit the connection rate on SMTP to 10 per second max
4975 listen smtp
4976 mode tcp
4977 bind :25
4978 rate-limit sessions 10
4979 server 127.0.0.1:1025
4980
Willy Tarreaua17c2d92011-07-25 08:16:20 +02004981 Note : when the maximum rate is reached, the frontend's status is not changed
4982 but its sockets appear as "WAITING" in the statistics if the
4983 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +01004984
4985 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
4986
4987
Willy Tarreau2e1dca82012-09-12 08:43:15 +02004988redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
4989redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
4990redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +02004991 Return an HTTP redirection if/unless a condition is matched
4992 May be used in sections : defaults | frontend | listen | backend
4993 no | yes | yes | yes
4994
4995 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +01004996 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02004997
Willy Tarreau0140f252008-11-19 21:07:09 +01004998 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +02004999 <loc> With "redirect location", the exact value in <loc> is placed into
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01005000 the HTTP "Location" header. When used in an "http-request" rule,
5001 <loc> value follows the log-format rules and can include some
5002 dynamic values (see Custom Log Format in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02005003
5004 <pfx> With "redirect prefix", the "Location" header is built from the
5005 concatenation of <pfx> and the complete URI path, including the
5006 query string, unless the "drop-query" option is specified (see
5007 below). As a special case, if <pfx> equals exactly "/", then
5008 nothing is inserted before the original URI. It allows one to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01005009 redirect to the same URL (for instance, to insert a cookie). When
5010 used in an "http-request" rule, <pfx> value follows the log-format
5011 rules and can include some dynamic values (see Custom Log Format
5012 in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02005013
5014 <sch> With "redirect scheme", then the "Location" header is built by
5015 concatenating <sch> with "://" then the first occurrence of the
5016 "Host" header, and then the URI path, including the query string
5017 unless the "drop-query" option is specified (see below). If no
5018 path is found or if the path is "*", then "/" is used instead. If
5019 no "Host" header is found, then an empty host component will be
5020 returned, which most recent browsers interprete as redirecting to
5021 the same host. This directive is mostly used to redirect HTTP to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01005022 HTTPS. When used in an "http-request" rule, <sch> value follows
5023 the log-format rules and can include some dynamic values (see
5024 Custom Log Format in section 8.2.4).
Willy Tarreau0140f252008-11-19 21:07:09 +01005025
5026 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +01005027 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
5028 with 302 used by default if no code is specified. 301 means
5029 "Moved permanently", and a browser may cache the Location. 302
5030 means "Moved permanently" and means that the browser should not
5031 cache the redirection. 303 is equivalent to 302 except that the
5032 browser will fetch the location with a GET method. 307 is just
5033 like 302 but makes it clear that the same method must be reused.
5034 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +01005035
5036 <option> There are several options which can be specified to adjust the
5037 expected behaviour of a redirection :
5038
5039 - "drop-query"
5040 When this keyword is used in a prefix-based redirection, then the
5041 location will be set without any possible query-string, which is useful
5042 for directing users to a non-secure page for instance. It has no effect
5043 with a location-type redirect.
5044
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01005045 - "append-slash"
5046 This keyword may be used in conjunction with "drop-query" to redirect
5047 users who use a URL not ending with a '/' to the same one with the '/'.
5048 It can be useful to ensure that search engines will only see one URL.
5049 For this, a return code 301 is preferred.
5050
Willy Tarreau0140f252008-11-19 21:07:09 +01005051 - "set-cookie NAME[=value]"
5052 A "Set-Cookie" header will be added with NAME (and optionally "=value")
5053 to the response. This is sometimes used to indicate that a user has
5054 been seen, for instance to protect against some types of DoS. No other
5055 cookie option is added, so the cookie will be a session cookie. Note
5056 that for a browser, a sole cookie name without an equal sign is
5057 different from a cookie with an equal sign.
5058
5059 - "clear-cookie NAME[=]"
5060 A "Set-Cookie" header will be added with NAME (and optionally "="), but
5061 with the "Max-Age" attribute set to zero. This will tell the browser to
5062 delete this cookie. It is useful for instance on logout pages. It is
5063 important to note that clearing the cookie "NAME" will not remove a
5064 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
5065 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02005066
5067 Example: move the login URL only to HTTPS.
5068 acl clear dst_port 80
5069 acl secure dst_port 8080
5070 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +01005071 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +01005072 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +01005073 acl cookie_set hdr_sub(cookie) SEEN=1
5074
5075 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +01005076 redirect prefix https://mysite.com if login_page !secure
5077 redirect prefix http://mysite.com drop-query if login_page !uid_given
5078 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +01005079 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +02005080
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01005081 Example: send redirects for request for articles without a '/'.
5082 acl missing_slash path_reg ^/article/[^/]*$
5083 redirect code 301 prefix / drop-query append-slash if missing_slash
5084
Willy Tarreau2e1dca82012-09-12 08:43:15 +02005085 Example: redirect all HTTP traffic to HTTPS when SSL is handled by haproxy.
David BERARDe7153042012-11-03 00:11:31 +01005086 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +02005087
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01005088 Example: append 'www.' prefix in front of all hosts not having it
5089 http-request redirect code 301 location www.%[hdr(host)]%[req.uri] \
5090 unless { hdr_beg(host) -i www }
5091
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005092 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02005093
5094
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005095redisp (deprecated)
5096redispatch (deprecated)
5097 Enable or disable session redistribution in case of connection failure
5098 May be used in sections: defaults | frontend | listen | backend
5099 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005100 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005101
5102 In HTTP mode, if a server designated by a cookie is down, clients may
5103 definitely stick to it because they cannot flush the cookie, so they will not
5104 be able to access the service anymore.
5105
5106 Specifying "redispatch" will allow the proxy to break their persistence and
5107 redistribute them to a working server.
5108
5109 It also allows to retry last connection to another server in case of multiple
5110 connection failures. Of course, it requires having "retries" set to a nonzero
5111 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01005112
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005113 This form is deprecated, do not use it in any new configuration, use the new
5114 "option redispatch" instead.
5115
5116 See also : "option redispatch"
5117
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005118
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01005119reqadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01005120 Add a header at the end of the HTTP request
5121 May be used in sections : defaults | frontend | listen | backend
5122 no | yes | yes | yes
5123 Arguments :
5124 <string> is the complete line to be added. Any space or known delimiter
5125 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005126 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01005127
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01005128 <cond> is an optional matching condition built from ACLs. It makes it
5129 possible to ignore this rule when other conditions are not met.
5130
Willy Tarreau303c0352008-01-17 19:01:39 +01005131 A new line consisting in <string> followed by a line feed will be added after
5132 the last header of an HTTP request.
5133
5134 Header transformations only apply to traffic which passes through HAProxy,
5135 and not to traffic generated by HAProxy, such as health-checks or error
5136 responses.
5137
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01005138 Example : add "X-Proto: SSL" to requests coming via port 81
5139 acl is-ssl dst_port 81
5140 reqadd X-Proto:\ SSL if is-ssl
5141
5142 See also: "rspadd", section 6 about HTTP header manipulation, and section 7
5143 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005144
5145
Willy Tarreau5321c422010-01-28 20:35:13 +01005146reqallow <search> [{if | unless} <cond>]
5147reqiallow <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005148 Definitely allow an HTTP request if a line matches a regular expression
5149 May be used in sections : defaults | frontend | listen | backend
5150 no | yes | yes | yes
5151 Arguments :
5152 <search> is the regular expression applied to HTTP headers and to the
5153 request line. This is an extended regular expression. Parenthesis
5154 grouping is supported and no preliminary backslash is required.
5155 Any space or known delimiter must be escaped using a backslash
5156 ('\'). The pattern applies to a full line at a time. The
5157 "reqallow" keyword strictly matches case while "reqiallow"
5158 ignores case.
5159
Willy Tarreau5321c422010-01-28 20:35:13 +01005160 <cond> is an optional matching condition built from ACLs. It makes it
5161 possible to ignore this rule when other conditions are not met.
5162
Willy Tarreau303c0352008-01-17 19:01:39 +01005163 A request containing any line which matches extended regular expression
5164 <search> will mark the request as allowed, even if any later test would
5165 result in a deny. The test applies both to the request line and to request
5166 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01005167 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01005168
5169 It is easier, faster and more powerful to use ACLs to write access policies.
5170 Reqdeny, reqallow and reqpass should be avoided in new designs.
5171
5172 Example :
5173 # allow www.* but refuse *.local
5174 reqiallow ^Host:\ www\.
5175 reqideny ^Host:\ .*\.local
5176
Willy Tarreau5321c422010-01-28 20:35:13 +01005177 See also: "reqdeny", "block", section 6 about HTTP header manipulation, and
5178 section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005179
5180
Willy Tarreau5321c422010-01-28 20:35:13 +01005181reqdel <search> [{if | unless} <cond>]
5182reqidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005183 Delete all headers matching a regular expression in an HTTP request
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 request line. This is an extended regular expression. Parenthesis
5189 grouping is supported and no preliminary backslash is required.
5190 Any space or known delimiter must be escaped using a backslash
5191 ('\'). The pattern applies to a full line at a time. The "reqdel"
5192 keyword strictly matches case while "reqidel" ignores case.
5193
Willy Tarreau5321c422010-01-28 20:35:13 +01005194 <cond> is an optional matching condition built from ACLs. It makes it
5195 possible to ignore this rule when other conditions are not met.
5196
Willy Tarreau303c0352008-01-17 19:01:39 +01005197 Any header line matching extended regular expression <search> in the request
5198 will be completely deleted. Most common use of this is to remove unwanted
5199 and/or dangerous headers or cookies from a request before passing it to the
5200 next servers.
5201
5202 Header transformations only apply to traffic which passes through HAProxy,
5203 and not to traffic generated by HAProxy, such as health-checks or error
5204 responses. Keep in mind that header names are not case-sensitive.
5205
5206 Example :
5207 # remove X-Forwarded-For header and SERVER cookie
5208 reqidel ^X-Forwarded-For:.*
5209 reqidel ^Cookie:.*SERVER=
5210
Willy Tarreau5321c422010-01-28 20:35:13 +01005211 See also: "reqadd", "reqrep", "rspdel", section 6 about HTTP header
5212 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005213
5214
Willy Tarreau5321c422010-01-28 20:35:13 +01005215reqdeny <search> [{if | unless} <cond>]
5216reqideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005217 Deny an HTTP request if a line matches a regular expression
5218 May be used in sections : defaults | frontend | listen | backend
5219 no | yes | yes | yes
5220 Arguments :
5221 <search> is the regular expression applied to HTTP headers and to the
5222 request line. This is an extended regular expression. Parenthesis
5223 grouping is supported and no preliminary backslash is required.
5224 Any space or known delimiter must be escaped using a backslash
5225 ('\'). The pattern applies to a full line at a time. The
5226 "reqdeny" keyword strictly matches case while "reqideny" ignores
5227 case.
5228
Willy Tarreau5321c422010-01-28 20:35:13 +01005229 <cond> is an optional matching condition built from ACLs. It makes it
5230 possible to ignore this rule when other conditions are not met.
5231
Willy Tarreau303c0352008-01-17 19:01:39 +01005232 A request containing any line which matches extended regular expression
5233 <search> will mark the request as denied, even if any later test would
5234 result in an allow. The test applies both to the request line and to request
5235 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01005236 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01005237
Willy Tarreauced27012008-01-17 20:35:34 +01005238 A denied request will generate an "HTTP 403 forbidden" response once the
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01005239 complete request has been parsed. This is consistent with what is practiced
Willy Tarreaud72758d2010-01-12 10:42:19 +01005240 using ACLs.
Willy Tarreauced27012008-01-17 20:35:34 +01005241
Willy Tarreau303c0352008-01-17 19:01:39 +01005242 It is easier, faster and more powerful to use ACLs to write access policies.
5243 Reqdeny, reqallow and reqpass should be avoided in new designs.
5244
5245 Example :
5246 # refuse *.local, then allow www.*
5247 reqideny ^Host:\ .*\.local
5248 reqiallow ^Host:\ www\.
5249
Willy Tarreau5321c422010-01-28 20:35:13 +01005250 See also: "reqallow", "rspdeny", "block", section 6 about HTTP header
5251 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005252
5253
Willy Tarreau5321c422010-01-28 20:35:13 +01005254reqpass <search> [{if | unless} <cond>]
5255reqipass <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005256 Ignore any HTTP request line matching a regular expression in next rules
5257 May be used in sections : defaults | frontend | listen | backend
5258 no | yes | yes | yes
5259 Arguments :
5260 <search> is the regular expression applied to HTTP headers and to the
5261 request line. This is an extended regular expression. Parenthesis
5262 grouping is supported and no preliminary backslash is required.
5263 Any space or known delimiter must be escaped using a backslash
5264 ('\'). The pattern applies to a full line at a time. The
5265 "reqpass" keyword strictly matches case while "reqipass" ignores
5266 case.
5267
Willy Tarreau5321c422010-01-28 20:35:13 +01005268 <cond> is an optional matching condition built from ACLs. It makes it
5269 possible to ignore this rule when other conditions are not met.
5270
Willy Tarreau303c0352008-01-17 19:01:39 +01005271 A request containing any line which matches extended regular expression
5272 <search> will skip next rules, without assigning any deny or allow verdict.
5273 The test applies both to the request line and to request headers. Keep in
5274 mind that URLs in request line are case-sensitive while header names are not.
5275
5276 It is easier, faster and more powerful to use ACLs to write access policies.
5277 Reqdeny, reqallow and reqpass should be avoided in new designs.
5278
5279 Example :
5280 # refuse *.local, then allow www.*, but ignore "www.private.local"
5281 reqipass ^Host:\ www.private\.local
5282 reqideny ^Host:\ .*\.local
5283 reqiallow ^Host:\ www\.
5284
Willy Tarreau5321c422010-01-28 20:35:13 +01005285 See also: "reqallow", "reqdeny", "block", section 6 about HTTP header
5286 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005287
5288
Willy Tarreau5321c422010-01-28 20:35:13 +01005289reqrep <search> <string> [{if | unless} <cond>]
5290reqirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005291 Replace a regular expression with a string in an HTTP request line
5292 May be used in sections : defaults | frontend | listen | backend
5293 no | yes | yes | yes
5294 Arguments :
5295 <search> is the regular expression applied to HTTP headers and to the
5296 request line. This is an extended regular expression. Parenthesis
5297 grouping is supported and no preliminary backslash is required.
5298 Any space or known delimiter must be escaped using a backslash
5299 ('\'). The pattern applies to a full line at a time. The "reqrep"
5300 keyword strictly matches case while "reqirep" ignores case.
5301
5302 <string> is the complete line to be added. Any space or known delimiter
5303 must be escaped using a backslash ('\'). References to matched
5304 pattern groups are possible using the common \N form, with N
5305 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005306 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01005307
Willy Tarreau5321c422010-01-28 20:35:13 +01005308 <cond> is an optional matching condition built from ACLs. It makes it
5309 possible to ignore this rule when other conditions are not met.
5310
Willy Tarreau303c0352008-01-17 19:01:39 +01005311 Any line matching extended regular expression <search> in the request (both
5312 the request line and header lines) will be completely replaced with <string>.
5313 Most common use of this is to rewrite URLs or domain names in "Host" headers.
5314
5315 Header transformations only apply to traffic which passes through HAProxy,
5316 and not to traffic generated by HAProxy, such as health-checks or error
5317 responses. Note that for increased readability, it is suggested to add enough
5318 spaces between the request and the response. Keep in mind that URLs in
5319 request line are case-sensitive while header names are not.
5320
5321 Example :
5322 # replace "/static/" with "/" at the beginning of any request path.
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04005323 reqrep ^([^\ :]*)\ /static/(.*) \1\ /\2
Willy Tarreau303c0352008-01-17 19:01:39 +01005324 # replace "www.mydomain.com" with "www" in the host name.
5325 reqirep ^Host:\ www.mydomain.com Host:\ www
5326
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04005327 See also: "reqadd", "reqdel", "rsprep", "tune.bufsize", section 6 about
5328 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005329
5330
Willy Tarreau5321c422010-01-28 20:35:13 +01005331reqtarpit <search> [{if | unless} <cond>]
5332reqitarpit <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005333 Tarpit an HTTP request containing a line matching a regular expression
5334 May be used in sections : defaults | frontend | listen | backend
5335 no | yes | yes | yes
5336 Arguments :
5337 <search> is the regular expression applied to HTTP headers and to the
5338 request line. This is an extended regular expression. Parenthesis
5339 grouping is supported and no preliminary backslash is required.
5340 Any space or known delimiter must be escaped using a backslash
5341 ('\'). The pattern applies to a full line at a time. The
5342 "reqtarpit" keyword strictly matches case while "reqitarpit"
5343 ignores case.
5344
Willy Tarreau5321c422010-01-28 20:35:13 +01005345 <cond> is an optional matching condition built from ACLs. It makes it
5346 possible to ignore this rule when other conditions are not met.
5347
Willy Tarreau303c0352008-01-17 19:01:39 +01005348 A request containing any line which matches extended regular expression
5349 <search> will be tarpitted, which means that it will connect to nowhere, will
Willy Tarreauced27012008-01-17 20:35:34 +01005350 be kept open for a pre-defined time, then will return an HTTP error 500 so
5351 that the attacker does not suspect it has been tarpitted. The status 500 will
5352 be reported in the logs, but the completion flags will indicate "PT". The
Willy Tarreau303c0352008-01-17 19:01:39 +01005353 delay is defined by "timeout tarpit", or "timeout connect" if the former is
5354 not set.
5355
5356 The goal of the tarpit is to slow down robots attacking servers with
5357 identifiable requests. Many robots limit their outgoing number of connections
5358 and stay connected waiting for a reply which can take several minutes to
5359 come. Depending on the environment and attack, it may be particularly
5360 efficient at reducing the load on the network and firewalls.
5361
Willy Tarreau5321c422010-01-28 20:35:13 +01005362 Examples :
Willy Tarreau303c0352008-01-17 19:01:39 +01005363 # ignore user-agents reporting any flavour of "Mozilla" or "MSIE", but
5364 # block all others.
5365 reqipass ^User-Agent:\.*(Mozilla|MSIE)
5366 reqitarpit ^User-Agent:
5367
Willy Tarreau5321c422010-01-28 20:35:13 +01005368 # block bad guys
5369 acl badguys src 10.1.0.3 172.16.13.20/28
5370 reqitarpit . if badguys
5371
5372 See also: "reqallow", "reqdeny", "reqpass", section 6 about HTTP header
5373 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005374
5375
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02005376retries <value>
5377 Set the number of retries to perform on a server after a connection failure
5378 May be used in sections: defaults | frontend | listen | backend
5379 yes | no | yes | yes
5380 Arguments :
5381 <value> is the number of times a connection attempt should be retried on
5382 a server when a connection either is refused or times out. The
5383 default value is 3.
5384
5385 It is important to understand that this value applies to the number of
5386 connection attempts, not full requests. When a connection has effectively
5387 been established to a server, there will be no more retry.
5388
5389 In order to avoid immediate reconnections to a server which is restarting,
5390 a turn-around timer of 1 second is applied before a retry occurs.
5391
5392 When "option redispatch" is set, the last retry may be performed on another
5393 server even if a cookie references a different server.
5394
5395 See also : "option redispatch"
5396
5397
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005398rspadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01005399 Add a header at the end of the HTTP response
5400 May be used in sections : defaults | frontend | listen | backend
5401 no | yes | yes | yes
5402 Arguments :
5403 <string> is the complete line to be added. Any space or known delimiter
5404 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005405 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01005406
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005407 <cond> is an optional matching condition built from ACLs. It makes it
5408 possible to ignore this rule when other conditions are not met.
5409
Willy Tarreau303c0352008-01-17 19:01:39 +01005410 A new line consisting in <string> followed by a line feed will be added after
5411 the last header of an HTTP response.
5412
5413 Header transformations only apply to traffic which passes through HAProxy,
5414 and not to traffic generated by HAProxy, such as health-checks or error
5415 responses.
5416
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005417 See also: "reqadd", section 6 about HTTP header manipulation, and section 7
5418 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005419
5420
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005421rspdel <search> [{if | unless} <cond>]
5422rspidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005423 Delete all headers matching a regular expression in an HTTP response
5424 May be used in sections : defaults | frontend | listen | backend
5425 no | yes | yes | yes
5426 Arguments :
5427 <search> is the regular expression applied to HTTP headers and to the
5428 response line. This is an extended regular expression, so
5429 parenthesis grouping is supported and no preliminary backslash
5430 is required. Any space or known delimiter must be escaped using
5431 a backslash ('\'). The pattern applies to a full line at a time.
5432 The "rspdel" keyword strictly matches case while "rspidel"
5433 ignores case.
5434
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005435 <cond> is an optional matching condition built from ACLs. It makes it
5436 possible to ignore this rule when other conditions are not met.
5437
Willy Tarreau303c0352008-01-17 19:01:39 +01005438 Any header line matching extended regular expression <search> in the response
5439 will be completely deleted. Most common use of this is to remove unwanted
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02005440 and/or sensitive headers or cookies from a response before passing it to the
Willy Tarreau303c0352008-01-17 19:01:39 +01005441 client.
5442
5443 Header transformations only apply to traffic which passes through HAProxy,
5444 and not to traffic generated by HAProxy, such as health-checks or error
5445 responses. Keep in mind that header names are not case-sensitive.
5446
5447 Example :
5448 # remove the Server header from responses
Willy Tarreau5e80e022013-05-25 08:31:25 +02005449 rspidel ^Server:.*
Willy Tarreau303c0352008-01-17 19:01:39 +01005450
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005451 See also: "rspadd", "rsprep", "reqdel", section 6 about HTTP header
5452 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005453
5454
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005455rspdeny <search> [{if | unless} <cond>]
5456rspideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005457 Block an HTTP response if a line matches a regular expression
5458 May be used in sections : defaults | frontend | listen | backend
5459 no | yes | yes | yes
5460 Arguments :
5461 <search> is the regular expression applied to HTTP headers and to the
5462 response line. This is an extended regular expression, so
5463 parenthesis grouping is supported and no preliminary backslash
5464 is required. Any space or known delimiter must be escaped using
5465 a backslash ('\'). The pattern applies to a full line at a time.
5466 The "rspdeny" keyword strictly matches case while "rspideny"
5467 ignores case.
5468
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005469 <cond> is an optional matching condition built from ACLs. It makes it
5470 possible to ignore this rule when other conditions are not met.
5471
Willy Tarreau303c0352008-01-17 19:01:39 +01005472 A response containing any line which matches extended regular expression
5473 <search> will mark the request as denied. The test applies both to the
5474 response line and to response headers. Keep in mind that header names are not
5475 case-sensitive.
5476
5477 Main use of this keyword is to prevent sensitive information leak and to
Willy Tarreauced27012008-01-17 20:35:34 +01005478 block the response before it reaches the client. If a response is denied, it
5479 will be replaced with an HTTP 502 error so that the client never retrieves
5480 any sensitive data.
Willy Tarreau303c0352008-01-17 19:01:39 +01005481
5482 It is easier, faster and more powerful to use ACLs to write access policies.
5483 Rspdeny should be avoided in new designs.
5484
5485 Example :
5486 # Ensure that no content type matching ms-word will leak
5487 rspideny ^Content-type:\.*/ms-word
5488
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005489 See also: "reqdeny", "acl", "block", section 6 about HTTP header manipulation
5490 and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005491
5492
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005493rsprep <search> <string> [{if | unless} <cond>]
5494rspirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005495 Replace a regular expression with a string in an HTTP response line
5496 May be used in sections : defaults | frontend | listen | backend
5497 no | yes | yes | yes
5498 Arguments :
5499 <search> is the regular expression applied to HTTP headers and to the
5500 response line. This is an extended regular expression, so
5501 parenthesis grouping is supported and no preliminary backslash
5502 is required. Any space or known delimiter must be escaped using
5503 a backslash ('\'). The pattern applies to a full line at a time.
5504 The "rsprep" keyword strictly matches case while "rspirep"
5505 ignores case.
5506
5507 <string> is the complete line to be added. Any space or known delimiter
5508 must be escaped using a backslash ('\'). References to matched
5509 pattern groups are possible using the common \N form, with N
5510 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005511 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01005512
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005513 <cond> is an optional matching condition built from ACLs. It makes it
5514 possible to ignore this rule when other conditions are not met.
5515
Willy Tarreau303c0352008-01-17 19:01:39 +01005516 Any line matching extended regular expression <search> in the response (both
5517 the response line and header lines) will be completely replaced with
5518 <string>. Most common use of this is to rewrite Location headers.
5519
5520 Header transformations only apply to traffic which passes through HAProxy,
5521 and not to traffic generated by HAProxy, such as health-checks or error
5522 responses. Note that for increased readability, it is suggested to add enough
5523 spaces between the request and the response. Keep in mind that header names
5524 are not case-sensitive.
5525
5526 Example :
5527 # replace "Location: 127.0.0.1:8080" with "Location: www.mydomain.com"
5528 rspirep ^Location:\ 127.0.0.1:8080 Location:\ www.mydomain.com
5529
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005530 See also: "rspadd", "rspdel", "reqrep", section 6 about HTTP header
5531 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005532
5533
David du Colombier486df472011-03-17 10:40:26 +01005534server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005535 Declare a server in a backend
5536 May be used in sections : defaults | frontend | listen | backend
5537 no | no | yes | yes
5538 Arguments :
5539 <name> is the internal name assigned to this server. This name will
Cyril Bonté941a0c62012-10-15 19:44:24 +02005540 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -05005541 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005542
David du Colombier486df472011-03-17 10:40:26 +01005543 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
5544 resolvable hostname is supported, but this name will be resolved
5545 during start-up. Address "0.0.0.0" or "*" has a special meaning.
5546 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +02005547 address as the one from the client connection. This is useful in
5548 transparent proxy architectures where the client's connection is
5549 intercepted and haproxy must forward to the original destination
5550 address. This is more or less what the "transparent" keyword does
5551 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +01005552 to report statistics. Optionally, an address family prefix may be
5553 used before the address to force the family regardless of the
5554 address format, which can be useful to specify a path to a unix
5555 socket with no slash ('/'). Currently supported prefixes are :
5556 - 'ipv4@' -> address is always IPv4
5557 - 'ipv6@' -> address is always IPv6
5558 - 'unix@' -> address is a path to a local unix socket
Willy Tarreaudad36a32013-03-11 01:20:04 +01005559 Any part of the address string may reference any number of
5560 environment variables by preceding their name with a dollar
5561 sign ('$') and optionally enclosing them with braces ('{}'),
5562 similarly to what is done in Bourne shell.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005563
Willy Tarreaub6205fd2012-09-24 12:27:33 +02005564 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005565 be sent to this port. If unset, the same port the client
5566 connected to will be used. The port may also be prefixed by a "+"
5567 or a "-". In this case, the server's port will be determined by
5568 adding this value to the client's port.
5569
5570 <param*> is a list of parameters for this server. The "server" keywords
5571 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005572 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005573
5574 Examples :
5575 server first 10.1.1.1:1080 cookie first check inter 1000
5576 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +01005577 server transp ipv4@
Willy Tarreaudad36a32013-03-11 01:20:04 +01005578 server backup ${SRV_BACKUP}:1080 backup
5579 server www1_dc1 ${LAN_DC1}.101:80
5580 server www1_dc2 ${LAN_DC2}.101:80
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005581
Mark Lamourinec2247f02012-01-04 13:02:01 -05005582 See also: "default-server", "http-send-name-header" and section 5 about
5583 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005584
5585
5586source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02005587source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +01005588source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005589 Set the source address for outgoing connections
5590 May be used in sections : defaults | frontend | listen | backend
5591 yes | no | yes | yes
5592 Arguments :
5593 <addr> is the IPv4 address HAProxy will bind to before connecting to a
5594 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +01005595
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005596 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +01005597 the most appropriate address to reach its destination. Optionally
5598 an address family prefix may be used before the address to force
5599 the family regardless of the address format, which can be useful
5600 to specify a path to a unix socket with no slash ('/'). Currently
5601 supported prefixes are :
5602 - 'ipv4@' -> address is always IPv4
5603 - 'ipv6@' -> address is always IPv6
5604 - 'unix@' -> address is a path to a local unix socket
Willy Tarreaudad36a32013-03-11 01:20:04 +01005605 Any part of the address string may reference any number of
5606 environment variables by preceding their name with a dollar
5607 sign ('$') and optionally enclosing them with braces ('{}'),
5608 similarly to what is done in Bourne shell.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005609
5610 <port> is an optional port. It is normally not needed but may be useful
5611 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02005612 the system will select a free port. Note that port ranges are not
5613 supported in the backend. If you want to force port ranges, you
5614 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005615
5616 <addr2> is the IP address to present to the server when connections are
5617 forwarded in full transparent proxy mode. This is currently only
5618 supported on some patched Linux kernels. When this address is
5619 specified, clients connecting to the server will be presented
5620 with this address, while health checks will still use the address
5621 <addr>.
5622
5623 <port2> is the optional port to present to the server when connections
5624 are forwarded in full transparent proxy mode (see <addr2> above).
5625 The default value of zero means the system will select a free
5626 port.
5627
Willy Tarreaubce70882009-09-07 11:51:47 +02005628 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
5629 This is the name of a comma-separated header list which can
5630 contain multiple IP addresses. By default, the last occurrence is
5631 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +01005632 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +02005633 by previous proxy, typically Stunnel. In order to use another
5634 occurrence from the last one, please see the <occ> parameter
5635 below. When the header (or occurrence) is not found, no binding
5636 is performed so that the proxy's default IP address is used. Also
5637 keep in mind that the header name is case insensitive, as for any
5638 HTTP header.
5639
5640 <occ> is the occurrence number of a value to be used in a multi-value
5641 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005642 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +02005643 address. Positive values indicate a position from the first
5644 occurrence, 1 being the first one. Negative values indicate
5645 positions relative to the last one, -1 being the last one. This
5646 is helpful for situations where an X-Forwarded-For header is set
5647 at the entry point of an infrastructure and must be used several
5648 proxy layers away. When this value is not specified, -1 is
5649 assumed. Passing a zero here disables the feature.
5650
Willy Tarreaud53f96b2009-02-04 18:46:54 +01005651 <name> is an optional interface name to which to bind to for outgoing
5652 traffic. On systems supporting this features (currently, only
5653 Linux), this allows one to bind all traffic to the server to
5654 this interface even if it is not the one the system would select
5655 based on routing tables. This should be used with extreme care.
5656 Note that using this option requires root privileges.
5657
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005658 The "source" keyword is useful in complex environments where a specific
5659 address only is allowed to connect to the servers. It may be needed when a
5660 private address must be used through a public gateway for instance, and it is
5661 known that the system cannot determine the adequate source address by itself.
5662
5663 An extension which is available on certain patched Linux kernels may be used
5664 through the "usesrc" optional keyword. It makes it possible to connect to the
5665 servers with an IP address which does not belong to the system itself. This
5666 is called "full transparent proxy mode". For this to work, the destination
5667 servers have to route their traffic back to this address through the machine
5668 running HAProxy, and IP forwarding must generally be enabled on this machine.
5669
5670 In this "full transparent proxy" mode, it is possible to force a specific IP
5671 address to be presented to the servers. This is not much used in fact. A more
5672 common use is to tell HAProxy to present the client's IP address. For this,
5673 there are two methods :
5674
5675 - present the client's IP and port addresses. This is the most transparent
5676 mode, but it can cause problems when IP connection tracking is enabled on
5677 the machine, because a same connection may be seen twice with different
5678 states. However, this solution presents the huge advantage of not
5679 limiting the system to the 64k outgoing address+port couples, because all
5680 of the client ranges may be used.
5681
5682 - present only the client's IP address and select a spare port. This
5683 solution is still quite elegant but slightly less transparent (downstream
5684 firewalls logs will not match upstream's). It also presents the downside
5685 of limiting the number of concurrent connections to the usual 64k ports.
5686 However, since the upstream and downstream ports are different, local IP
5687 connection tracking on the machine will not be upset by the reuse of the
5688 same session.
5689
5690 Note that depending on the transparent proxy technology used, it may be
5691 required to force the source address. In fact, cttproxy version 2 requires an
5692 IP address in <addr> above, and does not support setting of "0.0.0.0" as the
5693 IP address because it creates NAT entries which much match the exact outgoing
5694 address. Tproxy version 4 and some other kernel patches which work in pure
5695 forwarding mode generally will not have this limitation.
5696
5697 This option sets the default source for all servers in the backend. It may
5698 also be specified in a "defaults" section. Finer source address specification
5699 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005700 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005701
5702 Examples :
5703 backend private
5704 # Connect to the servers using our 192.168.1.200 source address
5705 source 192.168.1.200
5706
5707 backend transparent_ssl1
5708 # Connect to the SSL farm from the client's source address
5709 source 192.168.1.200 usesrc clientip
5710
5711 backend transparent_ssl2
5712 # Connect to the SSL farm from the client's source address and port
5713 # not recommended if IP conntrack is present on the local machine.
5714 source 192.168.1.200 usesrc client
5715
5716 backend transparent_ssl3
5717 # Connect to the SSL farm from the client's source address. It
5718 # is more conntrack-friendly.
5719 source 192.168.1.200 usesrc clientip
5720
5721 backend transparent_smtp
5722 # Connect to the SMTP farm from the client's source address/port
5723 # with Tproxy version 4.
5724 source 0.0.0.0 usesrc clientip
5725
Willy Tarreaubce70882009-09-07 11:51:47 +02005726 backend transparent_http
5727 # Connect to the servers using the client's IP as seen by previous
5728 # proxy.
5729 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
5730
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005731 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005732 the Linux kernel on www.balabit.com, the "bind" keyword.
5733
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005734
Willy Tarreau844e3c52008-01-11 16:28:18 +01005735srvtimeout <timeout> (deprecated)
5736 Set the maximum inactivity time on the server side.
5737 May be used in sections : defaults | frontend | listen | backend
5738 yes | no | yes | yes
5739 Arguments :
5740 <timeout> is the timeout value specified in milliseconds by default, but
5741 can be in any other unit if the number is suffixed by the unit,
5742 as explained at the top of this document.
5743
5744 The inactivity timeout applies when the server is expected to acknowledge or
5745 send data. In HTTP mode, this timeout is particularly important to consider
5746 during the first phase of the server's response, when it has to send the
5747 headers, as it directly represents the server's processing time for the
5748 request. To find out what value to put there, it's often good to start with
5749 what would be considered as unacceptable response times, then check the logs
5750 to observe the response time distribution, and adjust the value accordingly.
5751
5752 The value is specified in milliseconds by default, but can be in any other
5753 unit if the number is suffixed by the unit, as specified at the top of this
5754 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
5755 recommended that the client timeout remains equal to the server timeout in
5756 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01005757 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01005758 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreaud72758d2010-01-12 10:42:19 +01005759 seconds (eg: 4 or 5 seconds minimum).
Willy Tarreau844e3c52008-01-11 16:28:18 +01005760
5761 This parameter is specific to backends, but can be specified once for all in
5762 "defaults" sections. This is in fact one of the easiest solutions not to
5763 forget about it. An unspecified timeout results in an infinite timeout, which
5764 is not recommended. Such a usage is accepted and works but reports a warning
5765 during startup because it may results in accumulation of expired sessions in
5766 the system if the system's timeouts are not configured either.
5767
5768 This parameter is provided for compatibility but is currently deprecated.
5769 Please use "timeout server" instead.
5770
Willy Tarreauce887fd2012-05-12 12:50:00 +02005771 See also : "timeout server", "timeout tunnel", "timeout client" and
5772 "clitimeout".
Willy Tarreau844e3c52008-01-11 16:28:18 +01005773
5774
Cyril Bonté66c327d2010-10-12 00:14:37 +02005775stats admin { if | unless } <cond>
5776 Enable statistics admin level if/unless a condition is matched
5777 May be used in sections : defaults | frontend | listen | backend
5778 no | no | yes | yes
5779
5780 This statement enables the statistics admin level if/unless a condition is
5781 matched.
5782
5783 The admin level allows to enable/disable servers from the web interface. By
5784 default, statistics page is read-only for security reasons.
5785
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01005786 Note : Consider not using this feature in multi-process mode (nbproc > 1)
5787 unless you know what you do : memory is not shared between the
5788 processes, which can result in random behaviours.
5789
Cyril Bonté23b39d92011-02-10 22:54:44 +01005790 Currently, the POST request is limited to the buffer size minus the reserved
5791 buffer space, which means that if the list of servers is too long, the
5792 request won't be processed. It is recommended to alter few servers at a
5793 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +02005794
5795 Example :
5796 # statistics admin level only for localhost
5797 backend stats_localhost
5798 stats enable
5799 stats admin if LOCALHOST
5800
5801 Example :
5802 # statistics admin level always enabled because of the authentication
5803 backend stats_auth
5804 stats enable
5805 stats auth admin:AdMiN123
5806 stats admin if TRUE
5807
5808 Example :
5809 # statistics admin level depends on the authenticated user
5810 userlist stats-auth
5811 group admin users admin
5812 user admin insecure-password AdMiN123
5813 group readonly users haproxy
5814 user haproxy insecure-password haproxy
5815
5816 backend stats_auth
5817 stats enable
5818 acl AUTH http_auth(stats-auth)
5819 acl AUTH_ADMIN http_auth_group(stats-auth) admin
5820 stats http-request auth unless AUTH
5821 stats admin if AUTH_ADMIN
5822
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01005823 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
5824 "bind-process", section 3.4 about userlists and section 7 about
5825 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +02005826
5827
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005828stats auth <user>:<passwd>
5829 Enable statistics with authentication and grant access to an account
5830 May be used in sections : defaults | frontend | listen | backend
5831 yes | no | yes | yes
5832 Arguments :
5833 <user> is a user name to grant access to
5834
5835 <passwd> is the cleartext password associated to this user
5836
5837 This statement enables statistics with default settings, and restricts access
5838 to declared users only. It may be repeated as many times as necessary to
5839 allow as many users as desired. When a user tries to access the statistics
5840 without a valid account, a "401 Forbidden" response will be returned so that
5841 the browser asks the user to provide a valid user and password. The real
5842 which will be returned to the browser is configurable using "stats realm".
5843
5844 Since the authentication method is HTTP Basic Authentication, the passwords
5845 circulate in cleartext on the network. Thus, it was decided that the
5846 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02005847 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005848
5849 It is also possible to reduce the scope of the proxies which appear in the
5850 report using "stats scope".
5851
5852 Though this statement alone is enough to enable statistics reporting, it is
5853 recommended to set all other settings in order to avoid relying on default
5854 unobvious parameters.
5855
5856 Example :
5857 # public access (limited to this backend only)
5858 backend public_www
5859 server srv1 192.168.0.1:80
5860 stats enable
5861 stats hide-version
5862 stats scope .
5863 stats uri /admin?stats
5864 stats realm Haproxy\ Statistics
5865 stats auth admin1:AdMiN123
5866 stats auth admin2:AdMiN321
5867
5868 # internal monitoring access (unlimited)
5869 backend private_monitoring
5870 stats enable
5871 stats uri /admin?stats
5872 stats refresh 5s
5873
5874 See also : "stats enable", "stats realm", "stats scope", "stats uri"
5875
5876
5877stats enable
5878 Enable statistics reporting with default settings
5879 May be used in sections : defaults | frontend | listen | backend
5880 yes | no | yes | yes
5881 Arguments : none
5882
5883 This statement enables statistics reporting with default settings defined
5884 at build time. Unless stated otherwise, these settings are used :
5885 - stats uri : /haproxy?stats
5886 - stats realm : "HAProxy Statistics"
5887 - stats auth : no authentication
5888 - stats scope : no restriction
5889
5890 Though this statement alone is enough to enable statistics reporting, it is
5891 recommended to set all other settings in order to avoid relying on default
5892 unobvious parameters.
5893
5894 Example :
5895 # public access (limited to this backend only)
5896 backend public_www
5897 server srv1 192.168.0.1:80
5898 stats enable
5899 stats hide-version
5900 stats scope .
5901 stats uri /admin?stats
5902 stats realm Haproxy\ Statistics
5903 stats auth admin1:AdMiN123
5904 stats auth admin2:AdMiN321
5905
5906 # internal monitoring access (unlimited)
5907 backend private_monitoring
5908 stats enable
5909 stats uri /admin?stats
5910 stats refresh 5s
5911
5912 See also : "stats auth", "stats realm", "stats uri"
5913
5914
Willy Tarreaud63335a2010-02-26 12:56:52 +01005915stats hide-version
5916 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02005917 May be used in sections : defaults | frontend | listen | backend
5918 yes | no | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01005919 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02005920
Willy Tarreaud63335a2010-02-26 12:56:52 +01005921 By default, the stats page reports some useful status information along with
5922 the statistics. Among them is HAProxy's version. However, it is generally
5923 considered dangerous to report precise version to anyone, as it can help them
5924 target known weaknesses with specific attacks. The "stats hide-version"
5925 statement removes the version from the statistics report. This is recommended
5926 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02005927
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02005928 Though this statement alone is enough to enable statistics reporting, it is
5929 recommended to set all other settings in order to avoid relying on default
5930 unobvious parameters.
5931
Willy Tarreaud63335a2010-02-26 12:56:52 +01005932 Example :
5933 # public access (limited to this backend only)
5934 backend public_www
5935 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02005936 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01005937 stats hide-version
5938 stats scope .
5939 stats uri /admin?stats
5940 stats realm Haproxy\ Statistics
5941 stats auth admin1:AdMiN123
5942 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02005943
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02005944 # internal monitoring access (unlimited)
5945 backend private_monitoring
5946 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01005947 stats uri /admin?stats
5948 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +01005949
Willy Tarreaud63335a2010-02-26 12:56:52 +01005950 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02005951
Willy Tarreau983e01e2010-01-11 18:42:06 +01005952
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02005953stats http-request { allow | deny | auth [realm <realm>] }
5954 [ { if | unless } <condition> ]
5955 Access control for statistics
5956
5957 May be used in sections: defaults | frontend | listen | backend
5958 no | no | yes | yes
5959
5960 As "http-request", these set of options allow to fine control access to
5961 statistics. Each option may be followed by if/unless and acl.
5962 First option with matched condition (or option without condition) is final.
5963 For "deny" a 403 error will be returned, for "allow" normal processing is
5964 performed, for "auth" a 401/407 error code is returned so the client
5965 should be asked to enter a username and password.
5966
5967 There is no fixed limit to the number of http-request statements per
5968 instance.
5969
5970 See also : "http-request", section 3.4 about userlists and section 7
5971 about ACL usage.
5972
5973
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005974stats realm <realm>
5975 Enable statistics and set authentication realm
5976 May be used in sections : defaults | frontend | listen | backend
5977 yes | no | yes | yes
5978 Arguments :
5979 <realm> is the name of the HTTP Basic Authentication realm reported to
5980 the browser. The browser uses it to display it in the pop-up
5981 inviting the user to enter a valid username and password.
5982
5983 The realm is read as a single word, so any spaces in it should be escaped
5984 using a backslash ('\').
5985
5986 This statement is useful only in conjunction with "stats auth" since it is
5987 only related to authentication.
5988
5989 Though this statement alone is enough to enable statistics reporting, it is
5990 recommended to set all other settings in order to avoid relying on default
5991 unobvious parameters.
5992
5993 Example :
5994 # public access (limited to this backend only)
5995 backend public_www
5996 server srv1 192.168.0.1:80
5997 stats enable
5998 stats hide-version
5999 stats scope .
6000 stats uri /admin?stats
6001 stats realm Haproxy\ Statistics
6002 stats auth admin1:AdMiN123
6003 stats auth admin2:AdMiN321
6004
6005 # internal monitoring access (unlimited)
6006 backend private_monitoring
6007 stats enable
6008 stats uri /admin?stats
6009 stats refresh 5s
6010
6011 See also : "stats auth", "stats enable", "stats uri"
6012
6013
6014stats refresh <delay>
6015 Enable statistics with automatic refresh
6016 May be used in sections : defaults | frontend | listen | backend
6017 yes | no | yes | yes
6018 Arguments :
6019 <delay> is the suggested refresh delay, specified in seconds, which will
6020 be returned to the browser consulting the report page. While the
6021 browser is free to apply any delay, it will generally respect it
6022 and refresh the page this every seconds. The refresh interval may
6023 be specified in any other non-default time unit, by suffixing the
6024 unit after the value, as explained at the top of this document.
6025
6026 This statement is useful on monitoring displays with a permanent page
6027 reporting the load balancer's activity. When set, the HTML report page will
6028 include a link "refresh"/"stop refresh" so that the user can select whether
6029 he wants automatic refresh of the page or not.
6030
6031 Though this statement alone is enough to enable statistics reporting, it is
6032 recommended to set all other settings in order to avoid relying on default
6033 unobvious parameters.
6034
6035 Example :
6036 # public access (limited to this backend only)
6037 backend public_www
6038 server srv1 192.168.0.1:80
6039 stats enable
6040 stats hide-version
6041 stats scope .
6042 stats uri /admin?stats
6043 stats realm Haproxy\ Statistics
6044 stats auth admin1:AdMiN123
6045 stats auth admin2:AdMiN321
6046
6047 # internal monitoring access (unlimited)
6048 backend private_monitoring
6049 stats enable
6050 stats uri /admin?stats
6051 stats refresh 5s
6052
6053 See also : "stats auth", "stats enable", "stats realm", "stats uri"
6054
6055
6056stats scope { <name> | "." }
6057 Enable statistics and limit access scope
6058 May be used in sections : defaults | frontend | listen | backend
6059 yes | no | yes | yes
6060 Arguments :
6061 <name> is the name of a listen, frontend or backend section to be
6062 reported. The special name "." (a single dot) designates the
6063 section in which the statement appears.
6064
6065 When this statement is specified, only the sections enumerated with this
6066 statement will appear in the report. All other ones will be hidden. This
6067 statement may appear as many times as needed if multiple sections need to be
6068 reported. Please note that the name checking is performed as simple string
6069 comparisons, and that it is never checked that a give section name really
6070 exists.
6071
6072 Though this statement alone is enough to enable statistics reporting, it is
6073 recommended to set all other settings in order to avoid relying on default
6074 unobvious parameters.
6075
6076 Example :
6077 # public access (limited to this backend only)
6078 backend public_www
6079 server srv1 192.168.0.1:80
6080 stats enable
6081 stats hide-version
6082 stats scope .
6083 stats uri /admin?stats
6084 stats realm Haproxy\ Statistics
6085 stats auth admin1:AdMiN123
6086 stats auth admin2:AdMiN321
6087
6088 # internal monitoring access (unlimited)
6089 backend private_monitoring
6090 stats enable
6091 stats uri /admin?stats
6092 stats refresh 5s
6093
6094 See also : "stats auth", "stats enable", "stats realm", "stats uri"
6095
Willy Tarreaud63335a2010-02-26 12:56:52 +01006096
Willy Tarreauc9705a12010-07-27 20:05:50 +02006097stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +01006098 Enable reporting of a description on the statistics page.
6099 May be used in sections : defaults | frontend | listen | backend
6100 yes | no | yes | yes
6101
Willy Tarreauc9705a12010-07-27 20:05:50 +02006102 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +01006103 description from global section is automatically used instead.
6104
6105 This statement is useful for users that offer shared services to their
6106 customers, where node or description should be different for each customer.
6107
6108 Though this statement alone is enough to enable statistics reporting, it is
6109 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04006110 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +01006111
6112 Example :
6113 # internal monitoring access (unlimited)
6114 backend private_monitoring
6115 stats enable
6116 stats show-desc Master node for Europe, Asia, Africa
6117 stats uri /admin?stats
6118 stats refresh 5s
6119
6120 See also: "show-node", "stats enable", "stats uri" and "description" in
6121 global section.
6122
6123
6124stats show-legends
6125 Enable reporting additional informations on the statistics page :
6126 - cap: capabilities (proxy)
6127 - mode: one of tcp, http or health (proxy)
6128 - id: SNMP ID (proxy, socket, server)
6129 - IP (socket, server)
6130 - cookie (backend, server)
6131
6132 Though this statement alone is enough to enable statistics reporting, it is
6133 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04006134 unobvious parameters. Default behaviour is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +01006135
6136 See also: "stats enable", "stats uri".
6137
6138
6139stats show-node [ <name> ]
6140 Enable reporting of a host name on the statistics page.
6141 May be used in sections : defaults | frontend | listen | backend
6142 yes | no | yes | yes
6143 Arguments:
6144 <name> is an optional name to be reported. If unspecified, the
6145 node name from global section is automatically used instead.
6146
6147 This statement is useful for users that offer shared services to their
6148 customers, where node or description might be different on a stats page
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04006149 provided for each customer. Default behaviour is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +01006150
6151 Though this statement alone is enough to enable statistics reporting, it is
6152 recommended to set all other settings in order to avoid relying on default
6153 unobvious parameters.
6154
6155 Example:
6156 # internal monitoring access (unlimited)
6157 backend private_monitoring
6158 stats enable
6159 stats show-node Europe-1
6160 stats uri /admin?stats
6161 stats refresh 5s
6162
6163 See also: "show-desc", "stats enable", "stats uri", and "node" in global
6164 section.
6165
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006166
6167stats uri <prefix>
6168 Enable statistics and define the URI prefix to access them
6169 May be used in sections : defaults | frontend | listen | backend
6170 yes | no | yes | yes
6171 Arguments :
6172 <prefix> is the prefix of any URI which will be redirected to stats. This
6173 prefix may contain a question mark ('?') to indicate part of a
6174 query string.
6175
6176 The statistics URI is intercepted on the relayed traffic, so it appears as a
6177 page within the normal application. It is strongly advised to ensure that the
6178 selected URI will never appear in the application, otherwise it will never be
6179 possible to reach it in the application.
6180
6181 The default URI compiled in haproxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006182 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006183 It is generally a good idea to include a question mark in the URI so that
6184 intermediate proxies refrain from caching the results. Also, since any string
6185 beginning with the prefix will be accepted as a stats request, the question
6186 mark helps ensuring that no valid URI will begin with the same words.
6187
6188 It is sometimes very convenient to use "/" as the URI prefix, and put that
6189 statement in a "listen" instance of its own. That makes it easy to dedicate
6190 an address or a port to statistics only.
6191
6192 Though this statement alone is enough to enable statistics reporting, it is
6193 recommended to set all other settings in order to avoid relying on default
6194 unobvious parameters.
6195
6196 Example :
6197 # public access (limited to this backend only)
6198 backend public_www
6199 server srv1 192.168.0.1:80
6200 stats enable
6201 stats hide-version
6202 stats scope .
6203 stats uri /admin?stats
6204 stats realm Haproxy\ Statistics
6205 stats auth admin1:AdMiN123
6206 stats auth admin2:AdMiN321
6207
6208 # internal monitoring access (unlimited)
6209 backend private_monitoring
6210 stats enable
6211 stats uri /admin?stats
6212 stats refresh 5s
6213
6214 See also : "stats auth", "stats enable", "stats realm"
6215
6216
Willy Tarreaud63335a2010-02-26 12:56:52 +01006217stick match <pattern> [table <table>] [{if | unless} <cond>]
6218 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006219 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +01006220 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006221
6222 Arguments :
Willy Tarreau74ca5042013-06-11 23:12:07 +02006223 <pattern> is a pattern extraction rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006224 describes what elements of the incoming request or connection
6225 will be analysed in the hope to find a matching entry in a
6226 stickiness table. This rule is mandatory.
6227
6228 <table> is an optional stickiness table name. If unspecified, the same
6229 backend's table is used. A stickiness table is declared using
6230 the "stick-table" statement.
6231
6232 <cond> is an optional matching condition. It makes it possible to match
6233 on a certain criterion only when other conditions are met (or
6234 not met). For instance, it could be used to match on a source IP
6235 address except when a request passes through a known proxy, in
6236 which case we'd match on a header containing that IP address.
6237
6238 Some protocols or applications require complex stickiness rules and cannot
6239 always simply rely on cookies nor hashing. The "stick match" statement
6240 describes a rule to extract the stickiness criterion from an incoming request
6241 or connection. See section 7 for a complete list of possible patterns and
6242 transformation rules.
6243
6244 The table has to be declared using the "stick-table" statement. It must be of
6245 a type compatible with the pattern. By default it is the one which is present
6246 in the same backend. It is possible to share a table with other backends by
6247 referencing it using the "table" keyword. If another table is referenced,
6248 the server's ID inside the backends are used. By default, all server IDs
6249 start at 1 in each backend, so the server ordering is enough. But in case of
6250 doubt, it is highly recommended to force server IDs using their "id" setting.
6251
6252 It is possible to restrict the conditions where a "stick match" statement
6253 will apply, using "if" or "unless" followed by a condition. See section 7 for
6254 ACL based conditions.
6255
6256 There is no limit on the number of "stick match" statements. The first that
6257 applies and matches will cause the request to be directed to the same server
6258 as was used for the request which created the entry. That way, multiple
6259 matches can be used as fallbacks.
6260
6261 The stick rules are checked after the persistence cookies, so they will not
6262 affect stickiness if a cookie has already been used to select a server. That
6263 way, it becomes very easy to insert cookies and match on IP addresses in
6264 order to maintain stickiness between HTTP and HTTPS.
6265
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006266 Note : Consider not using this feature in multi-process mode (nbproc > 1)
6267 unless you know what you do : memory is not shared between the
6268 processes, which can result in random behaviours.
6269
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006270 Example :
6271 # forward SMTP users to the same server they just used for POP in the
6272 # last 30 minutes
6273 backend pop
6274 mode tcp
6275 balance roundrobin
6276 stick store-request src
6277 stick-table type ip size 200k expire 30m
6278 server s1 192.168.1.1:110
6279 server s2 192.168.1.1:110
6280
6281 backend smtp
6282 mode tcp
6283 balance roundrobin
6284 stick match src table pop
6285 server s1 192.168.1.1:25
6286 server s2 192.168.1.1:25
6287
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006288 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
6289 about ACLs and pattern extraction.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006290
6291
6292stick on <pattern> [table <table>] [{if | unless} <condition>]
6293 Define a request pattern to associate a user to a server
6294 May be used in sections : defaults | frontend | listen | backend
6295 no | no | yes | yes
6296
6297 Note : This form is exactly equivalent to "stick match" followed by
6298 "stick store-request", all with the same arguments. Please refer
6299 to both keywords for details. It is only provided as a convenience
6300 for writing more maintainable configurations.
6301
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006302 Note : Consider not using this feature in multi-process mode (nbproc > 1)
6303 unless you know what you do : memory is not shared between the
6304 processes, which can result in random behaviours.
6305
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006306 Examples :
6307 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +01006308 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006309
6310 # ...is strictly equivalent to this one :
6311 stick match src table pop if !localhost
6312 stick store-request src table pop if !localhost
6313
6314
6315 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
6316 # well as HTTP without cookie. Share the same table between both accesses.
6317 backend http
6318 mode http
6319 balance roundrobin
6320 stick on src table https
6321 cookie SRV insert indirect nocache
6322 server s1 192.168.1.1:80 cookie s1
6323 server s2 192.168.1.1:80 cookie s2
6324
6325 backend https
6326 mode tcp
6327 balance roundrobin
6328 stick-table type ip size 200k expire 30m
6329 stick on src
6330 server s1 192.168.1.1:443
6331 server s2 192.168.1.1:443
6332
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006333 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006334
6335
6336stick store-request <pattern> [table <table>] [{if | unless} <condition>]
6337 Define a request pattern used to create an entry in a stickiness table
6338 May be used in sections : defaults | frontend | listen | backend
6339 no | no | yes | yes
6340
6341 Arguments :
Willy Tarreau74ca5042013-06-11 23:12:07 +02006342 <pattern> is a pattern extraction rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006343 describes what elements of the incoming request or connection
6344 will be analysed, extracted and stored in the table once a
6345 server is selected.
6346
6347 <table> is an optional stickiness table name. If unspecified, the same
6348 backend's table is used. A stickiness table is declared using
6349 the "stick-table" statement.
6350
6351 <cond> is an optional storage condition. It makes it possible to store
6352 certain criteria only when some conditions are met (or not met).
6353 For instance, it could be used to store the source IP address
6354 except when the request passes through a known proxy, in which
6355 case we'd store a converted form of a header containing that IP
6356 address.
6357
6358 Some protocols or applications require complex stickiness rules and cannot
6359 always simply rely on cookies nor hashing. The "stick store-request" statement
6360 describes a rule to decide what to extract from the request and when to do
6361 it, in order to store it into a stickiness table for further requests to
6362 match it using the "stick match" statement. Obviously the extracted part must
6363 make sense and have a chance to be matched in a further request. Storing a
6364 client's IP address for instance often makes sense. Storing an ID found in a
6365 URL parameter also makes sense. Storing a source port will almost never make
6366 any sense because it will be randomly matched. See section 7 for a complete
6367 list of possible patterns and transformation rules.
6368
6369 The table has to be declared using the "stick-table" statement. It must be of
6370 a type compatible with the pattern. By default it is the one which is present
6371 in the same backend. It is possible to share a table with other backends by
6372 referencing it using the "table" keyword. If another table is referenced,
6373 the server's ID inside the backends are used. By default, all server IDs
6374 start at 1 in each backend, so the server ordering is enough. But in case of
6375 doubt, it is highly recommended to force server IDs using their "id" setting.
6376
6377 It is possible to restrict the conditions where a "stick store-request"
6378 statement will apply, using "if" or "unless" followed by a condition. This
6379 condition will be evaluated while parsing the request, so any criteria can be
6380 used. See section 7 for ACL based conditions.
6381
6382 There is no limit on the number of "stick store-request" statements, but
6383 there is a limit of 8 simultaneous stores per request or response. This
6384 makes it possible to store up to 8 criteria, all extracted from either the
6385 request or the response, regardless of the number of rules. Only the 8 first
6386 ones which match will be kept. Using this, it is possible to feed multiple
6387 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01006388 another protocol or access method. Using multiple store-request rules with
6389 the same table is possible and may be used to find the best criterion to rely
6390 on, by arranging the rules by decreasing preference order. Only the first
6391 extracted criterion for a given table will be stored. All subsequent store-
6392 request rules referencing the same table will be skipped and their ACLs will
6393 not be evaluated.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006394
6395 The "store-request" rules are evaluated once the server connection has been
6396 established, so that the table will contain the real server that processed
6397 the request.
6398
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006399 Note : Consider not using this feature in multi-process mode (nbproc > 1)
6400 unless you know what you do : memory is not shared between the
6401 processes, which can result in random behaviours.
6402
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006403 Example :
6404 # forward SMTP users to the same server they just used for POP in the
6405 # last 30 minutes
6406 backend pop
6407 mode tcp
6408 balance roundrobin
6409 stick store-request src
6410 stick-table type ip size 200k expire 30m
6411 server s1 192.168.1.1:110
6412 server s2 192.168.1.1:110
6413
6414 backend smtp
6415 mode tcp
6416 balance roundrobin
6417 stick match src table pop
6418 server s1 192.168.1.1:25
6419 server s2 192.168.1.1:25
6420
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006421 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
6422 about ACLs and pattern extraction.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006423
6424
Emeric Brun7c6b82e2010-09-24 16:34:28 +02006425stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Emeric Brunf099e792010-09-27 12:05:28 +02006426 size <size> [expire <expire>] [nopurge] [peers <peersect>]
6427 [store <data_type>]*
Godbach64cef792013-12-04 16:08:22 +08006428 Configure the stickiness table for the current section
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006429 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +02006430 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006431
6432 Arguments :
6433 ip a table declared with "type ip" will only store IPv4 addresses.
6434 This form is very compact (about 50 bytes per entry) and allows
6435 very fast entry lookup and stores with almost no overhead. This
6436 is mainly used to store client source IP addresses.
6437
David du Colombier9a6d3c92011-03-17 10:40:24 +01006438 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
6439 This form is very compact (about 60 bytes per entry) and allows
6440 very fast entry lookup and stores with almost no overhead. This
6441 is mainly used to store client source IP addresses.
6442
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006443 integer a table declared with "type integer" will store 32bit integers
6444 which can represent a client identifier found in a request for
6445 instance.
6446
6447 string a table declared with "type string" will store substrings of up
6448 to <len> characters. If the string provided by the pattern
6449 extractor is larger than <len>, it will be truncated before
6450 being stored. During matching, at most <len> characters will be
6451 compared between the string in the table and the extracted
6452 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +02006453 to 32 characters.
6454
6455 binary a table declared with "type binary" will store binary blocks
6456 of <len> bytes. If the block provided by the pattern
6457 extractor is larger than <len>, it will be truncated before
6458 being stored. If the block provided by the pattern extractor
6459 is shorter than <len>, it will be padded by 0. When not
6460 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006461
6462 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +02006463 "string" type table (See type "string" above). Or the number
6464 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006465 changing this parameter as memory usage will proportionally
6466 increase.
6467
6468 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +01006469 value directly impacts memory usage. Count approximately
6470 50 bytes per entry, plus the size of a string if any. The size
6471 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006472
6473 [nopurge] indicates that we refuse to purge older entries when the table
6474 is full. When not specified and the table is full when haproxy
6475 wants to store an entry in it, it will flush a few of the oldest
6476 entries in order to release some space for the new ones. This is
6477 most often the desired behaviour. In some specific cases, it
6478 be desirable to refuse new entries instead of purging the older
6479 ones. That may be the case when the amount of data to store is
6480 far above the hardware limits and we prefer not to offer access
6481 to new clients than to reject the ones already connected. When
6482 using this parameter, be sure to properly set the "expire"
6483 parameter (see below).
6484
Emeric Brunf099e792010-09-27 12:05:28 +02006485 <peersect> is the name of the peers section to use for replication. Entries
6486 which associate keys to server IDs are kept synchronized with
6487 the remote peers declared in this section. All entries are also
6488 automatically learned from the local peer (old process) during a
6489 soft restart.
6490
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006491 NOTE : peers can't be used in multi-process mode.
6492
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006493 <expire> defines the maximum duration of an entry in the table since it
6494 was last created, refreshed or matched. The expiration delay is
6495 defined using the standard time format, similarly as the various
6496 timeouts. The maximum duration is slightly above 24 days. See
6497 section 2.2 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +02006498 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006499 be removed once full. Be sure not to use the "nopurge" parameter
6500 if not expiration delay is specified.
6501
Willy Tarreau08d5f982010-06-06 13:34:54 +02006502 <data_type> is used to store additional information in the stick-table. This
6503 may be used by ACLs in order to control various criteria related
6504 to the activity of the client matching the stick-table. For each
6505 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +02006506 that the additional data can fit. Several data types may be
6507 stored with an entry. Multiple data types may be specified after
6508 the "store" keyword, as a comma-separated list. Alternatively,
6509 it is possible to repeat the "store" keyword followed by one or
6510 several data types. Except for the "server_id" type which is
6511 automatically detected and enabled, all data types must be
6512 explicitly declared to be stored. If an ACL references a data
6513 type which is not stored, the ACL will simply not match. Some
6514 data types require an argument which must be passed just after
6515 the type between parenthesis. See below for the supported data
6516 types and their arguments.
6517
6518 The data types that can be stored with an entry are the following :
6519 - server_id : this is an integer which holds the numeric ID of the server a
6520 request was assigned to. It is used by the "stick match", "stick store",
6521 and "stick on" rules. It is automatically enabled when referenced.
6522
6523 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
6524 integer which may be used for anything. Most of the time it will be used
6525 to put a special tag on some entries, for instance to note that a
6526 specific behaviour was detected and must be known for future matches.
6527
Willy Tarreauba2ffd12013-05-29 15:54:14 +02006528 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
6529 over a period. It is a positive 32-bit integer integer which may be used
6530 for anything. Just like <gpc0>, it counts events, but instead of keeping
6531 a cumulative count, it maintains the rate at which the counter is
6532 incremented. Most of the time it will be used to measure the frequency of
6533 occurrence of certain events (eg: requests to a specific URL).
6534
Willy Tarreauc9705a12010-07-27 20:05:50 +02006535 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
6536 the absolute number of connections received from clients which matched
6537 this entry. It does not mean the connections were accepted, just that
6538 they were received.
6539
6540 - conn_cur : Current Connections. It is a positive 32-bit integer which
6541 stores the concurrent connection counts for the entry. It is incremented
6542 once an incoming connection matches the entry, and decremented once the
6543 connection leaves. That way it is possible to know at any time the exact
6544 number of concurrent connections for an entry.
6545
6546 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
6547 integer parameter <period> which indicates in milliseconds the length
6548 of the period over which the average is measured. It reports the average
6549 incoming connection rate over that period, in connections per period. The
6550 result is an integer which can be matched using ACLs.
6551
6552 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
6553 the absolute number of sessions received from clients which matched this
6554 entry. A session is a connection that was accepted by the layer 4 rules.
6555
6556 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
6557 integer parameter <period> which indicates in milliseconds the length
6558 of the period over which the average is measured. It reports the average
6559 incoming session rate over that period, in sessions per period. The
6560 result is an integer which can be matched using ACLs.
6561
6562 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
6563 counts the absolute number of HTTP requests received from clients which
6564 matched this entry. It does not matter whether they are valid requests or
6565 not. Note that this is different from sessions when keep-alive is used on
6566 the client side.
6567
6568 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
6569 integer parameter <period> which indicates in milliseconds the length
6570 of the period over which the average is measured. It reports the average
6571 HTTP request rate over that period, in requests per period. The result is
6572 an integer which can be matched using ACLs. It does not matter whether
6573 they are valid requests or not. Note that this is different from sessions
6574 when keep-alive is used on the client side.
6575
6576 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
6577 counts the absolute number of HTTP requests errors induced by clients
6578 which matched this entry. Errors are counted on invalid and truncated
6579 requests, as well as on denied or tarpitted requests, and on failed
6580 authentications. If the server responds with 4xx, then the request is
6581 also counted as an error since it's an error triggered by the client
6582 (eg: vulnerability scan).
6583
6584 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
6585 integer parameter <period> which indicates in milliseconds the length
6586 of the period over which the average is measured. It reports the average
6587 HTTP request error rate over that period, in requests per period (see
6588 http_err_cnt above for what is accounted as an error). The result is an
6589 integer which can be matched using ACLs.
6590
6591 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
6592 integer which counts the cumulated amount of bytes received from clients
6593 which matched this entry. Headers are included in the count. This may be
6594 used to limit abuse of upload features on photo or video servers.
6595
6596 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
6597 integer parameter <period> which indicates in milliseconds the length
6598 of the period over which the average is measured. It reports the average
6599 incoming bytes rate over that period, in bytes per period. It may be used
6600 to detect users which upload too much and too fast. Warning: with large
6601 uploads, it is possible that the amount of uploaded data will be counted
6602 once upon termination, thus causing spikes in the average transfer speed
6603 instead of having a smooth one. This may partially be smoothed with
6604 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
6605 recommended for better fairness.
6606
6607 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
6608 integer which counts the cumulated amount of bytes sent to clients which
6609 matched this entry. Headers are included in the count. This may be used
6610 to limit abuse of bots sucking the whole site.
6611
6612 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
6613 an integer parameter <period> which indicates in milliseconds the length
6614 of the period over which the average is measured. It reports the average
6615 outgoing bytes rate over that period, in bytes per period. It may be used
6616 to detect users which download too much and too fast. Warning: with large
6617 transfers, it is possible that the amount of transferred data will be
6618 counted once upon termination, thus causing spikes in the average
6619 transfer speed instead of having a smooth one. This may partially be
6620 smoothed with "option contstats" though this is not perfect yet. Use of
6621 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +02006622
Willy Tarreauc00cdc22010-06-06 16:48:26 +02006623 There is only one stick-table per proxy. At the moment of writing this doc,
6624 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006625 to be required, simply create a dummy backend with a stick-table in it and
6626 reference it.
6627
6628 It is important to understand that stickiness based on learning information
6629 has some limitations, including the fact that all learned associations are
6630 lost upon restart. In general it can be good as a complement but not always
6631 as an exclusive stickiness.
6632
Willy Tarreauc9705a12010-07-27 20:05:50 +02006633 Last, memory requirements may be important when storing many data types.
6634 Indeed, storing all indicators above at once in each entry requires 116 bytes
6635 per entry, or 116 MB for a 1-million entries table. This is definitely not
6636 something that can be ignored.
6637
6638 Example:
6639 # Keep track of counters of up to 1 million IP addresses over 5 minutes
6640 # and store a general purpose counter and the average connection rate
6641 # computed over a sliding window of 30 seconds.
6642 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
6643
6644 See also : "stick match", "stick on", "stick store-request", section 2.2
David du Colombiera13d1b92011-03-17 10:40:22 +01006645 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006646
6647
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006648stick store-response <pattern> [table <table>] [{if | unless} <condition>]
6649 Define a request pattern used to create an entry in a stickiness table
6650 May be used in sections : defaults | frontend | listen | backend
6651 no | no | yes | yes
6652
6653 Arguments :
Willy Tarreau74ca5042013-06-11 23:12:07 +02006654 <pattern> is a pattern extraction rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006655 describes what elements of the response or connection will
6656 be analysed, extracted and stored in the table once a
6657 server is selected.
6658
6659 <table> is an optional stickiness table name. If unspecified, the same
6660 backend's table is used. A stickiness table is declared using
6661 the "stick-table" statement.
6662
6663 <cond> is an optional storage condition. It makes it possible to store
6664 certain criteria only when some conditions are met (or not met).
6665 For instance, it could be used to store the SSL session ID only
6666 when the response is a SSL server hello.
6667
6668 Some protocols or applications require complex stickiness rules and cannot
6669 always simply rely on cookies nor hashing. The "stick store-response"
6670 statement describes a rule to decide what to extract from the response and
6671 when to do it, in order to store it into a stickiness table for further
6672 requests to match it using the "stick match" statement. Obviously the
6673 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006674 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006675 See section 7 for a complete list of possible patterns and transformation
6676 rules.
6677
6678 The table has to be declared using the "stick-table" statement. It must be of
6679 a type compatible with the pattern. By default it is the one which is present
6680 in the same backend. It is possible to share a table with other backends by
6681 referencing it using the "table" keyword. If another table is referenced,
6682 the server's ID inside the backends are used. By default, all server IDs
6683 start at 1 in each backend, so the server ordering is enough. But in case of
6684 doubt, it is highly recommended to force server IDs using their "id" setting.
6685
6686 It is possible to restrict the conditions where a "stick store-response"
6687 statement will apply, using "if" or "unless" followed by a condition. This
6688 condition will be evaluated while parsing the response, so any criteria can
6689 be used. See section 7 for ACL based conditions.
6690
6691 There is no limit on the number of "stick store-response" statements, but
6692 there is a limit of 8 simultaneous stores per request or response. This
6693 makes it possible to store up to 8 criteria, all extracted from either the
6694 request or the response, regardless of the number of rules. Only the 8 first
6695 ones which match will be kept. Using this, it is possible to feed multiple
6696 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01006697 another protocol or access method. Using multiple store-response rules with
6698 the same table is possible and may be used to find the best criterion to rely
6699 on, by arranging the rules by decreasing preference order. Only the first
6700 extracted criterion for a given table will be stored. All subsequent store-
6701 response rules referencing the same table will be skipped and their ACLs will
6702 not be evaluated. However, even if a store-request rule references a table, a
6703 store-response rule may also use the same table. This means that each table
6704 may learn exactly one element from the request and one element from the
6705 response at once.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006706
6707 The table will contain the real server that processed the request.
6708
6709 Example :
6710 # Learn SSL session ID from both request and response and create affinity.
6711 backend https
6712 mode tcp
6713 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +02006714 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006715 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006716
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006717 acl clienthello req_ssl_hello_type 1
6718 acl serverhello rep_ssl_hello_type 2
6719
6720 # use tcp content accepts to detects ssl client and server hello.
6721 tcp-request inspect-delay 5s
6722 tcp-request content accept if clienthello
6723
6724 # no timeout on response inspect delay by default.
6725 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006726
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006727 # SSL session ID (SSLID) may be present on a client or server hello.
6728 # Its length is coded on 1 byte at offset 43 and its value starts
6729 # at offset 44.
6730
6731 # Match and learn on request if client hello.
6732 stick on payload_lv(43,1) if clienthello
6733
6734 # Learn on response if server hello.
6735 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +02006736
Emeric Brun6a1cefa2010-09-24 18:15:17 +02006737 server s1 192.168.1.1:443
6738 server s2 192.168.1.1:443
6739
6740 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
6741 extraction.
6742
6743
Willy Tarreaue9656522010-08-17 15:40:09 +02006744tcp-request connection <action> [{if | unless} <condition>]
6745 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +02006746 May be used in sections : defaults | frontend | listen | backend
6747 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +02006748 Arguments :
6749 <action> defines the action to perform if the condition applies. Valid
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006750 actions include : "accept", "reject", "track-sc0", "track-sc1",
6751 "track-sc2", and "expect-proxy". See below for more details.
Willy Tarreau1a687942010-05-23 22:40:30 +02006752
Willy Tarreaue9656522010-08-17 15:40:09 +02006753 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006754
6755 Immediately after acceptance of a new incoming connection, it is possible to
6756 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +02006757 or dropped or have its counters tracked. Those conditions cannot make use of
6758 any data contents because the connection has not been read from yet, and the
6759 buffers are not yet allocated. This is used to selectively and very quickly
6760 accept or drop connections from various sources with a very low overhead. If
6761 some contents need to be inspected in order to take the decision, the
6762 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006763
Willy Tarreaue9656522010-08-17 15:40:09 +02006764 The "tcp-request connection" rules are evaluated in their exact declaration
6765 order. If no rule matches or if there is no rule, the default action is to
6766 accept the incoming connection. There is no specific limit to the number of
6767 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006768
Willy Tarreaue9656522010-08-17 15:40:09 +02006769 Three types of actions are supported :
6770 - accept :
6771 accepts the connection if the condition is true (when used with "if")
6772 or false (when used with "unless"). The first such rule executed ends
6773 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006774
Willy Tarreaue9656522010-08-17 15:40:09 +02006775 - reject :
6776 rejects the connection if the condition is true (when used with "if")
6777 or false (when used with "unless"). The first such rule executed ends
6778 the rules evaluation. Rejected connections do not even become a
6779 session, which is why they are accounted separately for in the stats,
6780 as "denied connections". They are not considered for the session
6781 rate-limit and are not logged either. The reason is that these rules
6782 should only be used to filter extremely high connection rates such as
6783 the ones encountered during a massive DDoS attack. Under these extreme
6784 conditions, the simple action of logging each event would make the
6785 system collapse and would considerably lower the filtering capacity. If
6786 logging is absolutely desired, then "tcp-request content" rules should
6787 be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006788
Willy Tarreau4f0d9192013-06-11 20:40:55 +02006789 - expect-proxy layer4 :
6790 configures the client-facing connection to receive a PROXY protocol
6791 header before any byte is read from the socket. This is equivalent to
6792 having the "accept-proxy" keyword on the "bind" line, except that using
6793 the TCP rule allows the PROXY protocol to be accepted only for certain
6794 IP address ranges using an ACL. This is convenient when multiple layers
6795 of load balancers are passed through by traffic coming from public
6796 hosts.
6797
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006798 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +02006799 enables tracking of sticky counters from current connection. These
6800 rules do not stop evaluation and do not change default action. Two sets
6801 of counters may be simultaneously tracked by the same connection. The
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006802 first "track-sc0" rule executed enables tracking of the counters of the
6803 specified table as the first set. The first "track-sc1" rule executed
Willy Tarreaue9656522010-08-17 15:40:09 +02006804 enables tracking of the counters of the specified table as the second
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006805 set. The first "track-sc2" rule executed enables tracking of the
6806 counters of the specified table as the third set. It is a recommended
6807 practice to use the first set of counters for the per-frontend counters
6808 and the second set for the per-backend ones. But this is just a
6809 guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006810
Willy Tarreaue9656522010-08-17 15:40:09 +02006811 These actions take one or two arguments :
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006812 <key> is mandatory, and is a pattern extraction rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +02006813 in section 7.3. It describes what elements of the incoming
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006814 request or connection will be analysed, extracted, combined,
6815 and used to select which table entry to update the counters.
6816 Note that "tcp-request connection" cannot use content-based
6817 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006818
Willy Tarreaue9656522010-08-17 15:40:09 +02006819 <table> is an optional table to be used instead of the default one,
6820 which is the stick-table declared in the current proxy. All
6821 the counters for the matches and updates for the key will
6822 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006823
Willy Tarreaue9656522010-08-17 15:40:09 +02006824 Once a "track-sc*" rule is executed, the key is looked up in the table
6825 and if it is not found, an entry is allocated for it. Then a pointer to
6826 that entry is kept during all the session's life, and this entry's
6827 counters are updated as often as possible, every time the session's
6828 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006829 Counters are only updated for events that happen after the tracking has
6830 been started. For example, connection counters will not be updated when
6831 tracking layer 7 information, since the connection event happens before
6832 layer7 information is extracted.
6833
Willy Tarreaue9656522010-08-17 15:40:09 +02006834 If the entry tracks concurrent connection counters, one connection is
6835 counted for as long as the entry is tracked, and the entry will not
6836 expire during that time. Tracking counters also provides a performance
6837 advantage over just checking the keys, because only one table lookup is
6838 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006839
Willy Tarreaue9656522010-08-17 15:40:09 +02006840 Note that the "if/unless" condition is optional. If no condition is set on
6841 the action, it is simply performed unconditionally. That can be useful for
6842 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006843
Willy Tarreaue9656522010-08-17 15:40:09 +02006844 Example: accept all connections from white-listed hosts, reject too fast
6845 connection without counting them, and track accepted connections.
6846 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006847
Willy Tarreaue9656522010-08-17 15:40:09 +02006848 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006849 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006850 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006851
Willy Tarreaue9656522010-08-17 15:40:09 +02006852 Example: accept all connections from white-listed hosts, count all other
6853 connections and reject too fast ones. This results in abusive ones
6854 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006855
Willy Tarreaue9656522010-08-17 15:40:09 +02006856 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006857 tcp-request connection track-sc0 src
6858 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006859
Willy Tarreau4f0d9192013-06-11 20:40:55 +02006860 Example: enable the PROXY protocol for traffic coming from all known proxies.
6861
6862 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
6863
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006864 See section 7 about ACL usage.
6865
Willy Tarreaue9656522010-08-17 15:40:09 +02006866 See also : "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006867
6868
Willy Tarreaue9656522010-08-17 15:40:09 +02006869tcp-request content <action> [{if | unless} <condition>]
6870 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +02006871 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02006872 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +02006873 Arguments :
6874 <action> defines the action to perform if the condition applies. Valid
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006875 actions include : "accept", "reject", "track-sc0", "track-sc1",
6876 and "track-sc2". See "tcp-request connection" above for their
Willy Tarreaue25c9172013-05-28 18:32:20 +02006877 signification.
Willy Tarreau62644772008-07-16 18:36:06 +02006878
Willy Tarreaue9656522010-08-17 15:40:09 +02006879 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +02006880
Willy Tarreaue9656522010-08-17 15:40:09 +02006881 A request's contents can be analysed at an early stage of request processing
6882 called "TCP content inspection". During this stage, ACL-based rules are
6883 evaluated every time the request contents are updated, until either an
6884 "accept" or a "reject" rule matches, or the TCP request inspection delay
6885 expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +02006886
Willy Tarreaue9656522010-08-17 15:40:09 +02006887 The first difference between these rules and "tcp-request connection" rules
6888 is that "tcp-request content" rules can make use of contents to take a
6889 decision. Most often, these decisions will consider a protocol recognition or
6890 validity. The second difference is that content-based rules can be used in
6891 both frontends and backends. In frontends, they will be evaluated upon new
6892 connections. In backends, they will be evaluated once a session is assigned
6893 a backend. This means that a single frontend connection may be evaluated
6894 several times by one or multiple backends when a session gets reassigned
6895 (for instance after a client-side HTTP keep-alive request).
Willy Tarreau62644772008-07-16 18:36:06 +02006896
Willy Tarreaue9656522010-08-17 15:40:09 +02006897 Content-based rules are evaluated in their exact declaration order. If no
6898 rule matches or if there is no rule, the default action is to accept the
6899 contents. There is no specific limit to the number of rules which may be
6900 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +02006901
Willy Tarreaue9656522010-08-17 15:40:09 +02006902 Three types of actions are supported :
6903 - accept :
6904 - reject :
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006905 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Willy Tarreau62644772008-07-16 18:36:06 +02006906
Willy Tarreaue9656522010-08-17 15:40:09 +02006907 They have the same meaning as their counter-parts in "tcp-request connection"
6908 so please refer to that section for a complete description.
Willy Tarreau62644772008-07-16 18:36:06 +02006909
Willy Tarreaue9656522010-08-17 15:40:09 +02006910 Also, it is worth noting that if sticky counters are tracked from a rule
6911 defined in a backend, this tracking will automatically end when the session
6912 releases the backend. That allows per-backend counter tracking even in case
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006913 of HTTP keep-alive requests when the backend changes. This makes a subtle
6914 difference because tracking rules in "frontend" and "listen" section last for
6915 all the session, as opposed to the backend rules. The difference appears when
6916 some layer 7 information is tracked. While there is nothing mandatory about
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006917 it, it is recommended to use the track-sc0 pointer to track per-frontend
6918 counters and track-sc1 to track per-backend counters, but this is just a
Willy Tarreaue25c9172013-05-28 18:32:20 +02006919 guideline and all counters may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +02006920
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006921 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +02006922 the action, it is simply performed unconditionally. That can be useful for
6923 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +02006924
Willy Tarreaue9656522010-08-17 15:40:09 +02006925 It is perfectly possible to match layer 7 contents with "tcp-request content"
Willy Tarreauc0239e02012-04-16 14:42:55 +02006926 rules, since HTTP-specific ACL matches are able to preliminarily parse the
6927 contents of a buffer before extracting the required data. If the buffered
6928 contents do not parse as a valid HTTP message, then the ACL does not match.
6929 The parser which is involved there is exactly the same as for all other HTTP
6930 processing, so there is no risk of parsing something differently.
Willy Tarreau62644772008-07-16 18:36:06 +02006931
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006932 Tracking layer7 information is also possible provided that the information
6933 are present when the rule is processed. The current solution for making the
6934 rule engine wait for such information is to set an inspect delay and to
6935 condition its execution with an ACL relying on such information.
6936
Willy Tarreau62644772008-07-16 18:36:06 +02006937 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +02006938 # Accept HTTP requests containing a Host header saying "example.com"
6939 # and reject everything else.
6940 acl is_host_com hdr(Host) -i example.com
6941 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +02006942 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +02006943 tcp-request content reject
6944
6945 Example:
Willy Tarreau62644772008-07-16 18:36:06 +02006946 # reject SMTP connection if client speaks first
6947 tcp-request inspect-delay 30s
6948 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006949 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +02006950
6951 # Forward HTTPS connection only if client speaks
6952 tcp-request inspect-delay 30s
6953 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02006954 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +02006955 tcp-request content reject
6956
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006957 Example:
6958 # Track the last IP from X-Forwarded-For
6959 tcp-request inspect-delay 10s
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006960 tcp-request content track-sc0 hdr(x-forwarded-for,-1) if HTTP
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006961
6962 Example:
6963 # track request counts per "base" (concatenation of Host+URL)
6964 tcp-request inspect-delay 10s
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006965 tcp-request content track-sc0 base table req-rate if HTTP
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01006966
Willy Tarreaue9656522010-08-17 15:40:09 +02006967 Example: track per-frontend and per-backend counters, block abusers at the
6968 frontend when the backend detects abuse.
6969
6970 frontend http
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006971 # Use General Purpose Couter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +02006972 # protecting all our sites
6973 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006974 tcp-request connection track-sc0 src
6975 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +02006976 ...
6977 use_backend http_dynamic if { path_end .php }
6978
6979 backend http_dynamic
6980 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006981 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +02006982 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02006983 acl click_too_fast sc1_http_req_rate gt 10
6984 acl mark_as_abuser sc0_inc_gpc0 gt 0
6985 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +02006986 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +02006987
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006988 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +02006989
Willy Tarreaue9656522010-08-17 15:40:09 +02006990 See also : "tcp-request connection", "tcp-request inspect-delay"
Willy Tarreau62644772008-07-16 18:36:06 +02006991
6992
6993tcp-request inspect-delay <timeout>
6994 Set the maximum allowed time to wait for data during content inspection
6995 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02006996 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +02006997 Arguments :
6998 <timeout> is the timeout value specified in milliseconds by default, but
6999 can be in any other unit if the number is suffixed by the unit,
7000 as explained at the top of this document.
7001
7002 People using haproxy primarily as a TCP relay are often worried about the
7003 risk of passing any type of protocol to a server without any analysis. In
7004 order to be able to analyze the request contents, we must first withhold
7005 the data then analyze them. This statement simply enables withholding of
7006 data for at most the specified amount of time.
7007
Willy Tarreaufb356202010-08-03 14:02:05 +02007008 TCP content inspection applies very early when a connection reaches a
7009 frontend, then very early when the connection is forwarded to a backend. This
7010 means that a connection may experience a first delay in the frontend and a
7011 second delay in the backend if both have tcp-request rules.
7012
Willy Tarreau62644772008-07-16 18:36:06 +02007013 Note that when performing content inspection, haproxy will evaluate the whole
7014 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007015 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +02007016 a last check is performed upon expiration, this time considering that the
Willy Tarreaud869b242009-03-15 14:43:58 +01007017 contents are definitive. If no delay is set, haproxy will not wait at all
7018 and will immediately apply a verdict based on the available information.
7019 Obviously this is unlikely to be very useful and might even be racy, so such
7020 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +02007021
7022 As soon as a rule matches, the request is released and continues as usual. If
7023 the timeout is reached and no rule matches, the default policy will be to let
7024 it pass through unaffected.
7025
7026 For most protocols, it is enough to set it to a few seconds, as most clients
7027 send the full request immediately upon connection. Add 3 or more seconds to
7028 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +01007029 to use large values, for instance to ensure that the client never talks
Willy Tarreau62644772008-07-16 18:36:06 +02007030 before the server (eg: SMTP), or to wait for a client to talk before passing
7031 data to the server (eg: SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +02007032 least the inspection delay, otherwise it will expire first. If the client
7033 closes the connection or if the buffer is full, the delay immediately expires
7034 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +02007035
Willy Tarreau55165fe2009-05-10 12:02:55 +02007036 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +02007037 "timeout client".
7038
7039
Emeric Brun0a3b67f2010-09-24 15:34:53 +02007040tcp-response content <action> [{if | unless} <condition>]
7041 Perform an action on a session response depending on a layer 4-7 condition
7042 May be used in sections : defaults | frontend | listen | backend
7043 no | no | yes | yes
7044 Arguments :
7045 <action> defines the action to perform if the condition applies. Valid
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02007046 actions include : "accept", "close", "reject".
Emeric Brun0a3b67f2010-09-24 15:34:53 +02007047
7048 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
7049
7050 Response contents can be analysed at an early stage of response processing
7051 called "TCP content inspection". During this stage, ACL-based rules are
7052 evaluated every time the response contents are updated, until either an
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02007053 "accept", "close" or a "reject" rule matches, or a TCP response inspection
7054 delay is set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02007055
7056 Most often, these decisions will consider a protocol recognition or validity.
7057
7058 Content-based rules are evaluated in their exact declaration order. If no
7059 rule matches or if there is no rule, the default action is to accept the
7060 contents. There is no specific limit to the number of rules which may be
7061 inserted.
7062
7063 Two types of actions are supported :
7064 - accept :
7065 accepts the response if the condition is true (when used with "if")
7066 or false (when used with "unless"). The first such rule executed ends
7067 the rules evaluation.
7068
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02007069 - close :
7070 immediately closes the connection with the server if the condition is
7071 true (when used with "if"), or false (when used with "unless"). The
7072 first such rule executed ends the rules evaluation. The main purpose of
7073 this action is to force a connection to be finished between a client
7074 and a server after an exchange when the application protocol expects
7075 some long time outs to elapse first. The goal is to eliminate idle
7076 connections which take signifiant resources on servers with certain
7077 protocols.
7078
Emeric Brun0a3b67f2010-09-24 15:34:53 +02007079 - reject :
7080 rejects the response if the condition is true (when used with "if")
7081 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007082 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02007083
7084 Note that the "if/unless" condition is optional. If no condition is set on
7085 the action, it is simply performed unconditionally. That can be useful for
7086 for changing the default action to a reject.
7087
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007088 It is perfectly possible to match layer 7 contents with "tcp-response
7089 content" rules, but then it is important to ensure that a full response has
7090 been buffered, otherwise no contents will match. In order to achieve this,
7091 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +02007092 period.
7093
7094 See section 7 about ACL usage.
7095
7096 See also : "tcp-request content", "tcp-response inspect-delay"
7097
7098
7099tcp-response inspect-delay <timeout>
7100 Set the maximum allowed time to wait for a response during content inspection
7101 May be used in sections : defaults | frontend | listen | backend
7102 no | no | yes | yes
7103 Arguments :
7104 <timeout> is the timeout value specified in milliseconds by default, but
7105 can be in any other unit if the number is suffixed by the unit,
7106 as explained at the top of this document.
7107
7108 See also : "tcp-response content", "tcp-request inspect-delay".
7109
7110
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01007111timeout check <timeout>
7112 Set additional check timeout, but only after a connection has been already
7113 established.
7114
7115 May be used in sections: defaults | frontend | listen | backend
7116 yes | no | yes | yes
7117 Arguments:
7118 <timeout> is the timeout value specified in milliseconds by default, but
7119 can be in any other unit if the number is suffixed by the unit,
7120 as explained at the top of this document.
7121
7122 If set, haproxy uses min("timeout connect", "inter") as a connect timeout
7123 for check and "timeout check" as an additional read timeout. The "min" is
7124 used so that people running with *very* long "timeout connect" (eg. those
7125 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +01007126 (Please also note that there is no valid reason to have such long connect
7127 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
7128 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01007129
7130 If "timeout check" is not set haproxy uses "inter" for complete check
7131 timeout (connect + read) exactly like all <1.3.15 version.
7132
7133 In most cases check request is much simpler and faster to handle than normal
7134 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +01007135 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01007136
7137 This parameter is specific to backends, but can be specified once for all in
7138 "defaults" sections. This is in fact one of the easiest solutions not to
7139 forget about it.
7140
Willy Tarreau41a340d2008-01-22 12:25:31 +01007141 See also: "timeout connect", "timeout queue", "timeout server",
7142 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01007143
7144
Willy Tarreau0ba27502007-12-24 16:55:16 +01007145timeout client <timeout>
7146timeout clitimeout <timeout> (deprecated)
7147 Set the maximum inactivity time on the client side.
7148 May be used in sections : defaults | frontend | listen | backend
7149 yes | yes | yes | no
7150 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01007151 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01007152 can be in any other unit if the number is suffixed by the unit,
7153 as explained at the top of this document.
7154
7155 The inactivity timeout applies when the client is expected to acknowledge or
7156 send data. In HTTP mode, this timeout is particularly important to consider
7157 during the first phase, when the client sends the request, and during the
7158 response while it is reading data sent by the server. The value is specified
7159 in milliseconds by default, but can be in any other unit if the number is
7160 suffixed by the unit, as specified at the top of this document. In TCP mode
7161 (and to a lesser extent, in HTTP mode), it is highly recommended that the
7162 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007163 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01007164 losses by specifying timeouts that are slightly above multiples of 3 seconds
Willy Tarreauce887fd2012-05-12 12:50:00 +02007165 (eg: 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
7166 sessions (eg: WebSocket and HTTP), it's worth considering "timeout tunnel",
7167 which overrides "timeout client" and "timeout server" for tunnels.
Willy Tarreau0ba27502007-12-24 16:55:16 +01007168
7169 This parameter is specific to frontends, but can be specified once for all in
7170 "defaults" sections. This is in fact one of the easiest solutions not to
7171 forget about it. An unspecified timeout results in an infinite timeout, which
7172 is not recommended. Such a usage is accepted and works but reports a warning
7173 during startup because it may results in accumulation of expired sessions in
7174 the system if the system's timeouts are not configured either.
7175
7176 This parameter replaces the old, deprecated "clitimeout". It is recommended
7177 to use it to write new configurations. The form "timeout clitimeout" is
7178 provided only by backwards compatibility but its use is strongly discouraged.
7179
Willy Tarreauce887fd2012-05-12 12:50:00 +02007180 See also : "clitimeout", "timeout server", "timeout tunnel".
Willy Tarreau0ba27502007-12-24 16:55:16 +01007181
7182
7183timeout connect <timeout>
7184timeout contimeout <timeout> (deprecated)
7185 Set the maximum time to wait for a connection attempt to a server to succeed.
7186 May be used in sections : defaults | frontend | listen | backend
7187 yes | no | yes | yes
7188 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01007189 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01007190 can be in any other unit if the number is suffixed by the unit,
7191 as explained at the top of this document.
7192
7193 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007194 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01007195 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01007196 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01007197 connect timeout also presets both queue and tarpit timeouts to the same value
7198 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +01007199
7200 This parameter is specific to backends, but can be specified once for all in
7201 "defaults" sections. This is in fact one of the easiest solutions not to
7202 forget about it. An unspecified timeout results in an infinite timeout, which
7203 is not recommended. Such a usage is accepted and works but reports a warning
7204 during startup because it may results in accumulation of failed sessions in
7205 the system if the system's timeouts are not configured either.
7206
7207 This parameter replaces the old, deprecated "contimeout". It is recommended
7208 to use it to write new configurations. The form "timeout contimeout" is
7209 provided only by backwards compatibility but its use is strongly discouraged.
7210
Willy Tarreau41a340d2008-01-22 12:25:31 +01007211 See also: "timeout check", "timeout queue", "timeout server", "contimeout",
7212 "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +01007213
7214
Willy Tarreaub16a5742010-01-10 14:46:16 +01007215timeout http-keep-alive <timeout>
7216 Set the maximum allowed time to wait for a new HTTP request to appear
7217 May be used in sections : defaults | frontend | listen | backend
7218 yes | yes | yes | yes
7219 Arguments :
7220 <timeout> is the timeout value specified in milliseconds by default, but
7221 can be in any other unit if the number is suffixed by the unit,
7222 as explained at the top of this document.
7223
7224 By default, the time to wait for a new request in case of keep-alive is set
7225 by "timeout http-request". However this is not always convenient because some
7226 people want very short keep-alive timeouts in order to release connections
7227 faster, and others prefer to have larger ones but still have short timeouts
7228 once the request has started to present itself.
7229
7230 The "http-keep-alive" timeout covers these needs. It will define how long to
7231 wait for a new HTTP request to start coming after a response was sent. Once
7232 the first byte of request has been seen, the "http-request" timeout is used
7233 to wait for the complete request to come. Note that empty lines prior to a
7234 new request do not refresh the timeout and are not counted as a new request.
7235
7236 There is also another difference between the two timeouts : when a connection
7237 expires during timeout http-keep-alive, no error is returned, the connection
7238 just closes. If the connection expires in "http-request" while waiting for a
7239 connection to complete, a HTTP 408 error is returned.
7240
7241 In general it is optimal to set this value to a few tens to hundreds of
7242 milliseconds, to allow users to fetch all objects of a page at once but
7243 without waiting for further clicks. Also, if set to a very small value (eg:
7244 1 millisecond) it will probably only accept pipelined requests but not the
7245 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +02007246 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +01007247
7248 If this parameter is not set, the "http-request" timeout applies, and if both
7249 are not set, "timeout client" still applies at the lower level. It should be
7250 set in the frontend to take effect, unless the frontend is in TCP mode, in
7251 which case the HTTP backend's timeout will be used.
7252
7253 See also : "timeout http-request", "timeout client".
7254
7255
Willy Tarreau036fae02008-01-06 13:24:40 +01007256timeout http-request <timeout>
7257 Set the maximum allowed time to wait for a complete HTTP request
7258 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +02007259 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +01007260 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01007261 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +01007262 can be in any other unit if the number is suffixed by the unit,
7263 as explained at the top of this document.
7264
7265 In order to offer DoS protection, it may be required to lower the maximum
7266 accepted time to receive a complete HTTP request without affecting the client
7267 timeout. This helps protecting against established connections on which
7268 nothing is sent. The client timeout cannot offer a good protection against
7269 this abuse because it is an inactivity timeout, which means that if the
7270 attacker sends one character every now and then, the timeout will not
7271 trigger. With the HTTP request timeout, no matter what speed the client
7272 types, the request will be aborted if it does not complete in time.
7273
7274 Note that this timeout only applies to the header part of the request, and
7275 not to any data. As soon as the empty line is received, this timeout is not
Willy Tarreaub16a5742010-01-10 14:46:16 +01007276 used anymore. It is used again on keep-alive connections to wait for a second
7277 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +01007278
7279 Generally it is enough to set it to a few seconds, as most clients send the
7280 full request immediately upon connection. Add 3 or more seconds to cover TCP
7281 retransmits but that's all. Setting it to very low values (eg: 50 ms) will
7282 generally work on local networks as long as there are no packet losses. This
7283 will prevent people from sending bare HTTP requests using telnet.
7284
7285 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +02007286 chunk of the incoming request. It should be set in the frontend to take
7287 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
7288 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +01007289
Willy Tarreaub16a5742010-01-10 14:46:16 +01007290 See also : "timeout http-keep-alive", "timeout client".
Willy Tarreau036fae02008-01-06 13:24:40 +01007291
Willy Tarreau844e3c52008-01-11 16:28:18 +01007292
7293timeout queue <timeout>
7294 Set the maximum time to wait in the queue for a connection slot to be free
7295 May be used in sections : defaults | frontend | listen | backend
7296 yes | no | yes | yes
7297 Arguments :
7298 <timeout> is the timeout value specified in milliseconds by default, but
7299 can be in any other unit if the number is suffixed by the unit,
7300 as explained at the top of this document.
7301
7302 When a server's maxconn is reached, connections are left pending in a queue
7303 which may be server-specific or global to the backend. In order not to wait
7304 indefinitely, a timeout is applied to requests pending in the queue. If the
7305 timeout is reached, it is considered that the request will almost never be
7306 served, so it is dropped and a 503 error is returned to the client.
7307
7308 The "timeout queue" statement allows to fix the maximum time for a request to
7309 be left pending in a queue. If unspecified, the same value as the backend's
7310 connection timeout ("timeout connect") is used, for backwards compatibility
7311 with older versions with no "timeout queue" parameter.
7312
7313 See also : "timeout connect", "contimeout".
7314
7315
7316timeout server <timeout>
7317timeout srvtimeout <timeout> (deprecated)
7318 Set the maximum inactivity time on the server side.
7319 May be used in sections : defaults | frontend | listen | backend
7320 yes | no | yes | yes
7321 Arguments :
7322 <timeout> is the timeout value specified in milliseconds by default, but
7323 can be in any other unit if the number is suffixed by the unit,
7324 as explained at the top of this document.
7325
7326 The inactivity timeout applies when the server is expected to acknowledge or
7327 send data. In HTTP mode, this timeout is particularly important to consider
7328 during the first phase of the server's response, when it has to send the
7329 headers, as it directly represents the server's processing time for the
7330 request. To find out what value to put there, it's often good to start with
7331 what would be considered as unacceptable response times, then check the logs
7332 to observe the response time distribution, and adjust the value accordingly.
7333
7334 The value is specified in milliseconds by default, but can be in any other
7335 unit if the number is suffixed by the unit, as specified at the top of this
7336 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
7337 recommended that the client timeout remains equal to the server timeout in
7338 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007339 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01007340 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreauce887fd2012-05-12 12:50:00 +02007341 seconds (eg: 4 or 5 seconds minimum). If some long-lived sessions are mixed
7342 with short-lived sessions (eg: WebSocket and HTTP), it's worth considering
7343 "timeout tunnel", which overrides "timeout client" and "timeout server" for
7344 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007345
7346 This parameter is specific to backends, but can be specified once for all in
7347 "defaults" sections. This is in fact one of the easiest solutions not to
7348 forget about it. An unspecified timeout results in an infinite timeout, which
7349 is not recommended. Such a usage is accepted and works but reports a warning
7350 during startup because it may results in accumulation of expired sessions in
7351 the system if the system's timeouts are not configured either.
7352
7353 This parameter replaces the old, deprecated "srvtimeout". It is recommended
7354 to use it to write new configurations. The form "timeout srvtimeout" is
7355 provided only by backwards compatibility but its use is strongly discouraged.
7356
Willy Tarreauce887fd2012-05-12 12:50:00 +02007357 See also : "srvtimeout", "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +01007358
7359
7360timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +01007361 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +01007362 May be used in sections : defaults | frontend | listen | backend
7363 yes | yes | yes | yes
7364 Arguments :
7365 <timeout> is the tarpit duration specified in milliseconds by default, but
7366 can be in any other unit if the number is suffixed by the unit,
7367 as explained at the top of this document.
7368
7369 When a connection is tarpitted using "reqtarpit", it is maintained open with
7370 no activity for a certain amount of time, then closed. "timeout tarpit"
7371 defines how long it will be maintained open.
7372
7373 The value is specified in milliseconds by default, but can be in any other
7374 unit if the number is suffixed by the unit, as specified at the top of this
7375 document. If unspecified, the same value as the backend's connection timeout
7376 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +01007377 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007378
7379 See also : "timeout connect", "contimeout".
7380
7381
Willy Tarreauce887fd2012-05-12 12:50:00 +02007382timeout tunnel <timeout>
7383 Set the maximum inactivity time on the client and server side for tunnels.
7384 May be used in sections : defaults | frontend | listen | backend
7385 yes | no | yes | yes
7386 Arguments :
7387 <timeout> is the timeout value specified in milliseconds by default, but
7388 can be in any other unit if the number is suffixed by the unit,
7389 as explained at the top of this document.
7390
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007391 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +02007392 between a client and a server, and the connection remains inactive in both
7393 directions. This timeout supersedes both the client and server timeouts once
7394 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
7395 analyser remains attached to either connection (eg: tcp content rules are
7396 accepted). In HTTP, this timeout is used when a connection is upgraded (eg:
7397 when switching to the WebSocket protocol, or forwarding a CONNECT request
7398 to a proxy), or after the first response when no keepalive/close option is
7399 specified.
7400
7401 The value is specified in milliseconds by default, but can be in any other
7402 unit if the number is suffixed by the unit, as specified at the top of this
7403 document. Whatever the expected normal idle time, it is a good practice to
7404 cover at least one or several TCP packet losses by specifying timeouts that
7405 are slightly above multiples of 3 seconds (eg: 4 or 5 seconds minimum).
7406
7407 This parameter is specific to backends, but can be specified once for all in
7408 "defaults" sections. This is in fact one of the easiest solutions not to
7409 forget about it.
7410
7411 Example :
7412 defaults http
7413 option http-server-close
7414 timeout connect 5s
7415 timeout client 30s
7416 timeout client 30s
7417 timeout server 30s
7418 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
7419
7420 See also : "timeout client", "timeout server".
7421
7422
Willy Tarreau844e3c52008-01-11 16:28:18 +01007423transparent (deprecated)
7424 Enable client-side transparent proxying
7425 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01007426 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01007427 Arguments : none
7428
7429 This keyword was introduced in order to provide layer 7 persistence to layer
7430 3 load balancers. The idea is to use the OS's ability to redirect an incoming
7431 connection for a remote address to a local process (here HAProxy), and let
7432 this process know what address was initially requested. When this option is
7433 used, sessions without cookies will be forwarded to the original destination
7434 IP address of the incoming request (which should match that of another
7435 equipment), while requests with cookies will still be forwarded to the
7436 appropriate server.
7437
7438 The "transparent" keyword is deprecated, use "option transparent" instead.
7439
7440 Note that contrary to a common belief, this option does NOT make HAProxy
7441 present the client's IP to the server when establishing the connection.
7442
Willy Tarreau844e3c52008-01-11 16:28:18 +01007443 See also: "option transparent"
7444
William Lallemanda73203e2012-03-12 12:48:57 +01007445unique-id-format <string>
7446 Generate a unique ID for each request.
7447 May be used in sections : defaults | frontend | listen | backend
7448 yes | yes | yes | no
7449 Arguments :
7450 <string> is a log-format string.
7451
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007452 This keyword creates a ID for each request using the custom log format. A
7453 unique ID is useful to trace a request passing through many components of
7454 a complex infrastructure. The newly created ID may also be logged using the
7455 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +01007456
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007457 The format should be composed from elements that are guaranteed to be
7458 unique when combined together. For instance, if multiple haproxy instances
7459 are involved, it might be important to include the node name. It is often
7460 needed to log the incoming connection's source and destination addresses
7461 and ports. Note that since multiple requests may be performed over the same
7462 connection, including a request counter may help differentiate them.
7463 Similarly, a timestamp may protect against a rollover of the counter.
7464 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +01007465
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007466 It is recommended to use hexadecimal notation for many fields since it
7467 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +01007468
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007469 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01007470
7471 unique-id-format %{+X}o\ %Ci:%Cp_%Fi:%Fp_%Ts_%rt:%pid
7472
7473 will generate:
7474
7475 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
7476
7477 See also: "unique-id-header"
7478
7479unique-id-header <name>
7480 Add a unique ID header in the HTTP request.
7481 May be used in sections : defaults | frontend | listen | backend
7482 yes | yes | yes | no
7483 Arguments :
7484 <name> is the name of the header.
7485
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007486 Add a unique-id header in the HTTP request sent to the server, using the
7487 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +01007488
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007489 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01007490
7491 unique-id-format %{+X}o\ %Ci:%Cp_%Fi:%Fp_%Ts_%rt:%pid
7492 unique-id-header X-Unique-ID
7493
7494 will generate:
7495
7496 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
7497
7498 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +01007499
7500use_backend <backend> if <condition>
7501use_backend <backend> unless <condition>
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02007502 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007503 May be used in sections : defaults | frontend | listen | backend
7504 no | yes | yes | no
7505 Arguments :
7506 <backend> is the name of a valid backend or "listen" section.
7507
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007508 <condition> is a condition composed of ACLs, as described in section 7.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007509
7510 When doing content-switching, connections arrive on a frontend and are then
7511 dispatched to various backends depending on a number of conditions. The
7512 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02007513 "use_backend" keyword. While it is normally used with HTTP processing, it can
7514 also be used in pure TCP, either without content using stateless ACLs (eg:
7515 source address validation) or combined with a "tcp-request" rule to wait for
7516 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007517
7518 There may be as many "use_backend" rules as desired. All of these rules are
7519 evaluated in their declaration order, and the first one which matches will
7520 assign the backend.
7521
7522 In the first form, the backend will be used if the condition is met. In the
7523 second form, the backend will be used if the condition is not met. If no
7524 condition is valid, the backend defined with "default_backend" will be used.
7525 If no default backend is defined, either the servers in the same section are
7526 used (in case of a "listen" section) or, in case of a frontend, no server is
7527 used and a 503 service unavailable response is returned.
7528
Willy Tarreau51aecc72009-07-12 09:47:04 +02007529 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007530 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +02007531 and backend processing will immediately follow, or the backend will wait for
7532 a complete HTTP request to get in. This feature is useful when a frontend
7533 must decode several protocols on a unique port, one of them being HTTP.
7534
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02007535 See also: "default_backend", "tcp-request", and section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +01007536
Willy Tarreau036fae02008-01-06 13:24:40 +01007537
Willy Tarreau4a5cade2012-04-05 21:09:48 +02007538use-server <server> if <condition>
7539use-server <server> unless <condition>
7540 Only use a specific server if/unless an ACL-based condition is matched.
7541 May be used in sections : defaults | frontend | listen | backend
7542 no | no | yes | yes
7543 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007544 <server> is the name of a valid server in the same backend section.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02007545
7546 <condition> is a condition composed of ACLs, as described in section 7.
7547
7548 By default, connections which arrive to a backend are load-balanced across
7549 the available servers according to the configured algorithm, unless a
7550 persistence mechanism such as a cookie is used and found in the request.
7551
7552 Sometimes it is desirable to forward a particular request to a specific
7553 server without having to declare a dedicated backend for this server. This
7554 can be achieved using the "use-server" rules. These rules are evaluated after
7555 the "redirect" rules and before evaluating cookies, and they have precedence
7556 on them. There may be as many "use-server" rules as desired. All of these
7557 rules are evaluated in their declaration order, and the first one which
7558 matches will assign the server.
7559
7560 If a rule designates a server which is down, and "option persist" is not used
7561 and no force-persist rule was validated, it is ignored and evaluation goes on
7562 with the next rules until one matches.
7563
7564 In the first form, the server will be used if the condition is met. In the
7565 second form, the server will be used if the condition is not met. If no
7566 condition is valid, the processing continues and the server will be assigned
7567 according to other persistence mechanisms.
7568
7569 Note that even if a rule is matched, cookie processing is still performed but
7570 does not assign the server. This allows prefixed cookies to have their prefix
7571 stripped.
7572
7573 The "use-server" statement works both in HTTP and TCP mode. This makes it
7574 suitable for use with content-based inspection. For instance, a server could
7575 be selected in a farm according to the TLS SNI field. And if these servers
7576 have their weight set to zero, they will not be used for other traffic.
7577
7578 Example :
7579 # intercept incoming TLS requests based on the SNI field
7580 use-server www if { req_ssl_sni -i www.example.com }
7581 server www 192.168.0.1:443 weight 0
7582 use-server mail if { req_ssl_sni -i mail.example.com }
7583 server mail 192.168.0.1:587 weight 0
7584 use-server imap if { req_ssl_sni -i imap.example.com }
7585 server mail 192.168.0.1:993 weight 0
7586 # all the rest is forwarded to this server
7587 server default 192.168.0.2:443 check
7588
7589 See also: "use_backend", serction 5 about server and section 7 about ACLs.
7590
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007591
75925. Bind and Server options
7593--------------------------
7594
7595The "bind", "server" and "default-server" keywords support a number of settings
7596depending on some build options and on the system HAProxy was built on. These
7597settings generally each consist in one word sometimes followed by a value,
7598written on the same line as the "bind" or "server" line. All these options are
7599described in this section.
7600
7601
76025.1. Bind options
7603-----------------
7604
7605The "bind" keyword supports a certain number of settings which are all passed
7606as arguments on the same line. The order in which those arguments appear makes
7607no importance, provided that they appear after the bind address. All of these
7608parameters are optional. Some of them consist in a single words (booleans),
7609while other ones expect a value after them. In this case, the value must be
7610provided immediately after the setting name.
7611
7612The currently supported settings are the following ones.
7613
7614accept-proxy
7615 Enforces the use of the PROXY protocol over any connection accepted by any of
7616 the sockets declared on the same line. The PROXY protocol dictates the layer
7617 3/4 addresses of the incoming connection to be used everywhere an address is
7618 used, with the only exception of "tcp-request connection" rules which will
7619 only see the real connection address. Logs will reflect the addresses
7620 indicated in the protocol, unless it is violated, in which case the real
7621 address will still be used. This keyword combined with support from external
7622 components can be used as an efficient and reliable alternative to the
7623 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +02007624 usable. See also "tcp-request connection expect-proxy" for a finer-grained
7625 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007626
Willy Tarreauab861d32013-04-02 02:30:41 +02007627alpn <protocols>
7628 This enables the TLS ALPN extension and advertises the specified protocol
7629 list as supported on top of ALPN. The protocol list consists in a comma-
7630 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
7631 quotes). This requires that the SSL library is build with support for TLS
7632 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
7633 initial NPN extension.
7634
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007635backlog <backlog>
7636 Sets the socket's backlog to this value. If unspecified, the frontend's
7637 backlog is used instead, which generally defaults to the maxconn value.
7638
Emeric Brun7fb34422012-09-28 15:26:15 +02007639ecdhe <named curve>
7640 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +01007641 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
7642 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +02007643
Emeric Brunfd33a262012-10-11 16:28:27 +02007644ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +02007645 This setting is only available when support for OpenSSL was built in. It
7646 designates a PEM file from which to load CA certificates used to verify
7647 client's certificate.
7648
Emeric Brunb6dc9342012-09-28 17:55:37 +02007649ca-ignore-err [all|<errorID>,...]
7650 This setting is only available when support for OpenSSL was built in.
7651 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
7652 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
7653 error is ignored.
7654
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007655ciphers <ciphers>
7656 This setting is only available when support for OpenSSL was built in. It sets
7657 the string describing the list of cipher algorithms ("cipher suite") that are
7658 negociated during the SSL/TLS handshake. The format of the string is defined
7659 in "man 1 ciphers" from OpenSSL man pages, and can be for instance a string
7660 such as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes).
7661
Emeric Brunfd33a262012-10-11 16:28:27 +02007662crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +02007663 This setting is only available when support for OpenSSL was built in. It
7664 designates a PEM file from which to load certificate revocation list used
7665 to verify client's certificate.
7666
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007667crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +00007668 This setting is only available when support for OpenSSL was built in. It
7669 designates a PEM file containing both the required certificates and any
7670 associated private keys. This file can be built by concatenating multiple
7671 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
7672 requires an intermediate certificate, this can also be concatenated into this
7673 file.
7674
7675 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
7676 are loaded.
7677
7678 If a directory name is used instead of a PEM file, then all files found in
7679 that directory will be loaded. This directive may be specified multiple times
7680 in order to load certificates from multiple files or directories. The
7681 certificates will be presented to clients who provide a valid TLS Server Name
7682 Indication field matching one of their CN or alt subjects. Wildcards are
7683 supported, where a wildcard character '*' is used instead of the first
7684 hostname component (eg: *.example.org matches www.example.org but not
7685 www.sub.example.org).
7686
7687 If no SNI is provided by the client or if the SSL library does not support
7688 TLS extensions, or if the client provides an SNI hostname which does not
7689 match any certificate, then the first loaded certificate will be presented.
7690 This means that when loading certificates from a directory, it is highly
7691 recommended to load the default one first as a file.
7692
Emeric Brune032bfa2012-09-28 13:01:45 +02007693 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007694
Alex Davies0fbf0162013-03-02 16:04:50 +00007695 Some CAs (such as Godaddy) offer a drop down list of server types that do not
7696 include HAProxy when obtaining a certificate. If this happens be sure to
7697 choose a webserver that the CA believes requires a intermediate CA (for
7698 Godaddy, selection Apache Tomcat will get the correct bundle, but many
7699 others, e.g. nginx, result in a wrong bundle that will not work for some
7700 clients).
7701
Emeric Brunb6dc9342012-09-28 17:55:37 +02007702crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +00007703 This setting is only available when support for OpenSSL was built in. Sets a
7704 comma separated list of errorIDs to ignore during verify at depth == 0. If
7705 set to 'all', all errors are ignored. SSL handshake is not abored if an error
7706 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +02007707
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01007708crt-list <file>
7709 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02007710 designates a list of PEM file with an optional list of SNI filter per
7711 certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01007712
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02007713 <crtfile> [[!]<snifilter> ...]
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01007714
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02007715 Wildcards are supported in the SNI filter. Negative filter are also supported,
7716 only useful in combination with a wildcard filter to exclude a particular SNI.
7717 The certificates will be presented to clients who provide a valid TLS Server
7718 Name Indication field matching one of the SNI filters. If no SNI filter is
7719 specified, the CN and alt subjects are used. This directive may be specified
7720 multiple times. See the "crt" option for more information. The default
7721 certificate is still needed to meet OpenSSL expectations. If it is not used,
7722 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01007723
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007724defer-accept
7725 Is an optional keyword which is supported only on certain Linux kernels. It
7726 states that a connection will only be accepted once some data arrive on it,
7727 or at worst after the first retransmit. This should be used only on protocols
7728 for which the client talks first (eg: HTTP). It can slightly improve
7729 performance by ensuring that most of the request is already available when
7730 the connection is accepted. On the other hand, it will not be able to detect
7731 connections which don't talk. It is important to note that this option is
7732 broken in all kernels up to 2.6.31, as the connection is never accepted until
7733 the client talks. This can cause issues with front firewalls which would see
7734 an established connection while the proxy will only see it in SYN_RECV. This
7735 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
7736
Emeric Brun2cb7ae52012-10-05 14:14:21 +02007737force-sslv3
7738 This option enforces use of SSLv3 only on SSL connections instanciated from
7739 this listener. SSLv3 is generally less expensive than the TLS counterparts
7740 for high connection rates. See also "force-tls*", "no-sslv3", and "no-tls*".
7741
7742force-tlsv10
7743 This option enforces use of TLSv1.0 only on SSL connections instanciated from
7744 this listener. See also "force-tls*", "no-sslv3", and "no-tls*".
7745
7746force-tlsv11
7747 This option enforces use of TLSv1.1 only on SSL connections instanciated from
7748 this listener. See also "force-tls*", "no-sslv3", and "no-tls*".
7749
7750force-tlsv12
7751 This option enforces use of TLSv1.2 only on SSL connections instanciated from
7752 this listener. See also "force-tls*", "no-sslv3", and "no-tls*".
7753
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007754gid <gid>
7755 Sets the group of the UNIX sockets to the designated system gid. It can also
7756 be set by default in the global section's "unix-bind" statement. Note that
7757 some platforms simply ignore this. This setting is equivalent to the "group"
7758 setting except that the group ID is used instead of its name. This setting is
7759 ignored by non UNIX sockets.
7760
7761group <group>
7762 Sets the group of the UNIX sockets to the designated system group. It can
7763 also be set by default in the global section's "unix-bind" statement. Note
7764 that some platforms simply ignore this. This setting is equivalent to the
7765 "gid" setting except that the group name is used instead of its gid. This
7766 setting is ignored by non UNIX sockets.
7767
7768id <id>
7769 Fixes the socket ID. By default, socket IDs are automatically assigned, but
7770 sometimes it is more convenient to fix them to ease monitoring. This value
7771 must be strictly positive and unique within the listener/frontend. This
7772 option can only be used when defining only a single socket.
7773
7774interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +01007775 Restricts the socket to a specific interface. When specified, only packets
7776 received from that particular interface are processed by the socket. This is
7777 currently only supported on Linux. The interface must be a primary system
7778 interface, not an aliased interface. It is also possible to bind multiple
7779 frontends to the same address if they are bound to different interfaces. Note
7780 that binding to a network interface requires root privileges. This parameter
7781 is only compatible with TCPv4/TCPv6 sockets.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007782
Willy Tarreauabb175f2012-09-24 12:43:26 +02007783level <level>
7784 This setting is used with the stats sockets only to restrict the nature of
7785 the commands that can be issued on the socket. It is ignored by other
7786 sockets. <level> can be one of :
7787 - "user" is the least privileged level ; only non-sensitive stats can be
7788 read, and no change is allowed. It would make sense on systems where it
7789 is not easy to restrict access to the socket.
7790 - "operator" is the default level and fits most common uses. All data can
7791 be read, and only non-sensitive changes are permitted (eg: clear max
7792 counters).
7793 - "admin" should be used with care, as everything is permitted (eg: clear
7794 all counters).
7795
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007796maxconn <maxconn>
7797 Limits the sockets to this number of concurrent connections. Extraneous
7798 connections will remain in the system's backlog until a connection is
7799 released. If unspecified, the limit will be the same as the frontend's
7800 maxconn. Note that in case of port ranges or multiple addresses, the same
7801 value will be applied to each socket. This setting enables different
7802 limitations on expensive sockets, for instance SSL entries which may easily
7803 eat all memory.
7804
7805mode <mode>
7806 Sets the octal mode used to define access permissions on the UNIX socket. It
7807 can also be set by default in the global section's "unix-bind" statement.
7808 Note that some platforms simply ignore this. This setting is ignored by non
7809 UNIX sockets.
7810
7811mss <maxseg>
7812 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
7813 connections. This can be used to force a lower MSS for certain specific
7814 ports, for instance for connections passing through a VPN. Note that this
7815 relies on a kernel feature which is theoretically supported under Linux but
7816 was buggy in all versions prior to 2.6.28. It may or may not work on other
7817 operating systems. It may also not change the advertised value but change the
7818 effective size of outgoing segments. The commonly advertised value for TCPv4
7819 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
7820 positive, it will be used as the advertised MSS. If it is negative, it will
7821 indicate by how much to reduce the incoming connection's advertised MSS for
7822 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
7823
7824name <name>
7825 Sets an optional name for these sockets, which will be reported on the stats
7826 page.
7827
7828nice <nice>
7829 Sets the 'niceness' of connections initiated from the socket. Value must be
7830 in the range -1024..1024 inclusive, and defaults to zero. Positive values
7831 means that such connections are more friendly to others and easily offer
7832 their place in the scheduler. On the opposite, negative values mean that
7833 connections want to run with a higher priority than others. The difference
7834 only happens under high loads when the system is close to saturation.
7835 Negative values are appropriate for low-latency or administration services,
7836 and high values are generally recommended for CPU intensive tasks such as SSL
7837 processing or bulk transfers which are less sensible to latency. For example,
7838 it may make sense to use a positive value for an SMTP socket and a negative
7839 one for an RDP socket.
7840
Emeric Brun9b3009b2012-10-05 11:55:06 +02007841no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007842 This setting is only available when support for OpenSSL was built in. It
7843 disables support for SSLv3 on any sockets instanciated from the listener when
7844 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2cb7ae52012-10-05 14:14:21 +02007845 be enabled using any configuration option. See also "force-tls*",
7846 and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007847
Emeric Brun90ad8722012-10-02 14:00:59 +02007848no-tls-tickets
7849 This setting is only available when support for OpenSSL was built in. It
7850 disables the stateless session resumption (RFC 5077 TLS Ticket
7851 extension) and force to use stateful session resumption. Stateless
7852 session resumption is more expensive in CPU usage.
7853
Emeric Brun9b3009b2012-10-05 11:55:06 +02007854no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007855 This setting is only available when support for OpenSSL was built in. It
Emeric Brun2cb7ae52012-10-05 14:14:21 +02007856 disables support for TLSv1.0 on any sockets instanciated from the listener
7857 when SSL is supported. Note that SSLv2 is forced disabled in the code and
7858 cannot be enabled using any configuration option. See also "force-tls*",
7859 and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007860
Emeric Brun9b3009b2012-10-05 11:55:06 +02007861no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +02007862 This setting is only available when support for OpenSSL was built in. It
Emeric Brun2cb7ae52012-10-05 14:14:21 +02007863 disables support for TLSv1.1 on any sockets instanciated from the listener
7864 when SSL is supported. Note that SSLv2 is forced disabled in the code and
7865 cannot be enabled using any configuration option. See also "force-tls*",
7866 and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +02007867
Emeric Brun9b3009b2012-10-05 11:55:06 +02007868no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +02007869 This setting is only available when support for OpenSSL was built in. It
Emeric Brun2cb7ae52012-10-05 14:14:21 +02007870 disables support for TLSv1.2 on any sockets instanciated from the listener
7871 when SSL is supported. Note that SSLv2 is forced disabled in the code and
7872 cannot be enabled using any configuration option. See also "force-tls*",
7873 and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +02007874
Willy Tarreau6c9a3d52012-10-18 18:57:14 +02007875npn <protocols>
7876 This enables the NPN TLS extension and advertises the specified protocol list
7877 as supported on top of NPN. The protocol list consists in a comma-delimited
7878 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
7879 This requires that the SSL library is build with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +02007880 enabled (check with haproxy -vv). Note that the NPN extension has been
7881 replaced with the ALPN extension (see the "alpn" keyword).
Willy Tarreau6c9a3d52012-10-18 18:57:14 +02007882
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007883ssl
7884 This setting is only available when support for OpenSSL was built in. It
7885 enables SSL deciphering on connections instanciated from this listener. A
7886 certificate is necessary (see "crt" above). All contents in the buffers will
7887 appear in clear text, so that ACLs and HTTP processing will only have access
7888 to deciphered contents.
7889
Emmanuel Hocdet65623372013-01-24 17:17:15 +01007890strict-sni
7891 This setting is only available when support for OpenSSL was built in. The
7892 SSL/TLS negotiation is allow only if the client provided an SNI which match
7893 a certificate. The default certificate is not used.
7894 See the "crt" option for more information.
7895
Willy Tarreau1c862c52012-10-05 16:21:00 +02007896tfo
Lukas Tribus0defb902013-02-13 23:35:39 +01007897 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +02007898 enables TCP Fast Open on the listening socket, which means that clients which
7899 support this feature will be able to send a request and receive a response
7900 during the 3-way handshake starting from second connection, thus saving one
7901 round-trip after the first connection. This only makes sense with protocols
7902 that use high connection rates and where each round trip matters. This can
7903 possibly cause issues with many firewalls which do not accept data on SYN
7904 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +02007905 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
7906 need to build HAProxy with USE_TFO=1 if your libc doesn't define
7907 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +02007908
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007909transparent
7910 Is an optional keyword which is supported only on certain Linux kernels. It
7911 indicates that the addresses will be bound even if they do not belong to the
7912 local machine, and that packets targeting any of these addresses will be
7913 intercepted just as if the addresses were locally configured. This normally
7914 requires that IP forwarding is enabled. Caution! do not use this with the
7915 default address '*', as it would redirect any traffic for the specified port.
7916 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
7917 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
7918 kernel version. Some distribution kernels include backports of the feature,
7919 so check for support with your vendor.
7920
Willy Tarreau77e3af92012-11-24 15:07:23 +01007921v4v6
7922 Is an optional keyword which is supported only on most recent systems
7923 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
7924 and IPv6 when it uses the default address. Doing so is sometimes necessary
7925 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
7926 sockets, and is overriden by the "v6only" option.
7927
Willy Tarreau9b6700f2012-11-24 11:55:28 +01007928v6only
7929 Is an optional keyword which is supported only on most recent systems
7930 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
7931 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +01007932 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
7933 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +01007934
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007935uid <uid>
7936 Sets the owner of the UNIX sockets to the designated system uid. It can also
7937 be set by default in the global section's "unix-bind" statement. Note that
7938 some platforms simply ignore this. This setting is equivalent to the "user"
7939 setting except that the user numeric ID is used instead of its name. This
7940 setting is ignored by non UNIX sockets.
7941
7942user <user>
7943 Sets the owner of the UNIX sockets to the designated system user. It can also
7944 be set by default in the global section's "unix-bind" statement. Note that
7945 some platforms simply ignore this. This setting is equivalent to the "uid"
7946 setting except that the user name is used instead of its uid. This setting is
7947 ignored by non UNIX sockets.
7948
Emeric Brun1a073b42012-09-28 17:07:34 +02007949verify [none|optional|required]
7950 This setting is only available when support for OpenSSL was built in. If set
7951 to 'none', client certificate is not requested. This is the default. In other
7952 cases, a client certificate is requested. If the client does not provide a
7953 certificate after the request and if 'verify' is set to 'required', then the
7954 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +02007955 certificate provided by the client is always verified using CAs from
7956 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
7957 is aborted, regardless of the 'verify' option, unless the error code exactly
7958 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02007959
Willy Tarreaub6205fd2012-09-24 12:27:33 +020079605.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +01007961------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02007962
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01007963The "server" and "default-server" keywords support a certain number of settings
7964which are all passed as arguments on the server line. The order in which those
7965arguments appear does not count, and they are all optional. Some of those
7966settings are single words (booleans) while others expect one or several values
7967after them. In this case, the values must immediately follow the setting name.
7968Except default-server, all those settings must be specified after the server's
7969address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +02007970
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007971 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01007972 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +02007973
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007974The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +01007975
Willy Tarreauceb4ac92012-04-28 00:41:46 +02007976addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007977 Using the "addr" parameter, it becomes possible to use a different IP address
7978 to send health-checks. On some servers, it may be desirable to dedicate an IP
7979 address to specific component able to perform complex tests which are more
7980 suitable to health-checks than the application. This parameter is ignored if
7981 the "check" parameter is not set. See also the "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +02007982
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01007983 Supported in default-server: No
7984
Simon Hormand60d6912013-11-25 10:46:36 +09007985agent-check
7986 Enable an auxiliary agent check which is run independently of a regular
7987 health check. An agent health check is performed by making a TCP
7988 connection to the port set by the "agent-port" parameter" and reading
7989 an ASCII string. The string should have one of the following forms:
7990
7991 * An ASCII representation of an positive integer percentage.
7992 e.g. "75%"
7993
7994 Values in this format will set the weight proportional to the initial
7995 weight of a server as configured when haproxy starts.
7996
7997 * The string "drain".
7998
7999 This will cause the weight of a server to be set to 0, and thus it will
8000 not accept any new connections other than those that are accepted via
8001 persistence.
8002
8003 * The string "down", optionally followed by a description string.
8004
8005 Mark the server as down and log the description string as the reason.
8006
8007 * The string "stopped", optionally followed by a description string.
8008
8009 This currently has the same behaviour as "down".
8010
8011 * The string "fail", optionally followed by a description string.
8012
8013 This currently has the same behaviour as "down".
8014
Simon Horman2f1f9552013-11-25 10:46:37 +09008015 Failure to connect to the agent is not considered an error as connectivity
8016 is tested by the regular health check which is enabled by the "check"
8017 parameter.
8018
Simon Hormand60d6912013-11-25 10:46:36 +09008019 Requires the ""agent-port" parameter to be set.
8020 See also the "agent-check" parameter.
8021
8022 Supported in default-server: No
8023
8024agent-inter <delay>
8025 The "agent-inter" parameter sets the interval between two agent checks
8026 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
8027
8028 Just as with every other time-based parameter, it may be entered in any
8029 other explicit unit among { us, ms, s, m, h, d }. The "agent-inter"
8030 parameter also serves as a timeout for agent checks "timeout check" is
8031 not set. In order to reduce "resonance" effects when multiple servers are
8032 hosted on the same hardware, the agent and health checks of all servers
8033 are started with a small time offset between them. It is also possible to
8034 add some random noise in the agent and health checks interval using the
8035 global "spread-checks" keyword. This makes sense for instance when a lot
8036 of backends use the same servers.
8037
8038 See also the "agent-check" and "agent-port" parameters.
8039
8040 Supported in default-server: Yes
8041
8042agent-port <port>
8043 The "agent-port" parameter sets the TCP port used for agent checks.
8044
8045 See also the "agent-check" and "agent-inter" parameters.
8046
8047 Supported in default-server: Yes
8048
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008049backup
8050 When "backup" is present on a server line, the server is only used in load
8051 balancing when all other non-backup servers are unavailable. Requests coming
8052 with a persistence cookie referencing the server will always be served
8053 though. By default, only the first operational backup server is used, unless
8054 the "allbackups" option is set in the backend. See also the "allbackups"
8055 option.
Willy Tarreau6a06a402007-07-15 20:15:28 +02008056
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008057 Supported in default-server: No
8058
Emeric Brunef42d922012-10-11 16:11:36 +02008059ca-file <cafile>
8060 This setting is only available when support for OpenSSL was built in. It
8061 designates a PEM file from which to load CA certificates used to verify
8062 server's certificate.
8063
8064 Supported in default-server: No
8065
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008066check
8067 This option enables health checks on the server. By default, a server is
Patrick Mézardb7aeec62012-01-22 16:01:22 +01008068 always considered available. If "check" is set, the server is available when
8069 accepting periodic TCP connections, to ensure that it is really able to serve
8070 requests. The default address and port to send the tests to are those of the
8071 server, and the default source is the same as the one defined in the
8072 backend. It is possible to change the address using the "addr" parameter, the
8073 port using the "port" parameter, the source address using the "source"
8074 address, and the interval and timers using the "inter", "rise" and "fall"
Simon Hormanafc47ee2013-11-25 10:46:35 +09008075 parameters. The request method is define in the backend using the "httpchk",
8076 "smtpchk", "mysql-check", "pgsql-check" and "ssl-hello-chk" options. Please
8077 refer to those options and parameters for more information.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008078
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008079 Supported in default-server: No
8080
Willy Tarreau6c16adc2012-10-05 00:04:16 +02008081check-send-proxy
8082 This option forces emission of a PROXY protocol line with outgoing health
8083 checks, regardless of whether the server uses send-proxy or not for the
8084 normal traffic. By default, the PROXY protocol is enabled for health checks
8085 if it is already enabled for normal traffic and if no "port" nor "addr"
8086 directive is present. However, if such a directive is present, the
8087 "check-send-proxy" option needs to be used to force the use of the
8088 protocol. See also the "send-proxy" option for more information.
8089
8090 Supported in default-server: No
8091
Willy Tarreau763a95b2012-10-04 23:15:39 +02008092check-ssl
8093 This option forces encryption of all health checks over SSL, regardless of
8094 whether the server uses SSL or not for the normal traffic. This is generally
8095 used when an explicit "port" or "addr" directive is specified and SSL health
8096 checks are not inherited. It is important to understand that this option
8097 inserts an SSL transport layer below the ckecks, so that a simple TCP connect
8098 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
8099 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
8100 All SSL settings are common to health checks and traffic (eg: ciphers).
8101 See the "ssl" option for more information.
8102
8103 Supported in default-server: No
8104
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02008105ciphers <ciphers>
8106 This option sets the string describing the list of cipher algorithms that is
8107 is negociated during the SSL/TLS handshake with the server. The format of the
8108 string is defined in "man 1 ciphers". When SSL is used to communicate with
8109 servers on the local network, it is common to see a weaker set of algorithms
8110 than what is used over the internet. Doing so reduces CPU usage on both the
8111 server and haproxy while still keeping it compatible with deployed software.
8112 Some algorithms such as RC4-SHA1 are reasonably cheap. If no security at all
8113 is needed and just connectivity, using DES can be appropriate.
8114
Willy Tarreau763a95b2012-10-04 23:15:39 +02008115 Supported in default-server: No
8116
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008117cookie <value>
8118 The "cookie" parameter sets the cookie value assigned to the server to
8119 <value>. This value will be checked in incoming requests, and the first
8120 operational server possessing the same value will be selected. In return, in
8121 cookie insertion or rewrite modes, this value will be assigned to the cookie
8122 sent to the client. There is nothing wrong in having several servers sharing
8123 the same cookie value, and it is in fact somewhat common between normal and
8124 backup servers. See also the "cookie" keyword in backend section.
8125
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008126 Supported in default-server: No
8127
Emeric Brunef42d922012-10-11 16:11:36 +02008128crl-file <crlfile>
8129 This setting is only available when support for OpenSSL was built in. It
8130 designates a PEM file from which to load certificate revocation list used
8131 to verify server's certificate.
8132
8133 Supported in default-server: No
8134
Emeric Bruna7aa3092012-10-26 12:58:00 +02008135crt <cert>
8136 This setting is only available when support for OpenSSL was built in.
8137 It designates a PEM file from which to load both a certificate and the
8138 associated private key. This file can be built by concatenating both PEM
8139 files into one. This certificate will be sent if the server send a client
8140 certificate request.
8141
8142 Supported in default-server: No
8143
Willy Tarreau96839092010-03-29 10:02:24 +02008144disabled
8145 The "disabled" keyword starts the server in the "disabled" state. That means
8146 that it is marked down in maintenance mode, and no connection other than the
8147 ones allowed by persist mode will reach it. It is very well suited to setup
8148 new servers, because normal traffic will never reach them, while it is still
8149 possible to test the service by making use of the force-persist mechanism.
8150
8151 Supported in default-server: No
8152
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008153error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +01008154 If health observing is enabled, the "error-limit" parameter specifies the
8155 number of consecutive errors that triggers event selected by the "on-error"
8156 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01008157
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008158 Supported in default-server: Yes
8159
8160 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01008161
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008162fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008163 The "fall" parameter states that a server will be considered as dead after
8164 <count> consecutive unsuccessful health checks. This value defaults to 3 if
8165 unspecified. See also the "check", "inter" and "rise" parameters.
8166
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008167 Supported in default-server: Yes
8168
Emeric Brun8694b9a2012-10-05 14:39:07 +02008169force-sslv3
8170 This option enforces use of SSLv3 only when SSL is used to communicate with
8171 the server. SSLv3 is generally less expensive than the TLS counterparts for
8172 high connection rates. See also "no-tlsv*", "no-sslv3".
8173
8174 Supported in default-server: No
8175
8176force-tlsv10
8177 This option enforces use of TLSv1.0 only when SSL is used to communicate with
8178 the server. See also "no-tlsv*", "no-sslv3".
8179
8180 Supported in default-server: No
8181
8182force-tlsv11
8183 This option enforces use of TLSv1.1 only when SSL is used to communicate with
8184 the server. See also "no-tlsv*", "no-sslv3".
8185
8186 Supported in default-server: No
8187
8188force-tlsv12
8189 This option enforces use of TLSv1.2 only when SSL is used to communicate with
8190 the server. See also "no-tlsv*", "no-sslv3".
8191
8192 Supported in default-server: No
8193
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008194id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02008195 Set a persistent ID for the server. This ID must be positive and unique for
8196 the proxy. An unused ID will automatically be assigned if unset. The first
8197 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008198
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008199 Supported in default-server: No
8200
8201inter <delay>
8202fastinter <delay>
8203downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008204 The "inter" parameter sets the interval between two consecutive health checks
8205 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
8206 It is also possible to use "fastinter" and "downinter" to optimize delays
8207 between checks depending on the server state :
8208
8209 Server state | Interval used
8210 ---------------------------------+-----------------------------------------
8211 UP 100% (non-transitional) | "inter"
8212 ---------------------------------+-----------------------------------------
8213 Transitionally UP (going down), |
8214 Transitionally DOWN (going up), | "fastinter" if set, "inter" otherwise.
8215 or yet unchecked. |
8216 ---------------------------------+-----------------------------------------
8217 DOWN 100% (non-transitional) | "downinter" if set, "inter" otherwise.
8218 ---------------------------------+-----------------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +01008219
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008220 Just as with every other time-based parameter, they can be entered in any
8221 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
8222 serves as a timeout for health checks sent to servers if "timeout check" is
8223 not set. In order to reduce "resonance" effects when multiple servers are
Simon Hormand60d6912013-11-25 10:46:36 +09008224 hosted on the same hardware, the agent and health checks of all servers
8225 are started with a small time offset between them. It is also possible to
8226 add some random noise in the agent and health checks interval using the
8227 global "spread-checks" keyword. This makes sense for instance when a lot
8228 of backends use the same servers.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008229
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008230 Supported in default-server: Yes
8231
8232maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008233 The "maxconn" parameter specifies the maximal number of concurrent
8234 connections that will be sent to this server. If the number of incoming
8235 concurrent requests goes higher than this value, they will be queued, waiting
8236 for a connection to be released. This parameter is very important as it can
8237 save fragile servers from going down under extreme loads. If a "minconn"
8238 parameter is specified, the limit becomes dynamic. The default value is "0"
8239 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
8240 the backend's "fullconn" keyword.
8241
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008242 Supported in default-server: Yes
8243
8244maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008245 The "maxqueue" parameter specifies the maximal number of connections which
8246 will wait in the queue for this server. If this limit is reached, next
8247 requests will be redispatched to other servers instead of indefinitely
8248 waiting to be served. This will break persistence but may allow people to
8249 quickly re-log in when the server they try to connect to is dying. The
8250 default value is "0" which means the queue is unlimited. See also the
8251 "maxconn" and "minconn" parameters.
8252
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008253 Supported in default-server: Yes
8254
8255minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008256 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
8257 limit following the backend's load. The server will always accept at least
8258 <minconn> connections, never more than <maxconn>, and the limit will be on
8259 the ramp between both values when the backend has less than <fullconn>
8260 concurrent connections. This makes it possible to limit the load on the
8261 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008262 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008263 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01008264
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008265 Supported in default-server: Yes
8266
Emeric Brun9b3009b2012-10-05 11:55:06 +02008267no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02008268 This option disables support for SSLv3 when SSL is used to communicate with
8269 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emeric Brun8694b9a2012-10-05 14:39:07 +02008270 using any configuration option. See also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02008271
Willy Tarreau763a95b2012-10-04 23:15:39 +02008272 Supported in default-server: No
8273
Emeric Brunf9c5c472012-10-11 15:28:34 +02008274no-tls-tickets
8275 This setting is only available when support for OpenSSL was built in. It
8276 disables the stateless session resumption (RFC 5077 TLS Ticket
8277 extension) and force to use stateful session resumption. Stateless
8278 session resumption is more expensive in CPU usage for servers.
8279
8280 Supported in default-server: No
8281
Emeric Brun9b3009b2012-10-05 11:55:06 +02008282no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +02008283 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +02008284 the server. Note that SSLv2 is disabled in the code and cannot be enabled
8285 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun8694b9a2012-10-05 14:39:07 +02008286 often makes sense to disable it when communicating with local servers. See
8287 also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +02008288
Willy Tarreau763a95b2012-10-04 23:15:39 +02008289 Supported in default-server: No
8290
Emeric Brun9b3009b2012-10-05 11:55:06 +02008291no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +02008292 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +02008293 the server. Note that SSLv2 is disabled in the code and cannot be enabled
8294 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun8694b9a2012-10-05 14:39:07 +02008295 often makes sense to disable it when communicating with local servers. See
8296 also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +02008297
Willy Tarreau763a95b2012-10-04 23:15:39 +02008298 Supported in default-server: No
8299
Emeric Brun9b3009b2012-10-05 11:55:06 +02008300no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +02008301 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02008302 the server. Note that SSLv2 is disabled in the code and cannot be enabled
8303 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun8694b9a2012-10-05 14:39:07 +02008304 often makes sense to disable it when communicating with local servers. See
8305 also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02008306
Willy Tarreau763a95b2012-10-04 23:15:39 +02008307 Supported in default-server: No
8308
Simon Hormanfa461682011-06-25 09:39:49 +09008309non-stick
8310 Never add connections allocated to this sever to a stick-table.
8311 This may be used in conjunction with backup to ensure that
8312 stick-table persistence is disabled for backup servers.
8313
Willy Tarreau763a95b2012-10-04 23:15:39 +02008314 Supported in default-server: No
8315
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01008316observe <mode>
8317 This option enables health adjusting based on observing communication with
8318 the server. By default this functionality is disabled and enabling it also
8319 requires to enable health checks. There are two supported modes: "layer4" and
8320 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
8321 significant. In layer7, which is only allowed for http proxies, responses
8322 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +01008323 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01008324
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008325 Supported in default-server: No
8326
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01008327 See also the "check", "on-error" and "error-limit".
8328
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008329on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01008330 Select what should happen when enough consecutive errors are detected.
8331 Currently, four modes are available:
8332 - fastinter: force fastinter
8333 - fail-check: simulate a failed check, also forces fastinter (default)
8334 - sudden-death: simulate a pre-fatal failed health check, one more failed
8335 check will mark a server down, forces fastinter
8336 - mark-down: mark the server immediately down and force fastinter
8337
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008338 Supported in default-server: Yes
8339
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01008340 See also the "check", "observe" and "error-limit".
8341
Simon Hormane0d1bfb2011-06-21 14:34:58 +09008342on-marked-down <action>
8343 Modify what occurs when a server is marked down.
8344 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -07008345 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
8346 all connections to the server are immediately terminated when the server
8347 goes down. It might be used if the health check detects more complex cases
8348 than a simple connection status, and long timeouts would cause the service
8349 to remain unresponsive for too long a time. For instance, a health check
8350 might detect that a database is stuck and that there's no chance to reuse
8351 existing connections anymore. Connections killed this way are logged with
8352 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +09008353
8354 Actions are disabled by default
8355
8356 Supported in default-server: Yes
8357
Justin Karnegeseb2c24a2012-05-24 15:28:52 -07008358on-marked-up <action>
8359 Modify what occurs when a server is marked up.
8360 Currently one action is available:
8361 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
8362 done only if the server is not in backup state and if it is not disabled
8363 (it must have an effective weight > 0). This can be used sometimes to force
8364 an active server to take all the traffic back after recovery when dealing
8365 with long sessions (eg: LDAP, SQL, ...). Doing this can cause more trouble
8366 than it tries to solve (eg: incomplete transactions), so use this feature
8367 with extreme care. Sessions killed because a server comes up are logged
8368 with an 'U' termination code (for "Up").
8369
8370 Actions are disabled by default
8371
8372 Supported in default-server: Yes
8373
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008374port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008375 Using the "port" parameter, it becomes possible to use a different port to
8376 send health-checks. On some servers, it may be desirable to dedicate a port
8377 to a specific component able to perform complex tests which are more suitable
8378 to health-checks than the application. It is common to run a simple script in
8379 inetd for instance. This parameter is ignored if the "check" parameter is not
8380 set. See also the "addr" parameter.
8381
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008382 Supported in default-server: Yes
8383
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008384redir <prefix>
8385 The "redir" parameter enables the redirection mode for all GET and HEAD
8386 requests addressing this server. This means that instead of having HAProxy
8387 forward the request to the server, it will send an "HTTP 302" response with
8388 the "Location" header composed of this prefix immediately followed by the
8389 requested URI beginning at the leading '/' of the path component. That means
8390 that no trailing slash should be used after <prefix>. All invalid requests
8391 will be rejected, and all non-GET or HEAD requests will be normally served by
8392 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008393 mangling nor cookie insertion is possible in the response. However, cookies in
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008394 requests are still analysed, making this solution completely usable to direct
8395 users to a remote location in case of local disaster. Main use consists in
8396 increasing bandwidth for static servers by having the clients directly
8397 connect to them. Note: never use a relative location here, it would cause a
8398 loop between the client and HAProxy!
8399
8400 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
8401
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008402 Supported in default-server: No
8403
8404rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008405 The "rise" parameter states that a server will be considered as operational
8406 after <count> consecutive successful health checks. This value defaults to 2
8407 if unspecified. See also the "check", "inter" and "fall" parameters.
8408
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008409 Supported in default-server: Yes
8410
Willy Tarreau5ab04ec2011-03-20 10:32:26 +01008411send-proxy
8412 The "send-proxy" parameter enforces use of the PROXY protocol over any
8413 connection established to this server. The PROXY protocol informs the other
8414 end about the layer 3/4 addresses of the incoming connection, so that it can
8415 know the client's address or the public address it accessed to, whatever the
8416 upper layer protocol. For connections accepted by an "accept-proxy" listener,
8417 the advertised address will be used. Only TCPv4 and TCPv6 address families
8418 are supported. Other families such as Unix sockets, will report an UNKNOWN
8419 family. Servers using this option can fully be chained to another instance of
8420 haproxy listening with an "accept-proxy" setting. This setting must not be
Willy Tarreau6c16adc2012-10-05 00:04:16 +02008421 used if the server isn't aware of the protocol. When health checks are sent
8422 to the server, the PROXY protocol is automatically used when this option is
8423 set, unless there is an explicit "port" or "addr" directive, in which case an
8424 explicit "check-send-proxy" directive would also be needed to use the PROXY
8425 protocol. See also the "accept-proxy" option of the "bind" keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +01008426
8427 Supported in default-server: No
8428
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008429slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008430 The "slowstart" parameter for a server accepts a value in milliseconds which
8431 indicates after how long a server which has just come back up will run at
8432 full speed. Just as with every other time-based parameter, it can be entered
8433 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
8434 linearly from 0 to 100% during this time. The limitation applies to two
8435 parameters :
8436
8437 - maxconn: the number of connections accepted by the server will grow from 1
8438 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
8439
8440 - weight: when the backend uses a dynamic weighted algorithm, the weight
8441 grows linearly from 1 to 100%. In this case, the weight is updated at every
8442 health-check. For this reason, it is important that the "inter" parameter
8443 is smaller than the "slowstart", in order to maximize the number of steps.
8444
8445 The slowstart never applies when haproxy starts, otherwise it would cause
8446 trouble to running servers. It only applies when a server has been previously
8447 seen as failed.
8448
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008449 Supported in default-server: Yes
8450
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02008451source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02008452source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02008453source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008454 The "source" parameter sets the source address which will be used when
8455 connecting to the server. It follows the exact same parameters and principle
8456 as the backend "source" keyword, except that it only applies to the server
8457 referencing it. Please consult the "source" keyword for details.
8458
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02008459 Additionally, the "source" statement on a server line allows one to specify a
8460 source port range by indicating the lower and higher bounds delimited by a
8461 dash ('-'). Some operating systems might require a valid IP address when a
8462 source port range is specified. It is permitted to have the same IP/range for
8463 several servers. Doing so makes it possible to bypass the maximum of 64k
8464 total concurrent connections. The limit will then reach 64k connections per
8465 server.
8466
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008467 Supported in default-server: No
8468
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02008469ssl
Willy Tarreau44f65392013-06-25 07:56:20 +02008470 This option enables SSL ciphering on outgoing connections to the server. It
8471 is critical to verify server certificates using "verify" when using SSL to
8472 connect to servers, otherwise the communication is prone to trivial man in
8473 the-middle attacks rendering SSL useless. When this option is used, health
8474 checks are automatically sent in SSL too unless there is a "port" or an
8475 "addr" directive indicating the check should be sent to a different location.
8476 See the "check-ssl" optino to force SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +02008477
8478 Supported in default-server: No
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02008479
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008480track [<proxy>/]<server>
8481 This option enables ability to set the current state of the server by
8482 tracking another one. Only a server with checks enabled can be tracked
8483 so it is not possible for example to track a server that tracks another
8484 one. If <proxy> is omitted the current one is used. If disable-on-404 is
8485 used, it has to be enabled on both proxies.
8486
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008487 Supported in default-server: No
8488
Emeric Brunef42d922012-10-11 16:11:36 +02008489verify [none|required]
8490 This setting is only available when support for OpenSSL was built in. If set
8491 to 'none', server certificate is not verified. This is the default. In the
8492 other case, The certificate provided by the server is verified using CAs from
8493 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
Willy Tarreau44f65392013-06-25 07:56:20 +02008494 is aborted. It is critically important to verify server certificates when
8495 using SSL to connect to servers, otherwise the communication is prone to
8496 trivial man-in-the-middle attacks rendering SSL totally useless.
Emeric Brunef42d922012-10-11 16:11:36 +02008497
8498 Supported in default-server: No
8499
Evan Broderbe554312013-06-27 00:05:25 -07008500verifyhost <hostname>
8501 This setting is only available when support for OpenSSL was built in, and
8502 only takes effect if 'verify required' is also specified. When set, the
8503 hostnames in the subject and subjectAlternateNames of the certificate
8504 provided by the server are checked. If none of the hostnames in the
8505 certificate match the specified hostname, the handshake is aborted. The
8506 hostnames in the server-provided certificate may include wildcards.
8507
8508 Supported in default-server: No
8509
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008510weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008511 The "weight" parameter is used to adjust the server's weight relative to
8512 other servers. All servers will receive a load proportional to their weight
8513 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +02008514 load. The default weight is 1, and the maximal value is 256. A value of 0
8515 means the server will not participate in load-balancing but will still accept
8516 persistent connections. If this parameter is used to distribute the load
8517 according to server's capacity, it is recommended to start with values which
8518 can both grow and shrink, for instance between 10 and 100 to leave enough
8519 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008520
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008521 Supported in default-server: Yes
8522
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008523
85246. HTTP header manipulation
8525---------------------------
8526
8527In HTTP mode, it is possible to rewrite, add or delete some of the request and
8528response headers based on regular expressions. It is also possible to block a
8529request or a response if a particular header matches a regular expression,
8530which is enough to stop most elementary protocol attacks, and to protect
8531against information leak from the internal network. But there is a limitation
8532to this : since HAProxy's HTTP engine does not support keep-alive, only headers
8533passed during the first request of a TCP session will be seen. All subsequent
8534headers will be considered data only and not analyzed. Furthermore, HAProxy
8535never touches data contents, it stops analysis at the end of headers.
8536
Willy Tarreau816b9792009-09-15 21:25:21 +02008537There is an exception though. If HAProxy encounters an "Informational Response"
8538(status code 1xx), it is able to process all rsp* rules which can allow, deny,
8539rewrite or delete a header, but it will refuse to add a header to any such
8540messages as this is not HTTP-compliant. The reason for still processing headers
8541in such responses is to stop and/or fix any possible information leak which may
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008542happen, for instance because another downstream equipment would unconditionally
Willy Tarreau816b9792009-09-15 21:25:21 +02008543add a header, or if a server name appears there. When such messages are seen,
8544normal processing still occurs on the next non-informational messages.
8545
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008546This section covers common usage of the following keywords, described in detail
8547in section 4.2 :
8548
8549 - reqadd <string>
8550 - reqallow <search>
8551 - reqiallow <search>
8552 - reqdel <search>
8553 - reqidel <search>
8554 - reqdeny <search>
8555 - reqideny <search>
8556 - reqpass <search>
8557 - reqipass <search>
8558 - reqrep <search> <replace>
8559 - reqirep <search> <replace>
8560 - reqtarpit <search>
8561 - reqitarpit <search>
8562 - rspadd <string>
8563 - rspdel <search>
8564 - rspidel <search>
8565 - rspdeny <search>
8566 - rspideny <search>
8567 - rsprep <search> <replace>
8568 - rspirep <search> <replace>
8569
8570With all these keywords, the same conventions are used. The <search> parameter
8571is a POSIX extended regular expression (regex) which supports grouping through
8572parenthesis (without the backslash). Spaces and other delimiters must be
8573prefixed with a backslash ('\') to avoid confusion with a field delimiter.
8574Other characters may be prefixed with a backslash to change their meaning :
8575
8576 \t for a tab
8577 \r for a carriage return (CR)
8578 \n for a new line (LF)
8579 \ to mark a space and differentiate it from a delimiter
8580 \# to mark a sharp and differentiate it from a comment
8581 \\ to use a backslash in a regex
8582 \\\\ to use a backslash in the text (*2 for regex, *2 for haproxy)
8583 \xXX to write the ASCII hex code XX as in the C language
8584
8585The <replace> parameter contains the string to be used to replace the largest
8586portion of text matching the regex. It can make use of the special characters
8587above, and can reference a substring which is delimited by parenthesis in the
8588regex, by writing a backslash ('\') immediately followed by one digit from 0 to
85899 indicating the group position (0 designating the entire line). This practice
8590is very common to users of the "sed" program.
8591
8592The <string> parameter represents the string which will systematically be added
8593after the last header line. It can also use special character sequences above.
8594
8595Notes related to these keywords :
8596---------------------------------
8597 - these keywords are not always convenient to allow/deny based on header
8598 contents. It is strongly recommended to use ACLs with the "block" keyword
8599 instead, resulting in far more flexible and manageable rules.
8600
8601 - lines are always considered as a whole. It is not possible to reference
8602 a header name only or a value only. This is important because of the way
8603 headers are written (notably the number of spaces after the colon).
8604
8605 - the first line is always considered as a header, which makes it possible to
8606 rewrite or filter HTTP requests URIs or response codes, but in turn makes
8607 it harder to distinguish between headers and request line. The regex prefix
8608 ^[^\ \t]*[\ \t] matches any HTTP method followed by a space, and the prefix
8609 ^[^ \t:]*: matches any header name followed by a colon.
8610
8611 - for performances reasons, the number of characters added to a request or to
8612 a response is limited at build time to values between 1 and 4 kB. This
8613 should normally be far more than enough for most usages. If it is too short
8614 on occasional usages, it is possible to gain some space by removing some
8615 useless headers before adding new ones.
8616
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008617 - keywords beginning with "reqi" and "rspi" are the same as their counterpart
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008618 without the 'i' letter except that they ignore case when matching patterns.
8619
8620 - when a request passes through a frontend then a backend, all req* rules
8621 from the frontend will be evaluated, then all req* rules from the backend
8622 will be evaluated. The reverse path is applied to responses.
8623
8624 - req* statements are applied after "block" statements, so that "block" is
8625 always the first one, but before "use_backend" in order to permit rewriting
Willy Tarreaud72758d2010-01-12 10:42:19 +01008626 before switching.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008627
8628
Willy Tarreau74ca5042013-06-11 23:12:07 +020086297. Using ACLs and fetching samples
8630----------------------------------
8631
8632Haproxy is capable of extracting data from request or response streams, from
8633client or server information, from tables, environmental information etc...
8634The action of extracting such data is called fetching a sample. Once retrieved,
8635these samples may be used for various purposes such as a key to a stick-table,
8636but most common usages consist in matching them against predefined constant
8637data called patterns.
8638
8639
86407.1. ACL basics
8641---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008642
8643The use of Access Control Lists (ACL) provides a flexible solution to perform
8644content switching and generally to take decisions based on content extracted
8645from the request, the response or any environmental status. The principle is
8646simple :
8647
Willy Tarreau74ca5042013-06-11 23:12:07 +02008648 - extract a data sample from a stream, table or the environment
Willy Tarreaue6b11e42013-11-26 19:02:32 +01008649 - optionally apply some format conversion to the extracted sample
Willy Tarreau74ca5042013-06-11 23:12:07 +02008650 - apply one or multiple pattern matching methods on this sample
8651 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008652
Willy Tarreau74ca5042013-06-11 23:12:07 +02008653The actions generally consist in blocking a request, selecting a backend, or
8654adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008655
8656In order to define a test, the "acl" keyword is used. The syntax is :
8657
Willy Tarreau74ca5042013-06-11 23:12:07 +02008658 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008659
8660This creates a new ACL <aclname> or completes an existing one with new tests.
8661Those tests apply to the portion of request/response specified in <criterion>
8662and may be adjusted with optional flags [flags]. Some criteria also support
Willy Tarreaue6b11e42013-11-26 19:02:32 +01008663an operator which may be specified before the set of values. Optionally some
8664conversion operators may be applied to the sample, and they will be specified
8665as a comma-delimited list of keywords just after the first keyword. The values
8666are of the type supported by the criterion, and are separated by spaces.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008667
8668ACL names must be formed from upper and lower case letters, digits, '-' (dash),
8669'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
8670which means that "my_acl" and "My_Acl" are two different ACLs.
8671
8672There is no enforced limit to the number of ACLs. The unused ones do not affect
8673performance, they just consume a small amount of memory.
8674
Willy Tarreau74ca5042013-06-11 23:12:07 +02008675The criterion generally is the name of a sample fetch method, or one of its ACL
8676specific declinations. The default test method is implied by the output type of
8677this sample fetch method. The ACL declinations can describe alternate matching
Willy Tarreaue6b11e42013-11-26 19:02:32 +01008678methods of a same sample fetch method. The sample fetch methods are the only
8679ones supporting a conversion.
Willy Tarreau74ca5042013-06-11 23:12:07 +02008680
8681Sample fetch methods return data which can be of the following types :
8682 - boolean
8683 - integer (signed or unsigned)
8684 - IPv4 or IPv6 address
8685 - string
8686 - data block
8687
Willy Tarreaue6b11e42013-11-26 19:02:32 +01008688Converters transform any of these data into any of these. For example, some
8689converters might convert a string to a lower-case string while other ones
8690would turn a string to an IPv4 address, or apply a netmask to an IP address.
8691The resulting sample is of the type of the last converter applied to the list,
8692which defaults to the type of the sample fetch method.
8693
Willy Tarreau74ca5042013-06-11 23:12:07 +02008694The ACL engine can match these types against patterns of the following types :
8695 - boolean
8696 - integer or integer range
8697 - IP address / network
8698 - string (exact, substring, suffix, prefix, subdir, domain)
8699 - regular expression
8700 - hex block
8701
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008702The following ACL flags are currently supported :
8703
Willy Tarreau2b5285d2010-05-09 23:45:24 +02008704 -i : ignore case during matching of all subsequent patterns.
8705 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +02008706 -m : use a specific pattern matching method
Willy Tarreau6a06a402007-07-15 20:15:28 +02008707 -- : force end of flags. Useful when a string looks like one of the flags.
8708
Willy Tarreau74ca5042013-06-11 23:12:07 +02008709The "-f" flag is followed by the name of a file from which all lines will be
8710read as individual values. It is even possible to pass multiple "-f" arguments
8711if the patterns are to be loaded from multiple files. Empty lines as well as
8712lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
8713will be stripped. If it is absolutely necessary to insert a valid pattern
8714beginning with a sharp, just prefix it with a space so that it is not taken for
8715a comment. Depending on the data type and match method, haproxy may load the
8716lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
8717exact string matching. In this case, duplicates will automatically be removed.
8718
8719Also, note that the "-i" flag applies to subsequent entries and not to entries
8720loaded from files preceding it. For instance :
8721
8722 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
8723
8724In this example, each line of "exact-ua.lst" will be exactly matched against
8725the "user-agent" header of the request. Then each line of "generic-ua" will be
8726case-insensitively matched. Then the word "test" will be insensitively matched
8727as well.
8728
8729The "-m" flag is used to select a specific pattern matching method on the input
8730sample. All ACL-specific criteria imply a pattern matching method and generally
8731do not need this flag. However, this flag is useful with generic sample fetch
8732methods to describe how they're going to be matched against the patterns. This
8733is required for sample fetches which return data type for which there is no
8734obvious matching method (eg: string or binary). When "-m" is specified and
8735followed by a pattern matching method name, this method is used instead of the
8736default one for the criterion. This makes it possible to match contents in ways
8737that were not initially planned, or with sample fetch methods which return a
8738string. The matching method also affects the way the patterns are parsed.
8739
8740There are some restrictions however. Not all methods can be used with all
8741sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
8742be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +02008743
8744 - "found" : only check if the requested sample could be found in the stream,
8745 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +02008746 to pass any pattern to avoid confusion. This matching method is
8747 particularly useful to detect presence of certain contents such
8748 as headers, cookies, etc... even if they are empty and without
8749 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008750
8751 - "bool" : check the value as a boolean. It can only be applied to fetches
8752 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +02008753 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008754
8755 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +02008756 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008757
8758 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +02008759 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008760
8761 - "bin" : match the contents against an hexadecimal string representing a
8762 binary sequence. This may be used with binary or string samples.
8763
8764 - "len" : match the sample's length as an integer. This may be used with
8765 binary or string samples.
8766
Willy Tarreau74ca5042013-06-11 23:12:07 +02008767 - "str" : exact match : match the contents against a string. This may be
8768 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008769
Willy Tarreau74ca5042013-06-11 23:12:07 +02008770 - "sub" : substring match : check that the contents contain at least one of
8771 the provided string patterns. This may be used with binary or
8772 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008773
Willy Tarreau74ca5042013-06-11 23:12:07 +02008774 - "reg" : regex match : match the contents against a list of regular
8775 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008776
Willy Tarreau74ca5042013-06-11 23:12:07 +02008777 - "beg" : prefix match : check that the contents begin like the provided
8778 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008779
Willy Tarreau74ca5042013-06-11 23:12:07 +02008780 - "end" : suffix match : check that the contents end like the provided
8781 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008782
Willy Tarreau74ca5042013-06-11 23:12:07 +02008783 - "dir" : subdir match : check that a slash-delimited portion of the
8784 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008785 This may be used with binary or string samples.
8786
Willy Tarreau74ca5042013-06-11 23:12:07 +02008787 - "dom" : domain match : check that a dot-delimited portion of the contents
8788 exactly match one of the provided string patterns. This may be
8789 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02008790
8791For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
8792request, it is possible to do :
8793
8794 acl jsess_present cook(JSESSIONID) -m found
8795
8796In order to apply a regular expression on the 500 first bytes of data in the
8797buffer, one would use the following acl :
8798
8799 acl script_tag payload(0,500) -m reg -i <script>
8800
Willy Tarreaue6b11e42013-11-26 19:02:32 +01008801On systems where the regex library is much slower when using "-i", it is
8802possible to convert the sample to lowercase before matching, like this :
8803
8804 acl script_tag payload(0,500),lower -m reg <script>
8805
Willy Tarreau74ca5042013-06-11 23:12:07 +02008806All ACL-specific criteria imply a default matching method. Most often, these
8807criteria are composed by concatenating the name of the original sample fetch
8808method and the matching method. For example, "hdr_beg" applies the "beg" match
8809to samples retrieved using the "hdr" fetch method. Since all ACL-specific
8810criteria rely on a sample fetch method, it is always possible instead to use
8811the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +02008812
Willy Tarreau74ca5042013-06-11 23:12:07 +02008813If an alternate match is specified using "-m" on an ACL-specific criterion,
8814the mathing method is simply applied to the underlying sample fetch method. For
8815example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +02008816
Willy Tarreau74ca5042013-06-11 23:12:07 +02008817 acl short_form hdr_beg(host) www.
8818 acl alternate1 hdr_beg(host) -m beg www.
8819 acl alternate2 hdr_dom(host) -m beg www.
8820 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +02008821
Willy Tarreau2b5285d2010-05-09 23:45:24 +02008822
Willy Tarreaue6b11e42013-11-26 19:02:32 +01008823The table below summarizes the compatibility matrix between sample types
Willy Tarreau74ca5042013-06-11 23:12:07 +02008824and the pattern types to fetch against. It indicates for each compatible
8825combination the name of the matching method to be used, prefixed with "*" when
8826the method is implicit and will work by default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +01008827
Willy Tarreau74ca5042013-06-11 23:12:07 +02008828 +-------------------------------------------------+
8829 | Input sample type |
8830 +----------------------+---------+---------+---------+---------+---------+
8831 | pattern type | boolean | integer | IP | string | binary |
8832 +----------------------+---------+---------+---------+---------+---------+
8833 | none (presence only) | found | found | found | found | found |
8834 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01008835 | none (boolean value) | *bool | bool | | bool | |
Willy Tarreau74ca5042013-06-11 23:12:07 +02008836 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01008837 | integer (value) | int | *int | int | int | |
Willy Tarreau74ca5042013-06-11 23:12:07 +02008838 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01008839 | integer (length) | len | len | len | len | len |
Willy Tarreau74ca5042013-06-11 23:12:07 +02008840 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01008841 | IP address | | | *ip | ip | ip |
Willy Tarreau74ca5042013-06-11 23:12:07 +02008842 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01008843 | exact string | str | str | str | str | str |
Willy Tarreau74ca5042013-06-11 23:12:07 +02008844 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01008845 | prefix | beg | beg | beg | beg | beg |
Willy Tarreau74ca5042013-06-11 23:12:07 +02008846 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01008847 | suffix | end | end | end | end | end |
Willy Tarreau74ca5042013-06-11 23:12:07 +02008848 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01008849 | substring | sub | sub | sub | sub | sub |
Willy Tarreau74ca5042013-06-11 23:12:07 +02008850 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01008851 | subdir | dir | dir | dir | dir | dir |
Willy Tarreau74ca5042013-06-11 23:12:07 +02008852 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01008853 | domain | dom | dom | dom | dom | dom |
Willy Tarreau74ca5042013-06-11 23:12:07 +02008854 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01008855 | regex | reg | reg | reg | reg | reg |
Willy Tarreau74ca5042013-06-11 23:12:07 +02008856 +----------------------+---------+---------+---------+---------+---------+
8857 | hex block | | | | bin | bin |
8858 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +02008859
8860
Willy Tarreau74ca5042013-06-11 23:12:07 +020088617.1.1. Matching booleans
8862------------------------
8863
8864In order to match a boolean, no value is needed and all values are ignored.
8865Boolean matching is used by default for all fetch methods of type "boolean".
8866When boolean matching is used, the fetched value is returned as-is, which means
8867that a boolean "true" will always match and a boolean "false" will never match.
8868
8869Boolean matching may also be enforced using "-m bool" on fetch methods which
8870return an integer value. Then, integer value 0 is converted to the boolean
8871"false" and all other values are converted to "true".
8872
Willy Tarreau6a06a402007-07-15 20:15:28 +02008873
Willy Tarreau74ca5042013-06-11 23:12:07 +020088747.1.2. Matching integers
8875------------------------
8876
8877Integer matching applies by default to integer fetch methods. It can also be
8878enforced on boolean fetches using "-m int". In this case, "false" is converted
8879to the integer 0, and "true" is converted to the integer 1.
8880
8881Integer matching also supports integer ranges and operators. Note that integer
8882matching only applies to positive values. A range is a value expressed with a
8883lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +02008884
8885For instance, "1024:65535" is a valid range to represent a range of
8886unprivileged ports, and "1024:" would also work. "0:1023" is a valid
8887representation of privileged ports, and ":1023" would also work.
8888
Willy Tarreau62644772008-07-16 18:36:06 +02008889As a special case, some ACL functions support decimal numbers which are in fact
8890two integers separated by a dot. This is used with some version checks for
8891instance. All integer properties apply to those decimal numbers, including
8892ranges and operators.
8893
Willy Tarreau6a06a402007-07-15 20:15:28 +02008894For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +01008895operators with ranges does not make much sense and is strongly discouraged.
8896Similarly, it does not make much sense to perform order comparisons with a set
8897of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +02008898
Willy Tarreau0ba27502007-12-24 16:55:16 +01008899Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +02008900
8901 eq : true if the tested value equals at least one value
8902 ge : true if the tested value is greater than or equal to at least one value
8903 gt : true if the tested value is greater than at least one value
8904 le : true if the tested value is less than or equal to at least one value
8905 lt : true if the tested value is less than at least one value
8906
Willy Tarreau0ba27502007-12-24 16:55:16 +01008907For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +02008908
8909 acl negative-length hdr_val(content-length) lt 0
8910
Willy Tarreau62644772008-07-16 18:36:06 +02008911This one matches SSL versions between 3.0 and 3.1 (inclusive) :
8912
8913 acl sslv3 req_ssl_ver 3:3.1
8914
Willy Tarreau6a06a402007-07-15 20:15:28 +02008915
Willy Tarreau74ca5042013-06-11 23:12:07 +020089167.1.3. Matching strings
8917-----------------------
8918
8919String matching applies to string or binary fetch methods, and exists in 6
8920different forms :
8921
8922 - exact match (-m str) : the extracted string must exactly match the
8923 patterns ;
8924
8925 - substring match (-m sub) : the patterns are looked up inside the
8926 extracted string, and the ACL matches if any of them is found inside ;
8927
8928 - prefix match (-m beg) : the patterns are compared with the beginning of
8929 the extracted string, and the ACL matches if any of them matches.
8930
8931 - suffix match (-m end) : the patterns are compared with the end of the
8932 extracted string, and the ACL matches if any of them matches.
8933
8934 - subdir match (-m sub) : the patterns are looked up inside the extracted
8935 string, delimited with slashes ("/"), and the ACL matches if any of them
8936 matches.
8937
8938 - domain match (-m dom) : the patterns are looked up inside the extracted
8939 string, delimited with dots ("."), and the ACL matches if any of them
8940 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +02008941
8942String matching applies to verbatim strings as they are passed, with the
8943exception of the backslash ("\") which makes it possible to escape some
8944characters such as the space. If the "-i" flag is passed before the first
8945string, then the matching will be performed ignoring the case. In order
8946to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +01008947before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +02008948
8949
Willy Tarreau74ca5042013-06-11 23:12:07 +020089507.1.4. Matching regular expressions (regexes)
8951---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02008952
8953Just like with string matching, regex matching applies to verbatim strings as
8954they are passed, with the exception of the backslash ("\") which makes it
8955possible to escape some characters such as the space. If the "-i" flag is
8956passed before the first regex, then the matching will be performed ignoring
8957the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +01008958the "--" flag before the first string. Same principle applies of course to
8959match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +02008960
8961
Willy Tarreau74ca5042013-06-11 23:12:07 +020089627.1.5. Matching arbitrary data blocks
8963-------------------------------------
8964
8965It is possible to match some extracted samples against a binary block which may
8966not safely be represented as a string. For this, the patterns must be passed as
8967a series of hexadecimal digits in an even number, when the match method is set
8968to binary. Each sequence of two digits will represent a byte. The hexadecimal
8969digits may be used upper or lower case.
8970
8971Example :
8972 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
8973 acl hello payload(0,6) -m bin 48656c6c6f0a
8974
8975
89767.1.6. Matching IPv4 and IPv6 addresses
8977---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02008978
8979IPv4 addresses values can be specified either as plain addresses or with a
8980netmask appended, in which case the IPv4 address matches whenever it is
8981within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01008982host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +01008983difficult to read and debug configurations. If hostnames are used, you should
8984at least ensure that they are present in /etc/hosts so that the configuration
8985does not depend on any random DNS match at the moment the configuration is
8986parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +02008987
Willy Tarreauceb4ac92012-04-28 00:41:46 +02008988IPv6 may be entered in their usual form, with or without a netmask appended.
8989Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
8990trouble with randomly resolved IP addresses, host names are never allowed in
8991IPv6 patterns.
8992
8993HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
8994following situations :
8995 - tested address is IPv4, pattern address is IPv4, the match applies
8996 in IPv4 using the supplied mask if any.
8997 - tested address is IPv6, pattern address is IPv6, the match applies
8998 in IPv6 using the supplied mask if any.
8999 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
9000 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
9001 ::IPV4 or ::ffff:IPV4, otherwise it fails.
9002 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
9003 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
9004 applied in IPv6 using the supplied IPv6 mask.
9005
Willy Tarreau74ca5042013-06-11 23:12:07 +02009006
90077.2. Using ACLs to form conditions
9008----------------------------------
9009
9010Some actions are only performed upon a valid condition. A condition is a
9011combination of ACLs with operators. 3 operators are supported :
9012
9013 - AND (implicit)
9014 - OR (explicit with the "or" keyword or the "||" operator)
9015 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +02009016
Willy Tarreau74ca5042013-06-11 23:12:07 +02009017A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +02009018
Willy Tarreau74ca5042013-06-11 23:12:07 +02009019 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +02009020
Willy Tarreau74ca5042013-06-11 23:12:07 +02009021Such conditions are generally used after an "if" or "unless" statement,
9022indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +02009023
Willy Tarreau74ca5042013-06-11 23:12:07 +02009024For instance, to block HTTP requests to the "*" URL with methods other than
9025"OPTIONS", as well as POST requests without content-length, and GET or HEAD
9026requests with a content-length greater than 0, and finally every request which
9027is not either GET/HEAD/POST/OPTIONS !
9028
9029 acl missing_cl hdr_cnt(Content-length) eq 0
9030 block if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
9031 block if METH_GET HTTP_CONTENT
9032 block unless METH_GET or METH_POST or METH_OPTIONS
9033
9034To select a different backend for requests to static contents on the "www" site
9035and to every request on the "img", "video", "download" and "ftp" hosts :
9036
9037 acl url_static path_beg /static /images /img /css
9038 acl url_static path_end .gif .png .jpg .css .js
9039 acl host_www hdr_beg(host) -i www
9040 acl host_static hdr_beg(host) -i img. video. download. ftp.
9041
9042 # now use backend "static" for all static-only hosts, and for static urls
9043 # of host "www". Use backend "www" for the rest.
9044 use_backend static if host_static or host_www url_static
9045 use_backend www if host_www
9046
9047It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
9048expressions that are built on the fly without needing to be declared. They must
9049be enclosed between braces, with a space before and after each brace (because
9050the braces must be seen as independent words). Example :
9051
9052 The following rule :
9053
9054 acl missing_cl hdr_cnt(Content-length) eq 0
9055 block if METH_POST missing_cl
9056
9057 Can also be written that way :
9058
9059 block if METH_POST { hdr_cnt(Content-length) eq 0 }
9060
9061It is generally not recommended to use this construct because it's a lot easier
9062to leave errors in the configuration when written that way. However, for very
9063simple rules matching only one source IP address for instance, it can make more
9064sense to use them than to declare ACLs with random names. Another example of
9065good use is the following :
9066
9067 With named ACLs :
9068
9069 acl site_dead nbsrv(dynamic) lt 2
9070 acl site_dead nbsrv(static) lt 2
9071 monitor fail if site_dead
9072
9073 With anonymous ACLs :
9074
9075 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
9076
9077See section 4.2 for detailed help on the "block" and "use_backend" keywords.
9078
9079
90807.3. Fetching samples
9081---------------------
9082
9083Historically, sample fetch methods were only used to retrieve data to match
9084against patterns using ACLs. With the arrival of stick-tables, a new class of
9085sample fetch methods was created, most often sharing the same syntax as their
9086ACL counterpart. These sample fetch methods are also known as "fetches". As
9087of now, ACLs and fetches have converged. All ACL fetch methods have been made
9088available as fetch methods, and ACLs may use any sample fetch method as well.
9089
9090This section details all available sample fetch methods and their output type.
9091Some sample fetch methods have deprecated aliases that are used to maintain
9092compatibility with existing configurations. They are then explicitly marked as
9093deprecated and should not be used in new setups.
9094
9095The ACL derivatives are also indicated when available, with their respective
9096matching methods. These ones all have a well defined default pattern matching
9097method, so it is never necessary (though allowed) to pass the "-m" option to
9098indicate how the sample will be matched using ACLs.
9099
9100As indicated in the sample type versus matching compatibility matrix above,
9101when using a generic sample fetch method in an ACL, the "-m" option is
9102mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
9103the same keyword exists as an ACL keyword and as a standard fetch method, the
9104ACL engine will automatically pick the ACL-only one by default.
9105
9106Some of these keywords support one or multiple mandatory arguments, and one or
9107multiple optional arguments. These arguments are strongly typed and are checked
9108when the configuration is parsed so that there is no risk of running with an
9109incorrect argument (eg: an unresolved backend name). Fetch function arguments
9110are passed between parenthesis and are delimited by commas. When an argument
9111is optional, it will be indicated below between square brackets ('[ ]'). When
9112all arguments are optional, the parenthesis may be omitted.
9113
9114Thus, the syntax of a standard sample fetch method is one of the following :
9115 - name
9116 - name(arg1)
9117 - name(arg1,arg2)
9118
Willy Tarreaue6b11e42013-11-26 19:02:32 +01009119Sample fetch methods may be combined with transformations to be applied on top
9120of the fetched sample (also called "converters"). These combinations form what
9121is called "sample expressions" and the result is a "sample". Initially this
9122was only supported by "stick on" and "stick store-request" directives but this
9123has now be extended to all places where samples may be used (acls, log-format,
9124unique-id-format, add-header, ...).
9125
9126These transformations are enumerated as a series of specific keywords after the
9127sample fetch method. These keywords may equally be appended immediately after
9128the fetch keyword's argument, delimited by a comma. These keywords can also
9129support some arguments (eg: a netmask) which must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +01009130
Willy Tarreau74ca5042013-06-11 23:12:07 +02009131The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +01009132
Willy Tarreau74ca5042013-06-11 23:12:07 +02009133 lower Convert a string sample to lower case. This can only be placed
9134 after a string sample fetch function or after a transformation
9135 keyword returning a string type. The result is of type string.
Willy Tarreau6a06a402007-07-15 20:15:28 +02009136
Willy Tarreau74ca5042013-06-11 23:12:07 +02009137 upper Convert a string sample to upper case. This can only be placed
9138 after a string sample fetch function or after a transformation
9139 keyword returning a string type. The result is of type string.
Willy Tarreau6a06a402007-07-15 20:15:28 +02009140
Willy Tarreau74ca5042013-06-11 23:12:07 +02009141 ipmask(<mask>) Apply a mask to an IPv4 address, and use the result for
9142 lookups and storage. This can be used to make all hosts within
9143 a certain mask to share the same table entries and as such use
9144 the same server. The mask can be passed in dotted form (eg:
9145 255.255.255.0) or in CIDR form (eg: 24).
9146
Willy Tarreau276fae92013-07-25 14:36:01 +02009147 http_date([<offset>])
9148 Converts an integer supposed to contain a date since epoch to
9149 a string representing this date in a format suitable for use
9150 in HTTP header fields. If an offset value is specified, then
9151 it is a number of seconds that is added to the date before the
9152 conversion is operated. This is particularly useful to emit
9153 Date header fields, Expires values in responses when combined
9154 with a positive offset, or Last-Modified values when the
9155 offset is negative.
Willy Tarreau74ca5042013-06-11 23:12:07 +02009156
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +01009157 map(<map_file>[,<default_value>])
9158 map_<match_type>(<map_file>[,<default_value>])
9159 map_<match_type>_<output_type>(<map_file>[,<default_value>])
9160 Search the input value from <map_file> using the <match_type>
9161 matching method, and return the associated value converted to
9162 the type <output_type>. If the input value cannot be found in
9163 the <map_file>, the converter returns the <default_value>. If
9164 the <default_value> is not set, the converter fails and acts
9165 as if no input value could be fetched. If the <match_type> is
9166 not set, it defaults to "str". Likewise, if the <output_type>
9167 is not set, it defaults to "str". For convenience, the "map"
9168 keyword is an alias for "map_str" and maps a string to another
9169 string. The following array contains contains the list of all
9170 the map* converters.
9171
9172 It is important to avoid overlapping between the keys : IP
9173 addresses and strings are stored in trees, so the first of the
9174 finest match will be used. Other keys are stored in lists, so
9175 the first matching occurrence will be used.
9176
9177 +----+----------+---------+-------------+------------+
9178 | `-_ out | | | |
9179 | input `-_ | str | int | ip |
9180 | / match `-_ | | | |
9181 +---------------+---------+-------------+------------+
9182 | str / str | map_str | map_str_int | map_str_ip |
9183 | str / sub | map_sub | map_sub_int | map_sub_ip |
9184 | str / dir | map_dir | map_dir_int | map_dir_ip |
9185 | str / dom | map_dom | map_dom_int | map_dom_ip |
9186 | str / end | map_end | map_end_int | map_end_ip |
9187 | str / reg | map_reg | map_reg_int | map_reg_ip |
9188 | int / int | map_int | map_int_int | map_int_ip |
9189 | ip / ip | map_ip | map_ip_int | map_ip_ip |
9190 +---------------+---------+-------------+------------+
9191
9192 The file contains one key + value per line. Lines which start
9193 with '#' are ignored, just like empty lines. Leading tabs and
9194 spaces are stripped. The key is then the first "word" (series
9195 of non-space/tabs characters), and the value is what follows
9196 this series of space/tab till the end of the line excluding
9197 trailing spaces/tabs.
9198
9199 Example :
9200
9201 # this is a comment and is ignored
9202 2.22.246.0/23 United Kingdom \n
9203 <-><-----------><--><------------><---->
9204 | | | | `- trailing spaces ignored
9205 | | | `----------- value
9206 | | `--------------------- middle spaces ignored
9207 | `---------------------------- key
9208 `------------------------------------ leading spaces ignored
9209
Willy Tarreau74ca5042013-06-11 23:12:07 +020092107.3.1. Fetching samples from internal states
9211--------------------------------------------
9212
9213A first set of sample fetch methods applies to internal information which does
9214not even relate to any client information. These ones are sometimes used with
9215"monitor-fail" directives to report an internal status to external watchers.
9216The sample fetch methods described in this section are usable anywhere.
9217
9218always_false : boolean
9219 Always returns the boolean "false" value. It may be used with ACLs as a
9220 temporary replacement for another one when adjusting configurations.
9221
9222always_true : boolean
9223 Always returns the boolean "true" value. It may be used with ACLs as a
9224 temporary replacement for another one when adjusting configurations.
9225
9226avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +01009227 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +02009228 divided by the number of active servers. The current backend is used if no
9229 backend is specified. This is very similar to "queue" except that the size of
9230 the farm is considered, in order to give a more accurate measurement of the
9231 time it may take for a new connection to be processed. The main usage is with
9232 ACL to return a sorry page to new users when it becomes certain they will get
9233 a degraded service, or to pass to the backend servers in a header so that
9234 they decide to work in degraded mode or to disable some functions to speed up
9235 the processing a bit. Note that in the event there would not be any active
9236 server anymore, twice the number of queued connections would be considered as
9237 the measured value. This is a fair estimate, as we expect one server to get
9238 back soon anyway, but we still prefer to send new traffic to another backend
9239 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
9240 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +01009241
Willy Tarreau74ca5042013-06-11 23:12:07 +02009242be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +02009243 Applies to the number of currently established connections on the backend,
9244 possibly including the connection being evaluated. If no backend name is
9245 specified, the current one is used. But it is also possible to check another
9246 backend. It can be used to use a specific farm when the nominal one is full.
9247 See also the "fe_conn", "queue" and "be_sess_rate" criteria.
Willy Tarreau6a06a402007-07-15 20:15:28 +02009248
Willy Tarreau74ca5042013-06-11 23:12:07 +02009249be_sess_rate([<backend>]) : integer
9250 Returns an integer value corresponding to the sessions creation rate on the
9251 backend, in number of new sessions per second. This is used with ACLs to
9252 switch to an alternate backend when an expensive or fragile one reaches too
9253 high a session rate, or to limit abuse of service (eg. prevent sucking of an
9254 online dictionary). It can also be useful to add this element to logs using a
9255 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009256
9257 Example :
9258 # Redirect to an error page if the dictionary is requested too often
9259 backend dynamic
9260 mode http
9261 acl being_scanned be_sess_rate gt 100
9262 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +01009263
Willy Tarreau74ca5042013-06-11 23:12:07 +02009264connslots([<backend>]) : integer
9265 Returns an integer value corresponding to the number of connection slots
9266 still available in the backend, by totalizing the maximum amount of
9267 connections on all servers and the maximum queue size. This is probably only
9268 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -05009269
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +08009270 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +02009271 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +08009272 usage; see "use_backend" keyword) can be redirected to a different backend.
9273
Willy Tarreau55165fe2009-05-10 12:02:55 +02009274 'connslots' = number of available server connection slots, + number of
9275 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +08009276
Willy Tarreaua36af912009-10-10 12:02:45 +02009277 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +02009278 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +02009279 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +02009280 you want to be able to differentiate between different backends, and their
9281 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +02009282 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +02009283 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +08009284
Willy Tarreau55165fe2009-05-10 12:02:55 +02009285 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
9286 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +02009287 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +02009288 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +08009289
Willy Tarreau6236d3a2013-07-25 14:28:25 +02009290date([<offset>]) : integer
9291 Returns the current date as the epoch (number of seconds since 01/01/1970).
9292 If an offset value is specified, then it is a number of seconds that is added
9293 to the current date before returning the value. This is particularly useful
9294 to compute relative dates, as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +02009295 It is useful combined with the http_date converter.
9296
9297 Example :
9298
9299 # set an expires header to now+1 hour in every response
9300 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +02009301
Willy Tarreau595ec542013-06-12 21:34:28 +02009302env(<name>) : string
9303 Returns a string containing the value of environment variable <name>. As a
9304 reminder, environment variables are per-process and are sampled when the
9305 process starts. This can be useful to pass some information to a next hop
9306 server, or with ACLs to take specific action when the process is started a
9307 certain way.
9308
9309 Examples :
9310 # Pass the Via header to next hop with the local hostname in it
9311 http-request add-header Via 1.1\ %[env(HOSTNAME)]
9312
9313 # reject cookie-less requests when the STOP environment variable is set
9314 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
9315
Willy Tarreau74ca5042013-06-11 23:12:07 +02009316fe_conn([<frontend>]) : integer
9317 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +01009318 possibly including the connection being evaluated. If no frontend name is
9319 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +02009320 frontend. It can be used to return a sorry page before hard-blocking, or to
9321 use a specific backend to drain new requests when the farm is considered
9322 full. This is mostly used with ACLs but can also be used to pass some
9323 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
9324 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +02009325
Willy Tarreau74ca5042013-06-11 23:12:07 +02009326fe_sess_rate([<frontend>]) : integer
9327 Returns an integer value corresponding to the sessions creation rate on the
9328 frontend, in number of new sessions per second. This is used with ACLs to
9329 limit the incoming session rate to an acceptable range in order to prevent
9330 abuse of service at the earliest moment, for example when combined with other
9331 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
9332 down below the limit. It can also be useful to add this element to logs using
9333 a log-format directive. See also the "rate-limit sessions" directive for use
9334 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +01009335
9336 Example :
9337 # This frontend limits incoming mails to 10/s with a max of 100
9338 # concurrent connections. We accept any connection below 10/s, and
9339 # force excess clients to wait for 100 ms. Since clients are limited to
9340 # 100 max, there cannot be more than 10 incoming mails per second.
9341 frontend mail
9342 bind :25
9343 mode tcp
9344 maxconn 100
9345 acl too_fast fe_sess_rate ge 10
9346 tcp-request inspect-delay 100ms
9347 tcp-request content accept if ! too_fast
9348 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +01009349
Willy Tarreau74ca5042013-06-11 23:12:07 +02009350nbsrv([<backend>]) : integer
9351 Returns an integer value corresponding to the number of usable servers of
9352 either the current backend or the named backend. This is mostly used with
9353 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +01009354 switch to an alternate backend when the number of servers is too low to
9355 to handle some load. It is useful to report a failure when combined with
9356 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +01009357
Willy Tarreau74ca5042013-06-11 23:12:07 +02009358queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +01009359 Returns the total number of queued connections of the designated backend,
9360 including all the connections in server queues. If no backend name is
9361 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +02009362 one. This is useful with ACLs or to pass statistics to backend servers. This
9363 can be used to take actions when queuing goes above a known level, generally
9364 indicating a surge of traffic or a massive slowdown on the servers. One
9365 possible action could be to reject new users but still accept old ones. See
9366 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
9367
9368srv_conn([<backend>/]<server>) : integer
9369 Returns an integer value corresponding to the number of currently established
9370 connections on the designated server, possibly including the connection being
9371 evaluated. If <backend> is omitted, then the server is looked up in the
9372 current backend. It can be used to use a specific farm when one server is
9373 full, or to inform the server about our view of the number of active
9374 connections with it. See also the "fe_conn", "be_conn" and "queue" fetch
9375 methods.
9376
9377srv_is_up([<backend>/]<server>) : boolean
9378 Returns true when the designated server is UP, and false when it is either
9379 DOWN or in maintenance mode. If <backend> is omitted, then the server is
9380 looked up in the current backend. It is mainly used to take action based on
9381 an external status reported via a health check (eg: a geographical site's
9382 availability). Another possible use which is more of a hack consists in
9383 using dummy servers as boolean variables that can be enabled or disabled from
9384 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
9385
9386srv_sess_rate([<backend>/]<server>) : integer
9387 Returns an integer corresponding to the sessions creation rate on the
9388 designated server, in number of new sessions per second. If <backend> is
9389 omitted, then the server is looked up in the current backend. This is mosly
9390 used with ACLs but can make sense with logs too. This is used to switch to an
9391 alternate backend when an expensive or fragile one reaches too high a session
9392 rate, or to limit abuse of service (eg. prevent latent requests from
9393 overloading servers).
9394
9395 Example :
9396 # Redirect to a separate back
9397 acl srv1_full srv_sess_rate(be1/srv1) gt 50
9398 acl srv2_full srv_sess_rate(be1/srv2) gt 50
9399 use_backend be2 if srv1_full or srv2_full
9400
9401table_avl([<table>]) : integer
9402 Returns the total number of available entries in the current proxy's
9403 stick-table or in the designated stick-table. See also table_cnt.
9404
9405table_cnt([<table>]) : integer
9406 Returns the total number of entries currently in use in the current proxy's
9407 stick-table or in the designated stick-table. See also src_conn_cnt and
9408 table_avl for other entry counting methods.
9409
9410
94117.3.2. Fetching samples at Layer 4
9412----------------------------------
9413
9414The layer 4 usually describes just the transport layer which in haproxy is
9415closest to the connection, where no content is yet made available. The fetch
9416methods described here are usable as low as the "tcp-request connection" rule
9417sets unless they require some future information. Those generally include
9418TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009419the incoming connection. For retrieving a value from a sticky counters, the
9420counter number can be explicitly set as 0, 1, or 2 using the pre-defined
9421"sc0_", "sc1_", or "sc2_" prefix, or it can be specified as the first integer
Willy Tarreau0f791d42013-07-23 19:56:43 +02009422argument when using the "sc_" prefix. An optional table may be specified with
9423the "sc*" form, in which case the currently tracked key will be looked up into
9424this alternate table instead of the table currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +02009425
9426be_id : integer
9427 Returns an integer containing the current backend's id. It can be used in
9428 frontends with responses to check which backend processed the request.
9429
9430dst : ip
9431 This is the destination IPv4 address of the connection on the client side,
9432 which is the address the client connected to. It can be useful when running
9433 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
9434 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
9435 RFC 4291.
9436
9437dst_conn : integer
9438 Returns an integer value corresponding to the number of currently established
9439 connections on the same socket including the one being evaluated. It is
9440 normally used with ACLs but can as well be used to pass the information to
9441 servers in an HTTP header or in logs. It can be used to either return a sorry
9442 page before hard-blocking, or to use a specific backend to drain new requests
9443 when the socket is considered saturated. This offers the ability to assign
9444 different limits to different listening ports or addresses. See also the
9445 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009446
Willy Tarreau74ca5042013-06-11 23:12:07 +02009447dst_port : integer
9448 Returns an integer value corresponding to the destination TCP port of the
9449 connection on the client side, which is the port the client connected to.
9450 This might be used when running in transparent mode, when assigning dynamic
9451 ports to some clients for a whole application session, to stick all users to
9452 a same server, or to pass the destination port information to a server using
9453 an HTTP header.
9454
9455fe_id : integer
9456 Returns an integer containing the current frontend's id. It can be used in
9457 backends to check from which backend it was called, or to stick all users
9458 coming via a same frontend to the same server.
9459
Willy Tarreau0f791d42013-07-23 19:56:43 +02009460sc_bytes_in_rate(<ctr>,[<table>]) : integer
9461sc0_bytes_in_rate([<table>]) : integer
9462sc1_bytes_in_rate([<table>]) : integer
9463sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009464 Returns the average client-to-server bytes rate from the currently tracked
9465 counters, measured in amount of bytes over the period configured in the
9466 table. See also src_bytes_in_rate.
9467
Willy Tarreau0f791d42013-07-23 19:56:43 +02009468sc_bytes_out_rate(<ctr>,[<table>]) : integer
9469sc0_bytes_out_rate([<table>]) : integer
9470sc1_bytes_out_rate([<table>]) : integer
9471sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009472 Returns the average server-to-client bytes rate from the currently tracked
9473 counters, measured in amount of bytes over the period configured in the
9474 table. See also src_bytes_out_rate.
9475
Willy Tarreau0f791d42013-07-23 19:56:43 +02009476sc_clr_gpc0(<ctr>,[<table>]) : integer
9477sc0_clr_gpc0([<table>]) : integer
9478sc1_clr_gpc0([<table>]) : integer
9479sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +02009480 Clears the first General Purpose Counter associated to the currently tracked
9481 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +01009482 stored value is zero, so first invocation will always return zero. This is
9483 typically used as a second ACL in an expression in order to mark a connection
9484 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +02009485
9486 # block if 5 consecutive requests continue to come faster than 10 sess
9487 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009488 acl abuse sc0_http_req_rate gt 10
9489 acl kill sc0_inc_gpc0 gt 5
9490 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +02009491 tcp-request connection accept if !abuse save
9492 tcp-request connection reject if abuse kill
9493
Willy Tarreau0f791d42013-07-23 19:56:43 +02009494sc_conn_cnt(<ctr>,[<table>]) : integer
9495sc0_conn_cnt([<table>]) : integer
9496sc1_conn_cnt([<table>]) : integer
9497sc2_conn_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009498 Returns the cumulated number of incoming connections from currently tracked
9499 counters. See also src_conn_cnt.
9500
Willy Tarreau0f791d42013-07-23 19:56:43 +02009501sc_conn_cur(<ctr>,[<table>]) : integer
9502sc0_conn_cur([<table>]) : integer
9503sc1_conn_cur([<table>]) : integer
9504sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009505 Returns the current amount of concurrent connections tracking the same
9506 tracked counters. This number is automatically incremented when tracking
9507 begins and decremented when tracking stops. See also src_conn_cur.
9508
Willy Tarreau0f791d42013-07-23 19:56:43 +02009509sc_conn_rate(<ctr>,[<table>]) : integer
9510sc0_conn_rate([<table>]) : integer
9511sc1_conn_rate([<table>]) : integer
9512sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009513 Returns the average connection rate from the currently tracked counters,
9514 measured in amount of connections over the period configured in the table.
9515 See also src_conn_rate.
9516
Willy Tarreau0f791d42013-07-23 19:56:43 +02009517sc_get_gpc0(<ctr>,[<table>]) : integer
9518sc0_get_gpc0([<table>]) : integer
9519sc1_get_gpc0([<table>]) : integer
9520sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009521 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009522 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009523
Willy Tarreau0f791d42013-07-23 19:56:43 +02009524sc_gpc0_rate(<ctr>,[<table>]) : integer
9525sc0_gpc0_rate([<table>]) : integer
9526sc1_gpc0_rate([<table>]) : integer
9527sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009528 Returns the average increment rate of the first General Purpose Counter
9529 associated to the currently tracked counters. It reports the frequency
9530 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009531 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
9532 that the "gpc0_rate" counter must be stored in the stick-table for a value to
9533 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +02009534
Willy Tarreau0f791d42013-07-23 19:56:43 +02009535sc_http_err_cnt(<ctr>,[<table>]) : integer
9536sc0_http_err_cnt([<table>]) : integer
9537sc1_http_err_cnt([<table>]) : integer
9538sc2_http_err_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009539 Returns the cumulated number of HTTP errors from the currently tracked
9540 counters. This includes the both request errors and 4xx error responses.
9541 See also src_http_err_cnt.
9542
Willy Tarreau0f791d42013-07-23 19:56:43 +02009543sc_http_err_rate(<ctr>,[<table>]) : integer
9544sc0_http_err_rate([<table>]) : integer
9545sc1_http_err_rate([<table>]) : integer
9546sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009547 Returns the average rate of HTTP errors from the currently tracked counters,
9548 measured in amount of errors over the period configured in the table. This
9549 includes the both request errors and 4xx error responses. See also
9550 src_http_err_rate.
9551
Willy Tarreau0f791d42013-07-23 19:56:43 +02009552sc_http_req_cnt(<ctr>,[<table>]) : integer
9553sc0_http_req_cnt([<table>]) : integer
9554sc1_http_req_cnt([<table>]) : integer
9555sc2_http_req_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009556 Returns the cumulated number of HTTP requests from the currently tracked
9557 counters. This includes every started request, valid or not. See also
9558 src_http_req_cnt.
9559
Willy Tarreau0f791d42013-07-23 19:56:43 +02009560sc_http_req_rate(<ctr>,[<table>]) : integer
9561sc0_http_req_rate([<table>]) : integer
9562sc1_http_req_rate([<table>]) : integer
9563sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009564 Returns the average rate of HTTP requests from the currently tracked
9565 counters, measured in amount of requests over the period configured in
9566 the table. This includes every started request, valid or not. See also
9567 src_http_req_rate.
9568
Willy Tarreau0f791d42013-07-23 19:56:43 +02009569sc_inc_gpc0(<ctr>,[<table>]) : integer
9570sc0_inc_gpc0([<table>]) : integer
9571sc1_inc_gpc0([<table>]) : integer
9572sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009573 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +01009574 tracked counters, and returns its new value. Before the first invocation,
9575 the stored value is zero, so first invocation will increase it to 1 and will
9576 return 1. This is typically used as a second ACL in an expression in order
9577 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +02009578
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009579 acl abuse sc0_http_req_rate gt 10
9580 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +02009581 tcp-request connection reject if abuse kill
9582
Willy Tarreau0f791d42013-07-23 19:56:43 +02009583sc_kbytes_in(<ctr>,[<table>]) : integer
9584sc0_kbytes_in([<table>]) : integer
9585sc1_kbytes_in([<table>]) : integer
9586sc2_kbytes_in([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009587 Returns the amount of client-to-server data from the currently tracked
9588 counters, measured in kilobytes over the period configured in the table. The
9589 test is currently performed on 32-bit integers, which limits values to 4
9590 terabytes. See also src_kbytes_in.
9591
Willy Tarreau0f791d42013-07-23 19:56:43 +02009592sc_kbytes_out(<ctr>,[<table>]) : integer
9593sc0_kbytes_out([<table>]) : integer
9594sc1_kbytes_out([<table>]) : integer
9595sc2_kbytes_out([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009596 Returns the amount of server-to-client data from the currently tracked
9597 counters, measured in kilobytes over the period configured in the table. The
9598 test is currently performed on 32-bit integers, which limits values to 4
9599 terabytes. See also src_kbytes_out.
9600
Willy Tarreau0f791d42013-07-23 19:56:43 +02009601sc_sess_cnt(<ctr>,[<table>]) : integer
9602sc0_sess_cnt([<table>]) : integer
9603sc1_sess_cnt([<table>]) : integer
9604sc2_sess_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009605 Returns the cumulated number of incoming connections that were transformed
9606 into sessions, which means that they were accepted by a "tcp-request
9607 connection" rule, from the currently tracked counters. A backend may count
9608 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009609 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +02009610 with the client. See also src_sess_cnt.
9611
Willy Tarreau0f791d42013-07-23 19:56:43 +02009612sc_sess_rate(<ctr>,[<table>]) : integer
9613sc0_sess_rate([<table>]) : integer
9614sc1_sess_rate([<table>]) : integer
9615sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +02009616 Returns the average session rate from the currently tracked counters,
9617 measured in amount of sessions over the period configured in the table. A
9618 session is a connection that got past the early "tcp-request connection"
9619 rules. A backend may count more sessions than connections because each
9620 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009621 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +02009622
Willy Tarreau0f791d42013-07-23 19:56:43 +02009623sc_tracked(<ctr>,[<table>]) : boolean
9624sc0_tracked([<table>]) : boolean
9625sc1_tracked([<table>]) : boolean
9626sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +02009627 Returns true if the designated session counter is currently being tracked by
9628 the current session. This can be useful when deciding whether or not we want
9629 to set some values in a header passed to the server.
9630
Willy Tarreau0f791d42013-07-23 19:56:43 +02009631sc_trackers(<ctr>,[<table>]) : integer
9632sc0_trackers([<table>]) : integer
9633sc1_trackers([<table>]) : integer
9634sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +01009635 Returns the current amount of concurrent connections tracking the same
9636 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009637 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +01009638 that it does not rely on any stored information but on the table's reference
9639 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +02009640 may sometimes be more suited for layer7 tracking. It can be used to tell a
9641 server how many concurrent connections there are from a given address for
9642 example.
Willy Tarreau2406db42012-12-09 12:16:43 +01009643
Willy Tarreau74ca5042013-06-11 23:12:07 +02009644so_id : integer
9645 Returns an integer containing the current listening socket's id. It is useful
9646 in frontends involving many "bind" lines, or to stick all users coming via a
9647 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009648
Willy Tarreau74ca5042013-06-11 23:12:07 +02009649src : ip
9650 This is the source IPv4 address of the client of the session. It is of type
9651 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
9652 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
9653 TCP-level source address which is used, and not the address of a client
9654 behind a proxy. However if the "accept-proxy" bind directive is used, it can
9655 be the address of a client behind another PROXY-protocol compatible component
9656 for all rule sets except "tcp-request connection" which sees the real address.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009657
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +01009658 Example:
9659 # add an HTTP header in requests with the originating address' country
9660 http-request set-header X-Country %[src,map_ip(geoip.lst)]
9661
Willy Tarreau74ca5042013-06-11 23:12:07 +02009662src_bytes_in_rate([<table>]) : integer
9663 Returns the average bytes rate from the incoming connection's source address
9664 in the current proxy's stick-table or in the designated stick-table, measured
9665 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009666 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009667
Willy Tarreau74ca5042013-06-11 23:12:07 +02009668src_bytes_out_rate([<table>]) : integer
9669 Returns the average bytes rate to the incoming connection's source address in
9670 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +02009671 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009672 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009673
Willy Tarreau74ca5042013-06-11 23:12:07 +02009674src_clr_gpc0([<table>]) : integer
9675 Clears the first General Purpose Counter associated to the incoming
9676 connection's source address in the current proxy's stick-table or in the
9677 designated stick-table, and returns its previous value. If the address is not
9678 found, an entry is created and 0 is returned. This is typically used as a
9679 second ACL in an expression in order to mark a connection when a first ACL
9680 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +02009681
9682 # block if 5 consecutive requests continue to come faster than 10 sess
9683 # per second, and reset the counter as soon as the traffic slows down.
9684 acl abuse src_http_req_rate gt 10
9685 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +01009686 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +02009687 tcp-request connection accept if !abuse save
9688 tcp-request connection reject if abuse kill
9689
Willy Tarreau74ca5042013-06-11 23:12:07 +02009690src_conn_cnt([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +02009691 Returns the cumulated number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +02009692 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +02009693 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009694 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009695
Willy Tarreau74ca5042013-06-11 23:12:07 +02009696src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +02009697 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +02009698 current incoming connection's source address in the current proxy's
9699 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009700 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009701
Willy Tarreau74ca5042013-06-11 23:12:07 +02009702src_conn_rate([<table>]) : integer
9703 Returns the average connection rate from the incoming connection's source
9704 address in the current proxy's stick-table or in the designated stick-table,
9705 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009706 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009707
Willy Tarreau74ca5042013-06-11 23:12:07 +02009708src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +02009709 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +02009710 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +02009711 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009712 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009713
Willy Tarreau74ca5042013-06-11 23:12:07 +02009714src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009715 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +02009716 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009717 stick-table or in the designated stick-table. It reports the frequency
9718 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009719 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
9720 that the "gpc0_rate" counter must be stored in the stick-table for a value to
9721 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009722
Willy Tarreau74ca5042013-06-11 23:12:07 +02009723src_http_err_cnt([<table>]) : integer
9724 Returns the cumulated number of HTTP errors from the incoming connection's
9725 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +02009726 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009727 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +02009728 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009729
Willy Tarreau74ca5042013-06-11 23:12:07 +02009730src_http_err_rate([<table>]) : integer
9731 Returns the average rate of HTTP errors from the incoming connection's source
9732 address in the current proxy's stick-table or in the designated stick-table,
9733 measured in amount of errors over the period configured in the table. This
9734 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009735 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009736
Willy Tarreau74ca5042013-06-11 23:12:07 +02009737src_http_req_cnt([<table>]) : integer
9738 Returns the cumulated number of HTTP requests from the incoming connection's
9739 source address in the current proxy's stick-table or in the designated stick-
9740 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009741 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009742
Willy Tarreau74ca5042013-06-11 23:12:07 +02009743src_http_req_rate([<table>]) : integer
9744 Returns the average rate of HTTP requests from the incoming connection's
9745 source address in the current proxy's stick-table or in the designated stick-
9746 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +02009747 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009748 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009749
Willy Tarreau74ca5042013-06-11 23:12:07 +02009750src_inc_gpc0([<table>]) : integer
9751 Increments the first General Purpose Counter associated to the incoming
9752 connection's source address in the current proxy's stick-table or in the
9753 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009754 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +02009755 This is typically used as a second ACL in an expression in order to mark a
9756 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +02009757
9758 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +01009759 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +02009760 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +02009761
Willy Tarreau74ca5042013-06-11 23:12:07 +02009762src_kbytes_in([<table>]) : integer
9763 Returns the amount of data received from the incoming connection's source
9764 address in the current proxy's stick-table or in the designated stick-table,
9765 measured in kilobytes over the period configured in the table. If the address
9766 is not found, zero is returned. The test is currently performed on 32-bit
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009767 integers, which limits values to 4 terabytes. See also
9768 sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009769
Willy Tarreau74ca5042013-06-11 23:12:07 +02009770src_kbytes_out([<table>]) : integer
9771 Returns the amount of data sent to the incoming connection's source address
9772 in the current proxy's stick-table or in the designated stick-table, measured
Willy Tarreauc9705a12010-07-27 20:05:50 +02009773 in kilobytes over the period configured in the table. If the address is not
9774 found, zero is returned. The test is currently performed on 32-bit integers,
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009775 which limits values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +02009776
Willy Tarreau74ca5042013-06-11 23:12:07 +02009777src_port : integer
9778 Returns an integer value corresponding to the TCP source port of the
9779 connection on the client side, which is the port the client connected from.
9780 Usage of this function is very limited as modern protocols do not care much
9781 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +01009782
Willy Tarreau74ca5042013-06-11 23:12:07 +02009783src_sess_cnt([<table>]) : integer
9784 Returns the cumulated number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +02009785 connection's source IPv4 address in the current proxy's stick-table or in the
9786 designated stick-table, that were transformed into sessions, which means that
9787 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009788 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009789
Willy Tarreau74ca5042013-06-11 23:12:07 +02009790src_sess_rate([<table>]) : integer
9791 Returns the average session rate from the incoming connection's source
9792 address in the current proxy's stick-table or in the designated stick-table,
9793 measured in amount of sessions over the period configured in the table. A
9794 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +02009795 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009796
Willy Tarreau74ca5042013-06-11 23:12:07 +02009797src_updt_conn_cnt([<table>]) : integer
9798 Creates or updates the entry associated to the incoming connection's source
9799 address in the current proxy's stick-table or in the designated stick-table.
9800 This table must be configured to store the "conn_cnt" data type, otherwise
9801 the match will be ignored. The current count is incremented by one, and the
9802 expiration timer refreshed. The updated count is returned, so this match
9803 can't return zero. This was used to reject service abusers based on their
9804 source address. Note: it is recommended to use the more complete "track-sc*"
9805 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +02009806
9807 Example :
9808 # This frontend limits incoming SSH connections to 3 per 10 second for
9809 # each source address, and rejects excess connections until a 10 second
9810 # silence is observed. At most 20 addresses are tracked.
9811 listen ssh
9812 bind :22
9813 mode tcp
9814 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +02009815 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +02009816 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +02009817 server local 127.0.0.1:22
9818
Willy Tarreau74ca5042013-06-11 23:12:07 +02009819srv_id : integer
9820 Returns an integer containing the server's id when processing the response.
9821 While it's almost only used with ACLs, it may be used for logging or
9822 debugging.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +02009823
Hervé COMMOWICK35ed8012010-12-15 14:04:51 +01009824
Willy Tarreau74ca5042013-06-11 23:12:07 +020098257.3.3. Fetching samples at Layer 5
9826----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +02009827
Willy Tarreau74ca5042013-06-11 23:12:07 +02009828The layer 5 usually describes just the session layer which in haproxy is
9829closest to the session once all the connection handshakes are finished, but
9830when no content is yet made available. The fetch methods described here are
9831usable as low as the "tcp-request content" rule sets unless they require some
9832future information. Those generally include the results of SSL negociations.
Willy Tarreauc735a072011-03-29 00:57:02 +02009833
Willy Tarreau74ca5042013-06-11 23:12:07 +02009834ssl_c_ca_err : integer
9835 When the incoming connection was made over an SSL/TLS transport layer,
9836 returns the ID of the first error detected during verification of the client
9837 certificate at depth > 0, or 0 if no error was encountered during this
9838 verification process. Please refer to your SSL library's documentation to
9839 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +02009840
Willy Tarreau74ca5042013-06-11 23:12:07 +02009841ssl_c_ca_err_depth : integer
9842 When the incoming connection was made over an SSL/TLS transport layer,
9843 returns the depth in the CA chain of the first error detected during the
9844 verification of the client certificate. If no error is encountered, 0 is
9845 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +01009846
Willy Tarreau74ca5042013-06-11 23:12:07 +02009847ssl_c_err : integer
9848 When the incoming connection was made over an SSL/TLS transport layer,
9849 returns the ID of the first error detected during verification at depth 0, or
9850 0 if no error was encountered during this verification process. Please refer
9851 to your SSL library's documentation to find the exhaustive list of error
9852 codes.
Willy Tarreau62644772008-07-16 18:36:06 +02009853
Willy Tarreau74ca5042013-06-11 23:12:07 +02009854ssl_c_i_dn([<entry>[,<occ>]]) : string
9855 When the incoming connection was made over an SSL/TLS transport layer,
9856 returns the full distinguished name of the issuer of the certificate
9857 presented by the client when no <entry> is specified, or the value of the
9858 first given entry found from the beginning of the DN. If a positive/negative
9859 occurrence number is specified as the optional second argument, it returns
9860 the value of the nth given entry value from the beginning/end of the DN.
9861 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
9862 "ssl_c_i_dn(CN)" retrieves the common name.
Willy Tarreau62644772008-07-16 18:36:06 +02009863
Willy Tarreau74ca5042013-06-11 23:12:07 +02009864 ACL derivatives :
9865 ssl_c_i_dn([<entry>[,<occ>]]) : exact string match
Willy Tarreaub6672b52011-12-12 17:23:41 +01009866
Willy Tarreau74ca5042013-06-11 23:12:07 +02009867ssl_c_key_alg : string
9868 Returns the name of the algorithm used to generate the key of the certificate
9869 presented by the client when the incoming connection was made over an SSL/TLS
9870 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +02009871
Willy Tarreau74ca5042013-06-11 23:12:07 +02009872 ACL derivatives :
9873 ssl_c_key_alg : exact string match
Willy Tarreau2492d5b2009-07-11 00:06:00 +02009874
Willy Tarreau74ca5042013-06-11 23:12:07 +02009875ssl_c_notafter : string
9876 Returns the end date presented by the client as a formatted string
9877 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
9878 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +02009879
Willy Tarreau74ca5042013-06-11 23:12:07 +02009880 ACL derivatives :
9881 ssl_c_notafter : exact string match
Emeric Brunbede3d02009-06-30 17:54:00 +02009882
Willy Tarreau74ca5042013-06-11 23:12:07 +02009883ssl_c_notbefore : string
9884 Returns the start date presented by the client as a formatted string
9885 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
9886 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +01009887
Willy Tarreau74ca5042013-06-11 23:12:07 +02009888 ACL derivatives :
9889 ssl_c_notbefore : exact string match
Willy Tarreaub6672b52011-12-12 17:23:41 +01009890
Willy Tarreau74ca5042013-06-11 23:12:07 +02009891ssl_c_s_dn([<entry>[,<occ>]]) : string
9892 When the incoming connection was made over an SSL/TLS transport layer,
9893 returns the full distinguished name of the subject of the certificate
9894 presented by the client when no <entry> is specified, or the value of the
9895 first given entry found from the beginning of the DN. If a positive/negative
9896 occurrence number is specified as the optional second argument, it returns
9897 the value of the nth given entry value from the beginning/end of the DN.
9898 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
9899 "ssl_c_s_dn(CN)" retrieves the common name.
Willy Tarreaub6672b52011-12-12 17:23:41 +01009900
Willy Tarreau74ca5042013-06-11 23:12:07 +02009901 ACL derivatives :
9902 ssl_c_s_dn([<entry>[,<occ>]]) : exact string match
Willy Tarreau7875d092012-09-10 08:20:03 +02009903
Willy Tarreau74ca5042013-06-11 23:12:07 +02009904ssl_c_serial : binary
9905 Returns the serial of the certificate presented by the client when the
9906 incoming connection was made over an SSL/TLS transport layer. When used for
9907 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +02009908
Willy Tarreau74ca5042013-06-11 23:12:07 +02009909 ACL derivatives :
9910 ssl_c_serial : hex block match
Emeric Brun2525b6b2012-10-18 15:59:43 +02009911
Willy Tarreau74ca5042013-06-11 23:12:07 +02009912ssl_c_sha1 : binary
9913 Returns the SHA-1 fingerprint of the certificate presented by the client when
9914 the incoming connection was made over an SSL/TLS transport layer. This can be
9915 used to stick a client to a server, or to pass this information to a server.
Emeric Brun2525b6b2012-10-18 15:59:43 +02009916
Willy Tarreau74ca5042013-06-11 23:12:07 +02009917ssl_c_sig_alg : string
9918 Returns the name of the algorithm used to sign the certificate presented by
9919 the client when the incoming connection was made over an SSL/TLS transport
9920 layer.
Emeric Brun87855892012-10-17 17:39:35 +02009921
Willy Tarreau74ca5042013-06-11 23:12:07 +02009922 ACL derivatives :
9923 ssl_c_sig_alg : exact string match
9924
9925ssl_c_used : boolean
9926 Returns true if current SSL session uses a client certificate even if current
9927 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +02009928
Willy Tarreau74ca5042013-06-11 23:12:07 +02009929ssl_c_verify : integer
9930 Returns the verify result error ID when the incoming connection was made over
9931 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
9932 refer to your SSL library's documentation for an exhaustive list of error
9933 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +02009934
Willy Tarreau74ca5042013-06-11 23:12:07 +02009935ssl_c_version : integer
9936 Returns the version of the certificate presented by the client when the
9937 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +02009938
Willy Tarreau74ca5042013-06-11 23:12:07 +02009939ssl_f_i_dn([<entry>[,<occ>]]) : string
9940 When the incoming connection was made over an SSL/TLS transport layer,
9941 returns the full distinguished name of the issuer of the certificate
9942 presented by the frontend when no <entry> is specified, or the value of the
9943 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +02009944 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +02009945 the value of the nth given entry value from the beginning/end of the DN.
9946 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
9947 "ssl_f_i_dn(CN)" retrieves the common name.
Emeric Brun87855892012-10-17 17:39:35 +02009948
Willy Tarreau74ca5042013-06-11 23:12:07 +02009949 ACL derivatives :
9950 ssl_f_i_dn([<entry>[,<occ>]]) : exact string match
Willy Tarreau8d598402012-10-22 17:58:39 +02009951
Willy Tarreau74ca5042013-06-11 23:12:07 +02009952ssl_f_key_alg : string
9953 Returns the name of the algorithm used to generate the key of the certificate
9954 presented by the frontend when the incoming connection was made over an
9955 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +02009956
Willy Tarreau74ca5042013-06-11 23:12:07 +02009957 ACL derivatives :
9958 ssl_f_key_alg : exact string match
Emeric Brun9143d372012-12-20 15:44:16 +01009959
Willy Tarreau74ca5042013-06-11 23:12:07 +02009960ssl_f_notafter : string
9961 Returns the end date presented by the frontend as a formatted string
9962 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
9963 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +02009964
Willy Tarreau74ca5042013-06-11 23:12:07 +02009965 ACL derivatives :
9966 ssl_f_notafter : exact string match
Emeric Bruna7359fd2012-10-17 15:03:11 +02009967
Willy Tarreau74ca5042013-06-11 23:12:07 +02009968ssl_f_notbefore : string
9969 Returns the start date presented by the frontend as a formatted string
9970 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
9971 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +02009972
Willy Tarreau74ca5042013-06-11 23:12:07 +02009973 ACL derivatives :
9974 ssl_f_notbefore : exact string match
Emeric Brun7f56e742012-10-19 18:15:40 +02009975
Willy Tarreau74ca5042013-06-11 23:12:07 +02009976ssl_f_s_dn([<entry>[,<occ>]]) : string
9977 When the incoming connection was made over an SSL/TLS transport layer,
9978 returns the full distinguished name of the subject of the certificate
9979 presented by the frontend when no <entry> is specified, or the value of the
9980 first given entry found from the beginning of the DN. If a positive/negative
9981 occurrence number is specified as the optional second argument, it returns
9982 the value of the nth given entry value from the beginning/end of the DN.
9983 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
9984 "ssl_f_s_dn(CN)" retrieves the common name.
Emeric Brunce5ad802012-10-22 14:11:22 +02009985
Willy Tarreau74ca5042013-06-11 23:12:07 +02009986 ACL derivatives :
9987 ssl_f_s_dn([<entry>[,<occ>]]) : exact string match
Emeric Brunce5ad802012-10-22 14:11:22 +02009988
Willy Tarreau74ca5042013-06-11 23:12:07 +02009989ssl_f_serial : binary
9990 Returns the serial of the certificate presented by the frontend when the
9991 incoming connection was made over an SSL/TLS transport layer. When used for
9992 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +02009993
Willy Tarreau74ca5042013-06-11 23:12:07 +02009994 ACL derivatives :
9995 ssl_f_serial : hex block match
Willy Tarreau8d598402012-10-22 17:58:39 +02009996
Willy Tarreau74ca5042013-06-11 23:12:07 +02009997ssl_f_sig_alg : string
9998 Returns the name of the algorithm used to sign the certificate presented by
9999 the frontend when the incoming connection was made over an SSL/TLS transport
10000 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020010001
Willy Tarreau74ca5042013-06-11 23:12:07 +020010002 ACL derivatives :
10003 ssl_f_sig_alg : exact string match
Emeric Bruna7359fd2012-10-17 15:03:11 +020010004
Willy Tarreau74ca5042013-06-11 23:12:07 +020010005ssl_f_version : integer
10006 Returns the version of the certificate presented by the frontend when the
10007 incoming connection was made over an SSL/TLS transport layer.
10008
10009ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020010010 Returns true when the front connection was made via an SSL/TLS transport
10011 layer and is locally deciphered. This means it has matched a socket declared
10012 with a "bind" line having the "ssl" option.
10013
Willy Tarreau74ca5042013-06-11 23:12:07 +020010014 Example :
10015 # This passes "X-Proto: https" to servers when client connects over SSL
10016 listen http-https
10017 bind :80
10018 bind :443 ssl crt /etc/haproxy.pem
10019 http-request add-header X-Proto https if { ssl_fc }
10020
10021ssl_fc_alg_keysize : integer
10022 Returns the symmetric cipher key size supported in bits when the incoming
10023 connection was made over an SSL/TLS transport layer.
10024
10025ssl_fc_alpn : string
10026 This extracts the Application Layer Protocol Negociation field from an
10027 incoming connection made via a TLS transport layer and locally deciphered by
10028 haproxy. The result is a string containing the protocol name advertised by
10029 the client. The SSL library must have been built with support for TLS
10030 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
10031 not advertised unless the "alpn" keyword on the "bind" line specifies a
10032 protocol list. Also, nothing forces the client to pick a protocol from this
10033 list, any other one may be requested. The TLS ALPN extension is meant to
10034 replace the TLS NPN extension. See also "ssl_fc_npn".
10035
10036 ACL derivatives :
10037 ssl_fc_alpn : exact string match
Emeric Brun589fcad2012-10-16 14:13:26 +020010038
Willy Tarreau74ca5042013-06-11 23:12:07 +020010039ssl_fc_cipher : string
10040 Returns the name of the used cipher when the incoming connection was made
10041 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +020010042
Willy Tarreau74ca5042013-06-11 23:12:07 +020010043 ACL derivatives :
10044 ssl_fc_cipher : exact string match
Emeric Brun589fcad2012-10-16 14:13:26 +020010045
Willy Tarreau74ca5042013-06-11 23:12:07 +020010046ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020010047 Returns true if a client certificate is present in an incoming connection over
10048 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +010010049 Note: on SSL session resumption with Session ID or TLS ticket, client
10050 certificate is not present in the current connection but may be retrieved
10051 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
10052 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +020010053
Willy Tarreau74ca5042013-06-11 23:12:07 +020010054ssl_fc_has_sni : boolean
10055 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020010056 in an incoming connection was made over an SSL/TLS transport layer. Returns
10057 true when the incoming connection presents a TLS SNI field. This requires
10058 that the SSL library is build with support for TLS extensions enabled (check
10059 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +020010060
Willy Tarreau74ca5042013-06-11 23:12:07 +020010061ssl_fc_npn : string
10062 This extracts the Next Protocol Negociation field from an incoming connection
10063 made via a TLS transport layer and locally deciphered by haproxy. The result
10064 is a string containing the protocol name advertised by the client. The SSL
10065 library must have been built with support for TLS extensions enabled (check
10066 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
10067 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
10068 forces the client to pick a protocol from this list, any other one may be
10069 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +020010070
Willy Tarreau74ca5042013-06-11 23:12:07 +020010071 ACL derivatives :
10072 ssl_fc_npn : exact string match
Emeric Brun589fcad2012-10-16 14:13:26 +020010073
Willy Tarreau74ca5042013-06-11 23:12:07 +020010074ssl_fc_protocol : string
10075 Returns the name of the used protocol when the incoming connection was made
10076 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +020010077
Willy Tarreau74ca5042013-06-11 23:12:07 +020010078 ACL derivatives :
10079 ssl_fc_protocol : exact string match
10080
10081ssl_fc_session_id : binary
10082 Returns the SSL ID of the front connection when the incoming connection was
10083 made over an SSL/TLS transport layer. It is useful to stick a given client to
10084 a server. It is important to note that some browsers refresh their session ID
10085 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +020010086
Willy Tarreau74ca5042013-06-11 23:12:07 +020010087ssl_fc_sni : string
10088 This extracts the Server Name Indication TLS extension (SNI) field from an
10089 incoming connection made via an SSL/TLS transport layer and locally
10090 deciphered by haproxy. The result (when present) typically is a string
10091 matching the HTTPS host name (253 chars or less). The SSL library must have
10092 been built with support for TLS extensions enabled (check haproxy -vv).
10093
10094 This fetch is different from "req_ssl_sni" above in that it applies to the
10095 connection being deciphered by haproxy and not to SSL contents being blindly
10096 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +020010097 requires that the SSL library is build with support for TLS extensions
10098 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +020010099
Willy Tarreau74ca5042013-06-11 23:12:07 +020010100 ACL derivatives :
10101 ssl_fc_sni : exact string match
10102 ssl_fc_sni_end : suffix match
10103 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +020010104
Willy Tarreau74ca5042013-06-11 23:12:07 +020010105ssl_fc_use_keysize : integer
10106 Returns the symmetric cipher key size used in bits when the incoming
10107 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020010108
Willy Tarreaub6fb4202008-07-20 11:18:28 +020010109
Willy Tarreau74ca5042013-06-11 23:12:07 +0200101107.3.4. Fetching samples from buffer contents (Layer 6)
10111------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +020010112
Willy Tarreau74ca5042013-06-11 23:12:07 +020010113Fetching samples from buffer contents is a bit different from the previous
10114sample fetches above because the sampled data are ephemeral. These data can
10115only be used when they're available and will be lost when they're forwarded.
10116For this reason, samples fetched from buffer contents during a request cannot
10117be used in a response for example. Even while the data are being fetched, they
10118can change. Sometimes it is necessary to set some delays or combine multiple
10119sample fetch methods to ensure that the expected data are complete and usable,
10120for example through TCP request content inspection. Please see the "tcp-request
10121content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +020010122
Willy Tarreau74ca5042013-06-11 23:12:07 +020010123payload(<offset>,<length>) : binary (deprecated)
10124 This is an alias for "req.payload" when used in the context of a request (eg:
10125 "stick on", "stick match"), and for "res.payload" when used in the context of
10126 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010010127
Willy Tarreau74ca5042013-06-11 23:12:07 +020010128payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
10129 This is an alias for "req.payload_lv" when used in the context of a request
10130 (eg: "stick on", "stick match"), and for "res.payload_lv" when used in the
10131 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010010132
Willy Tarreau74ca5042013-06-11 23:12:07 +020010133req.len : integer
10134req_len : integer (deprecated)
10135 Returns an integer value corresponding to the number of bytes present in the
10136 request buffer. This is mostly used in ACL. It is important to understand
10137 that this test does not return false as long as the buffer is changing. This
10138 means that a check with equality to zero will almost always immediately match
10139 at the beginning of the session, while a test for more data will wait for
10140 that data to come in and return false only when haproxy is certain that no
10141 more data will come in. This test was designed to be used with TCP request
10142 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020010143
Willy Tarreau74ca5042013-06-11 23:12:07 +020010144req.payload(<offset>,<length>) : binary
10145 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020010146 in the request buffer. As a special case, if the <length> argument is zero,
10147 the the whole buffer from <offset> to the end is extracted. This can be used
10148 with ACLs in order to check for the presence of some content in a buffer at
10149 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020010150
Willy Tarreau74ca5042013-06-11 23:12:07 +020010151 ACL alternatives :
10152 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020010153
Willy Tarreau74ca5042013-06-11 23:12:07 +020010154req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
10155 This extracts a binary block whose size is specified at <offset1> for <length>
10156 bytes, and which starts at <offset2> if specified or just after the length in
10157 the request buffer. The <offset2> parameter also supports relative offsets if
10158 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020010159
Willy Tarreau74ca5042013-06-11 23:12:07 +020010160 ACL alternatives :
10161 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020010162
Willy Tarreau74ca5042013-06-11 23:12:07 +020010163 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020010164
Willy Tarreau74ca5042013-06-11 23:12:07 +020010165req.proto_http : boolean
10166req_proto_http : boolean (deprecated)
10167 Returns true when data in the request buffer look like HTTP and correctly
10168 parses as such. It is the same parser as the common HTTP request parser which
10169 is used so there should be no surprises. The test does not match until the
10170 request is complete, failed or timed out. This test may be used to report the
10171 protocol in TCP logs, but the biggest use is to block TCP request analysis
10172 until a complete HTTP request is present in the buffer, for example to track
10173 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020010174
Willy Tarreau74ca5042013-06-11 23:12:07 +020010175 Example:
10176 # track request counts per "base" (concatenation of Host+URL)
10177 tcp-request inspect-delay 10s
10178 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010179 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020010180
Willy Tarreau74ca5042013-06-11 23:12:07 +020010181req.rdp_cookie([<name>]) : string
10182rdp_cookie([<name>]) : string (deprecated)
10183 When the request buffer looks like the RDP protocol, extracts the RDP cookie
10184 <name>, or any cookie if unspecified. The parser only checks for the first
10185 cookie, as illustrated in the RDP protocol specification. The cookie name is
10186 case insensitive. Generally the "MSTS" cookie name will be used, as it can
10187 contain the user name of the client connecting to the server if properly
10188 configured on the client. The "MSTSHASH" cookie is often used as well for
10189 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020010190
Willy Tarreau74ca5042013-06-11 23:12:07 +020010191 This differs from "balance rdp-cookie" in that any balancing algorithm may be
10192 used and thus the distribution of clients to backend servers is not linked to
10193 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
10194 such as "balance roundrobin" or "balance leastconn" will lead to a more even
10195 distribution of clients to backend servers than the hash used by "balance
10196 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020010197
Willy Tarreau74ca5042013-06-11 23:12:07 +020010198 ACL derivatives :
10199 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020010200
Willy Tarreau74ca5042013-06-11 23:12:07 +020010201 Example :
10202 listen tse-farm
10203 bind 0.0.0.0:3389
10204 # wait up to 5s for an RDP cookie in the request
10205 tcp-request inspect-delay 5s
10206 tcp-request content accept if RDP_COOKIE
10207 # apply RDP cookie persistence
10208 persist rdp-cookie
10209 # Persist based on the mstshash cookie
10210 # This is only useful makes sense if
10211 # balance rdp-cookie is not used
10212 stick-table type string size 204800
10213 stick on req.rdp_cookie(mstshash)
10214 server srv1 1.1.1.1:3389
10215 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +020010216
Willy Tarreau74ca5042013-06-11 23:12:07 +020010217 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
10218 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020010219
Willy Tarreau74ca5042013-06-11 23:12:07 +020010220req.rdp_cookie_cnt([name]) : integer
10221rdp_cookie_cnt([name]) : integer (deprecated)
10222 Tries to parse the request buffer as RDP protocol, then returns an integer
10223 corresponding to the number of RDP cookies found. If an optional cookie name
10224 is passed, only cookies matching this name are considered. This is mostly
10225 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020010226
Willy Tarreau74ca5042013-06-11 23:12:07 +020010227 ACL derivatives :
10228 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020010229
Willy Tarreau74ca5042013-06-11 23:12:07 +020010230req.ssl_hello_type : integer
10231req_ssl_hello_type : integer (deprecated)
10232 Returns an integer value containing the type of the SSL hello message found
10233 in the request buffer if the buffer contains data that parse as a complete
10234 SSL (v3 or superior) client hello message. Note that this only applies to raw
10235 contents found in the request buffer and not to contents deciphered via an
10236 SSL data layer, so this will not work with "bind" lines having the "ssl"
10237 option. This is mostly used in ACL to detect presence of an SSL hello message
10238 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020010239
Willy Tarreau74ca5042013-06-11 23:12:07 +020010240req.ssl_sni : string
10241req_ssl_sni : string (deprecated)
10242 Returns a string containing the value of the Server Name TLS extension sent
10243 by a client in a TLS stream passing through the request buffer if the buffer
10244 contains data that parse as a complete SSL (v3 or superior) client hello
10245 message. Note that this only applies to raw contents found in the request
10246 buffer and not to contents deciphered via an SSL data layer, so this will not
10247 work with "bind" lines having the "ssl" option. SNI normally contains the
10248 name of the host the client tries to connect to (for recent browsers). SNI is
10249 useful for allowing or denying access to certain hosts when SSL/TLS is used
10250 by the client. This test was designed to be used with TCP request content
10251 inspection. If content switching is needed, it is recommended to first wait
10252 for a complete client hello (type 1), like in the example below. See also
10253 "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020010254
Willy Tarreau74ca5042013-06-11 23:12:07 +020010255 ACL derivatives :
10256 req_ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020010257
Willy Tarreau74ca5042013-06-11 23:12:07 +020010258 Examples :
10259 # Wait for a client hello for at most 5 seconds
10260 tcp-request inspect-delay 5s
10261 tcp-request content accept if { req_ssl_hello_type 1 }
10262 use_backend bk_allow if { req_ssl_sni -f allowed_sites }
10263 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +020010264
Willy Tarreau74ca5042013-06-11 23:12:07 +020010265res.ssl_hello_type : integer
10266rep_ssl_hello_type : integer (deprecated)
10267 Returns an integer value containing the type of the SSL hello message found
10268 in the response buffer if the buffer contains data that parses as a complete
10269 SSL (v3 or superior) hello message. Note that this only applies to raw
10270 contents found in the response buffer and not to contents deciphered via an
10271 SSL data layer, so this will not work with "server" lines having the "ssl"
10272 option. This is mostly used in ACL to detect presence of an SSL hello message
10273 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau51539362012-05-08 12:46:28 +020010274
Willy Tarreau74ca5042013-06-11 23:12:07 +020010275req.ssl_ver : integer
10276req_ssl_ver : integer (deprecated)
10277 Returns an integer value containing the version of the SSL/TLS protocol of a
10278 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
10279 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
10280 composed of the major version multiplied by 65536, added to the minor
10281 version. Note that this only applies to raw contents found in the request
10282 buffer and not to contents deciphered via an SSL data layer, so this will not
10283 work with "bind" lines having the "ssl" option. The ACL version of the test
10284 matches against a decimal notation in the form MAJOR.MINOR (eg: 3.1). This
10285 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010286
Willy Tarreau74ca5042013-06-11 23:12:07 +020010287 ACL derivatives :
10288 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +010010289
Willy Tarreau47e8eba2013-09-11 23:28:46 +020010290res.len : integer
10291 Returns an integer value corresponding to the number of bytes present in the
10292 response buffer. This is mostly used in ACL. It is important to understand
10293 that this test does not return false as long as the buffer is changing. This
10294 means that a check with equality to zero will almost always immediately match
10295 at the beginning of the session, while a test for more data will wait for
10296 that data to come in and return false only when haproxy is certain that no
10297 more data will come in. This test was designed to be used with TCP response
10298 content inspection.
10299
Willy Tarreau74ca5042013-06-11 23:12:07 +020010300res.payload(<offset>,<length>) : binary
10301 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020010302 in the response buffer. As a special case, if the <length> argument is zero,
10303 the the whole buffer from <offset> to the end is extracted. This can be used
10304 with ACLs in order to check for the presence of some content in a buffer at
10305 any location.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010306
Willy Tarreau74ca5042013-06-11 23:12:07 +020010307res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
10308 This extracts a binary block whose size is specified at <offset1> for <length>
10309 bytes, and which starts at <offset2> if specified or just after the length in
10310 the response buffer. The <offset2> parameter also supports relative offsets
10311 if prepended with a '+' or '-' sign.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010312
Willy Tarreau74ca5042013-06-11 23:12:07 +020010313 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010314
Willy Tarreau74ca5042013-06-11 23:12:07 +020010315wait_end : boolean
10316 This fetch either returns true when the inspection period is over, or does
10317 not fetch. It is only used in ACLs, in conjunction with content analysis to
10318 avoid returning a wrong verdict early. It may also be used to delay some
10319 actions, such as a delayed reject for some special addresses. Since it either
10320 stops the rules evaluation or immediately returns true, it is recommended to
10321 use this acl as the last one in a rule. Please note that the default ACL
10322 "WAIT_END" is always usable without prior declaration. This test was designed
10323 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010324
Willy Tarreau74ca5042013-06-11 23:12:07 +020010325 Examples :
10326 # delay every incoming request by 2 seconds
10327 tcp-request inspect-delay 2s
10328 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +010010329
Willy Tarreau74ca5042013-06-11 23:12:07 +020010330 # don't immediately tell bad guys they are rejected
10331 tcp-request inspect-delay 10s
10332 acl goodguys src 10.0.0.0/24
10333 acl badguys src 10.0.1.0/24
10334 tcp-request content accept if goodguys
10335 tcp-request content reject if badguys WAIT_END
10336 tcp-request content reject
10337
10338
103397.3.5. Fetching HTTP samples (Layer 7)
10340--------------------------------------
10341
10342It is possible to fetch samples from HTTP contents, requests and responses.
10343This application layer is also called layer 7. It is only possible to fetch the
10344data in this section when a full HTTP request or response has been parsed from
10345its respective request or response buffer. This is always the case with all
10346HTTP specific rules and for sections running with "mode http". When using TCP
10347content inspection, it may be necessary to support an inspection delay in order
10348to let the request or response come in first. These fetches may require a bit
10349more CPU resources than the layer 4 ones, but not much since the request and
10350response are indexed.
10351
10352base : string
10353 This returns the concatenation of the first Host header and the path part of
10354 the request, which starts at the first slash and ends before the question
10355 mark. It can be useful in virtual hosted environments to detect URL abuses as
10356 well as to improve shared caches efficiency. Using this with a limited size
10357 stick table also allows one to collect statistics about most commonly
10358 requested objects by host/path. With ACLs it can allow simple content
10359 switching rules involving the host and the path at the same time, such as
10360 "www.example.com/favicon.ico". See also "path" and "uri".
10361
10362 ACL derivatives :
10363 base : exact string match
10364 base_beg : prefix match
10365 base_dir : subdir match
10366 base_dom : domain match
10367 base_end : suffix match
10368 base_len : length match
10369 base_reg : regex match
10370 base_sub : substring match
10371
10372base32 : integer
10373 This returns a 32-bit hash of the value returned by the "base" fetch method
10374 above. This is useful to track per-URL activity on high traffic sites without
10375 having to store all URLs. Instead a shorter hash is stored, saving a lot of
10376 memory. The output type is an unsigned integer.
10377
10378base32+src : binary
10379 This returns the concatenation of the base32 fetch above and the src fetch
10380 below. The resulting type is of type binary, with a size of 8 or 20 bytes
10381 depending on the source address family. This can be used to track per-IP,
10382 per-URL counters.
10383
10384req.cook([<name>]) : string
10385cook([<name>]) : string (deprecated)
10386 This extracts the last occurrence of the cookie name <name> on a "Cookie"
10387 header line from the request, and returns its value as string. If no name is
10388 specified, the first cookie value is returned. When used with ACLs, all
10389 matching cookies are evaluated. Spaces around the name and the value are
10390 ignored as requested by the Cookie header specification (RFC6265). The cookie
10391 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
10392 well return an empty value if it is present. Use the "found" match to detect
10393 presence. Use the res.cook() variant for response cookies sent by the server.
10394
10395 ACL derivatives :
10396 cook([<name>]) : exact string match
10397 cook_beg([<name>]) : prefix match
10398 cook_dir([<name>]) : subdir match
10399 cook_dom([<name>]) : domain match
10400 cook_end([<name>]) : suffix match
10401 cook_len([<name>]) : length match
10402 cook_reg([<name>]) : regex match
10403 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +010010404
Willy Tarreau74ca5042013-06-11 23:12:07 +020010405req.cook_cnt([<name>]) : integer
10406cook_cnt([<name>]) : integer (deprecated)
10407 Returns an integer value representing the number of occurrences of the cookie
10408 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010409
Willy Tarreau74ca5042013-06-11 23:12:07 +020010410req.cook_val([<name>]) : integer
10411cook_val([<name>]) : integer (deprecated)
10412 This extracts the last occurrence of the cookie name <name> on a "Cookie"
10413 header line from the request, and converts its value to an integer which is
10414 returned. If no name is specified, the first cookie value is returned. When
10415 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +020010416
Willy Tarreau74ca5042013-06-11 23:12:07 +020010417cookie([<name>]) : string (deprecated)
10418 This extracts the last occurrence of the cookie name <name> on a "Cookie"
10419 header line from the request, or a "Set-Cookie" header from the response, and
10420 returns its value as a string. A typical use is to get multiple clients
10421 sharing a same profile use the same server. This can be similar to what
10422 "appsession" does with the "request-learn" statement, but with support for
10423 multi-peer synchronization and state keeping across restarts. If no name is
10424 specified, the first cookie value is returned. This fetch should not be used
10425 anymore and should be replaced by req.cook() or res.cook() instead as it
10426 ambiguously uses the direction based on the context where it is used.
10427 See also : "appsession".
Willy Tarreaud63335a2010-02-26 12:56:52 +010010428
Willy Tarreau74ca5042013-06-11 23:12:07 +020010429hdr([<name>[,<occ>]]) : string
10430 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
10431 used on responses. Please refer to these respective fetches for more details.
10432 In case of doubt about the fetch direction, please use the explicit ones.
10433 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
10434 unambiguouslly apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010435
Willy Tarreau74ca5042013-06-11 23:12:07 +020010436req.fhdr(<name>[,<occ>]) : string
10437 This extracts the last occurrence of header <name> in an HTTP request. When
10438 used from an ACL, all occurrences are iterated over until a match is found.
10439 Optionally, a specific occurrence might be specified as a position number.
10440 Positive values indicate a position from the first occurrence, with 1 being
10441 the first one. Negative values indicate positions relative to the last one,
10442 with -1 being the last one. It differs from req.hdr() in that any commas
10443 present in the value are returned and are not used as delimiters. This is
10444 sometimes useful with headers such as User-Agent.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010445
Willy Tarreau74ca5042013-06-11 23:12:07 +020010446req.fhdr_cnt([<name>]) : integer
10447 Returns an integer value representing the number of occurrences of request
10448 header field name <name>, or the total number of header fields if <name> is
10449 not specified. Contrary to its req.hdr_cnt() cousin, this function returns
10450 the number of full line headers and does not stop on commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010451
Willy Tarreau74ca5042013-06-11 23:12:07 +020010452req.hdr([<name>[,<occ>]]) : string
10453 This extracts the last occurrence of header <name> in an HTTP request. When
10454 used from an ACL, all occurrences are iterated over until a match is found.
10455 Optionally, a specific occurrence might be specified as a position number.
10456 Positive values indicate a position from the first occurrence, with 1 being
10457 the first one. Negative values indicate positions relative to the last one,
10458 with -1 being the last one. A typical use is with the X-Forwarded-For header
10459 once converted to IP, associated with an IP stick-table. The function
10460 considers any comma as a delimiter for distinct values. If full-line headers
10461 are desired instead, use req.fhdr(). Please carefully check RFC2616 to know
10462 how certain headers are supposed to be parsed. Also, some of them are case
10463 insensitive (eg: Connection).
Willy Tarreaud63335a2010-02-26 12:56:52 +010010464
Willy Tarreau74ca5042013-06-11 23:12:07 +020010465 ACL derivatives :
10466 hdr([<name>[,<occ>]]) : exact string match
10467 hdr_beg([<name>[,<occ>]]) : prefix match
10468 hdr_dir([<name>[,<occ>]]) : subdir match
10469 hdr_dom([<name>[,<occ>]]) : domain match
10470 hdr_end([<name>[,<occ>]]) : suffix match
10471 hdr_len([<name>[,<occ>]]) : length match
10472 hdr_reg([<name>[,<occ>]]) : regex match
10473 hdr_sub([<name>[,<occ>]]) : substring match
10474
10475req.hdr_cnt([<name>]) : integer
10476hdr_cnt([<header>]) : integer (deprecated)
10477 Returns an integer value representing the number of occurrences of request
10478 header field name <name>, or the total number of header field values if
10479 <name> is not specified. It is important to remember that one header line may
10480 count as several headers if it has several values. The function considers any
10481 comma as a delimiter for distinct values. If full-line headers are desired
10482 instead, req.fhdr_cnt() should be used instead. With ACLs, it can be used to
10483 detect presence, absence or abuse of a specific header, as well as to block
10484 request smuggling attacks by rejecting requests which contain more than one
10485 of certain headers. See "req.hdr" for more information on header matching.
10486
10487req.hdr_ip([<name>[,<occ>]]) : ip
10488hdr_ip([<name>[,<occ>]]) : ip (deprecated)
10489 This extracts the last occurrence of header <name> in an HTTP request,
10490 converts it to an IPv4 or IPv6 address and returns this address. When used
10491 with ACLs, all occurrences are checked, and if <name> is omitted, every value
10492 of every header is checked. Optionally, a specific occurrence might be
10493 specified as a position number. Positive values indicate a position from the
10494 first occurrence, with 1 being the first one. Negative values indicate
10495 positions relative to the last one, with -1 being the last one. A typical use
10496 is with the X-Forwarded-For and X-Client-IP headers.
10497
10498req.hdr_val([<name>[,<occ>]]) : integer
10499hdr_val([<name>[,<occ>]]) : integer (deprecated)
10500 This extracts the last occurrence of header <name> in an HTTP request, and
10501 converts it to an integer value. When used with ACLs, all occurrences are
10502 checked, and if <name> is omitted, every value of every header is checked.
10503 Optionally, a specific occurrence might be specified as a position number.
10504 Positive values indicate a position from the first occurrence, with 1 being
10505 the first one. Negative values indicate positions relative to the last one,
10506 with -1 being the last one. A typical use is with the X-Forwarded-For header.
10507
10508http_auth(<userlist>) : boolean
10509 Returns a boolean indicating whether the authentication data received from
10510 the client match a username & password stored in the specified userlist. This
10511 fetch function is not really useful outside of ACLs. Currently only http
10512 basic auth is supported.
10513
10514http_auth_group(<userlist>) : group
10515 Returns a boolean indicating whether the authentication data received from
10516 the client match a username & password stored in the specified userlist, and
10517 whether that username belongs to one of the groups supplied in ACL patterns.
10518 This fetch function is not really useful outside of ACLs. Currently only http
10519 basic auth is supported.
10520
10521 ACL derivatives :
10522 http_auth_group(<userlist>) : user group match
10523
10524http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020010525 Returns true when the request being processed is the first one of the
10526 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020010527 from some requests when a request is not the first one, or to help grouping
10528 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020010529
Willy Tarreau74ca5042013-06-11 23:12:07 +020010530method : integer + string
10531 Returns an integer value corresponding to the method in the HTTP request. For
10532 example, "GET" equals 1 (check sources to establish the matching). Value 9
10533 means "other method" and may be converted to a string extracted from the
10534 stream. This should not be used directly as a sample, this is only meant to
10535 be used from ACLs, which transparently convert methods from patterns to these
10536 integer + string values. Some predefined ACL already check for most common
10537 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010538
Willy Tarreau74ca5042013-06-11 23:12:07 +020010539 ACL derivatives :
10540 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020010541
Willy Tarreau74ca5042013-06-11 23:12:07 +020010542 Example :
10543 # only accept GET and HEAD requests
10544 acl valid_method method GET HEAD
10545 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020010546
Willy Tarreau74ca5042013-06-11 23:12:07 +020010547path : string
10548 This extracts the request's URL path, which starts at the first slash and
10549 ends before the question mark (without the host part). A typical use is with
10550 prefetch-capable caches, and with portals which need to aggregate multiple
10551 information from databases and keep them in caches. Note that with outgoing
10552 caches, it would be wiser to use "url" instead. With ACLs, it's typically
10553 used to match exact file names (eg: "/login.php"), or directory parts using
10554 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010555
Willy Tarreau74ca5042013-06-11 23:12:07 +020010556 ACL derivatives :
10557 path : exact string match
10558 path_beg : prefix match
10559 path_dir : subdir match
10560 path_dom : domain match
10561 path_end : suffix match
10562 path_len : length match
10563 path_reg : regex match
10564 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020010565
Willy Tarreau74ca5042013-06-11 23:12:07 +020010566req.ver : string
10567req_ver : string (deprecated)
10568 Returns the version string from the HTTP request, for example "1.1". This can
10569 be useful for logs, but is mostly there for ACL. Some predefined ACL already
10570 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010571
Willy Tarreau74ca5042013-06-11 23:12:07 +020010572 ACL derivatives :
10573 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020010574
Willy Tarreau74ca5042013-06-11 23:12:07 +020010575res.comp : boolean
10576 Returns the boolean "true" value if the response has been compressed by
10577 HAProxy, otherwise returns boolean "false". This may be used to add
10578 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010579
Willy Tarreau74ca5042013-06-11 23:12:07 +020010580res.comp_algo : string
10581 Returns a string containing the name of the algorithm used if the response
10582 was compressed by HAProxy, for example : "deflate". This may be used to add
10583 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010584
Willy Tarreau74ca5042013-06-11 23:12:07 +020010585res.cook([<name>]) : string
10586scook([<name>]) : string (deprecated)
10587 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
10588 header line from the response, and returns its value as string. If no name is
10589 specified, the first cookie value is returned.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020010590
Willy Tarreau74ca5042013-06-11 23:12:07 +020010591 ACL derivatives :
10592 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020010593
Willy Tarreau74ca5042013-06-11 23:12:07 +020010594res.cook_cnt([<name>]) : integer
10595scook_cnt([<name>]) : integer (deprecated)
10596 Returns an integer value representing the number of occurrences of the cookie
10597 <name> in the response, or all cookies if <name> is not specified. This is
10598 mostly useful when combined with ACLs to detect suspicious responses.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010599
Willy Tarreau74ca5042013-06-11 23:12:07 +020010600res.cook_val([<name>]) : integer
10601scook_val([<name>]) : integer (deprecated)
10602 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
10603 header line from the response, and converts its value to an integer which is
10604 returned. If no name is specified, the first cookie value is returned.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010605
Willy Tarreau74ca5042013-06-11 23:12:07 +020010606res.fhdr([<name>[,<occ>]]) : string
10607 This extracts the last occurrence of header <name> in an HTTP response, or of
10608 the last header if no <name> is specified. Optionally, a specific occurrence
10609 might be specified as a position number. Positive values indicate a position
10610 from the first occurrence, with 1 being the first one. Negative values
10611 indicate positions relative to the last one, with -1 being the last one. It
10612 differs from res.hdr() in that any commas present in the value are returned
10613 and are not used as delimiters. If this is not desired, the res.hdr() fetch
10614 should be used instead. This is sometimes useful with headers such as Date or
10615 Expires.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010616
Willy Tarreau74ca5042013-06-11 23:12:07 +020010617res.fhdr_cnt([<name>]) : integer
10618 Returns an integer value representing the number of occurrences of response
10619 header field name <name>, or the total number of header fields if <name> is
10620 not specified. Contrary to its res.hdr_cnt() cousin, this function returns
10621 the number of full line headers and does not stop on commas. If this is not
10622 desired, the res.hdr_cnt() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010623
Willy Tarreau74ca5042013-06-11 23:12:07 +020010624res.hdr([<name>[,<occ>]]) : string
10625shdr([<name>[,<occ>]]) : string (deprecated)
10626 This extracts the last occurrence of header <name> in an HTTP response, or of
10627 the last header if no <name> is specified. Optionally, a specific occurrence
10628 might be specified as a position number. Positive values indicate a position
10629 from the first occurrence, with 1 being the first one. Negative values
10630 indicate positions relative to the last one, with -1 being the last one. This
10631 can be useful to learn some data into a stick-table. The function considers
10632 any comma as a delimiter for distinct values. If this is not desired, the
10633 res.fhdr() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010634
Willy Tarreau74ca5042013-06-11 23:12:07 +020010635 ACL derivatives :
10636 shdr([<name>[,<occ>]]) : exact string match
10637 shdr_beg([<name>[,<occ>]]) : prefix match
10638 shdr_dir([<name>[,<occ>]]) : subdir match
10639 shdr_dom([<name>[,<occ>]]) : domain match
10640 shdr_end([<name>[,<occ>]]) : suffix match
10641 shdr_len([<name>[,<occ>]]) : length match
10642 shdr_reg([<name>[,<occ>]]) : regex match
10643 shdr_sub([<name>[,<occ>]]) : substring match
10644
10645res.hdr_cnt([<name>]) : integer
10646shdr_cnt([<name>]) : integer (deprecated)
10647 Returns an integer value representing the number of occurrences of response
10648 header field name <name>, or the total number of header fields if <name> is
10649 not specified. The function considers any comma as a delimiter for distinct
10650 values. If this is not desired, the res.fhdr_cnt() fetch should be used
10651 instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010652
Willy Tarreau74ca5042013-06-11 23:12:07 +020010653res.hdr_ip([<name>[,<occ>]]) : ip
10654shdr_ip([<name>[,<occ>]]) : ip (deprecated)
10655 This extracts the last occurrence of header <name> in an HTTP response,
10656 convert it to an IPv4 or IPv6 address and returns this address. Optionally, a
10657 specific occurrence might be specified as a position number. Positive values
10658 indicate a position from the first occurrence, with 1 being the first one.
10659 Negative values indicate positions relative to the last one, with -1 being
10660 the last one. This can be useful to learn some data into a stick table.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010661
Willy Tarreau74ca5042013-06-11 23:12:07 +020010662res.hdr_val([<name>[,<occ>]]) : integer
10663shdr_val([<name>[,<occ>]]) : integer (deprecated)
10664 This extracts the last occurrence of header <name> in an HTTP response, and
10665 converts it to an integer value. Optionally, a specific occurrence might be
10666 specified as a position number. Positive values indicate a position from the
10667 first occurrence, with 1 being the first one. Negative values indicate
10668 positions relative to the last one, with -1 being the last one. This can be
10669 useful to learn some data into a stick table.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010010670
Willy Tarreau74ca5042013-06-11 23:12:07 +020010671res.ver : string
10672resp_ver : string (deprecated)
10673 Returns the version string from the HTTP response, for example "1.1". This
10674 can be useful for logs, but is mostly there for ACL.
Willy Tarreau0e698542011-09-16 08:32:32 +020010675
Willy Tarreau74ca5042013-06-11 23:12:07 +020010676 ACL derivatives :
10677 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010010678
Willy Tarreau74ca5042013-06-11 23:12:07 +020010679set-cookie([<name>]) : string (deprecated)
10680 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
10681 header line from the response and uses the corresponding value to match. This
10682 can be comparable to what "appsession" does with default options, but with
10683 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010010684
Willy Tarreau74ca5042013-06-11 23:12:07 +020010685 This fetch function is deprecated and has been superseded by the "res.cook"
10686 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010010687
Willy Tarreau74ca5042013-06-11 23:12:07 +020010688 See also : "appsession"
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010689
Willy Tarreau74ca5042013-06-11 23:12:07 +020010690status : integer
10691 Returns an integer containing the HTTP status code in the HTTP response, for
10692 example, 302. It is mostly used within ACLs and integer ranges, for example,
10693 to remove any Location header if the response is not a 3xx.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010694
Willy Tarreau74ca5042013-06-11 23:12:07 +020010695url : string
10696 This extracts the request's URL as presented in the request. A typical use is
10697 with prefetch-capable caches, and with portals which need to aggregate
10698 multiple information from databases and keep them in caches. With ACLs, using
10699 "path" is preferred over using "url", because clients may send a full URL as
10700 is normally done with proxies. The only real use is to match "*" which does
10701 not match in "path", and for which there is already a predefined ACL. See
10702 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010703
Willy Tarreau74ca5042013-06-11 23:12:07 +020010704 ACL derivatives :
10705 url : exact string match
10706 url_beg : prefix match
10707 url_dir : subdir match
10708 url_dom : domain match
10709 url_end : suffix match
10710 url_len : length match
10711 url_reg : regex match
10712 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010713
Willy Tarreau74ca5042013-06-11 23:12:07 +020010714url_ip : ip
10715 This extracts the IP address from the request's URL when the host part is
10716 presented as an IP address. Its use is very limited. For instance, a
10717 monitoring system might use this field as an alternative for the source IP in
10718 order to test what path a given source address would follow, or to force an
10719 entry in a table for a given source address. With ACLs it can be used to
10720 restrict access to certain systems through a proxy, for example when combined
10721 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010722
Willy Tarreau74ca5042013-06-11 23:12:07 +020010723url_port : integer
10724 This extracts the port part from the request's URL. Note that if the port is
10725 not specified in the request, port 80 is assumed. With ACLs it can be used to
10726 restrict access to certain systems through a proxy, for example when combined
10727 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010728
Willy Tarreau74ca5042013-06-11 23:12:07 +020010729urlp(<name>[,<delim>]) : string
10730url_param(<name>[,<delim>]) : string
10731 This extracts the first occurrence of the parameter <name> in the query
10732 string, which begins after either '?' or <delim>, and which ends before '&',
10733 ';' or <delim>. The parameter name is case-sensitive. The result is a string
10734 corresponding to the value of the parameter <name> as presented in the
10735 request (no URL decoding is performed). This can be used for session
10736 stickiness based on a client ID, to extract an application cookie passed as a
10737 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
10738 this fetch do not iterate over multiple parameters and stop at the first one
10739 as well.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010740
Willy Tarreau74ca5042013-06-11 23:12:07 +020010741 ACL derivatives :
10742 urlp(<name>[,<delim>]) : exact string match
10743 urlp_beg(<name>[,<delim>]) : prefix match
10744 urlp_dir(<name>[,<delim>]) : subdir match
10745 urlp_dom(<name>[,<delim>]) : domain match
10746 urlp_end(<name>[,<delim>]) : suffix match
10747 urlp_len(<name>[,<delim>]) : length match
10748 urlp_reg(<name>[,<delim>]) : regex match
10749 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010750
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010751
Willy Tarreau74ca5042013-06-11 23:12:07 +020010752 Example :
10753 # match http://example.com/foo?PHPSESSIONID=some_id
10754 stick on urlp(PHPSESSIONID)
10755 # match http://example.com/foo;JSESSIONID=some_id
10756 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020010757
Willy Tarreau74ca5042013-06-11 23:12:07 +020010758urlp_val(<name>[,<delim>]) : integer
10759 See "urlp" above. This one extracts the URL parameter <name> in the request
10760 and converts it to an integer value. This can be used for session stickiness
10761 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020010762
Willy Tarreau198a7442008-01-17 12:05:32 +010010763
Willy Tarreau74ca5042013-06-11 23:12:07 +0200107647.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010765---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010010766
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010767Some predefined ACLs are hard-coded so that they do not have to be declared in
10768every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020010769order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010010770
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010771ACL name Equivalent to Usage
10772---------------+-----------------------------+---------------------------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010773FALSE always_false never match
Willy Tarreau2492d5b2009-07-11 00:06:00 +020010774HTTP req_proto_http match if protocol is valid HTTP
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010775HTTP_1.0 req_ver 1.0 match HTTP version 1.0
10776HTTP_1.1 req_ver 1.1 match HTTP version 1.1
Willy Tarreaud63335a2010-02-26 12:56:52 +010010777HTTP_CONTENT hdr_val(content-length) gt 0 match an existing content-length
10778HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
10779HTTP_URL_SLASH url_beg / match URL beginning with "/"
10780HTTP_URL_STAR url * match URL equal to "*"
10781LOCALHOST src 127.0.0.1/8 match connection from local host
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010782METH_CONNECT method CONNECT match HTTP CONNECT method
10783METH_GET method GET HEAD match HTTP GET or HEAD method
10784METH_HEAD method HEAD match HTTP HEAD method
10785METH_OPTIONS method OPTIONS match HTTP OPTIONS method
10786METH_POST method POST match HTTP POST method
10787METH_TRACE method TRACE match HTTP TRACE method
Emeric Brunbede3d02009-06-30 17:54:00 +020010788RDP_COOKIE req_rdp_cookie_cnt gt 0 match presence of an RDP cookie
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010789REQ_CONTENT req_len gt 0 match data in the request buffer
Willy Tarreaud63335a2010-02-26 12:56:52 +010010790TRUE always_true always match
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010791WAIT_END wait_end wait for end of content analysis
10792---------------+-----------------------------+---------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010010793
Willy Tarreaub937b7e2010-01-12 15:27:54 +010010794
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200107958. Logging
10796----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010010797
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010798One of HAProxy's strong points certainly lies is its precise logs. It probably
10799provides the finest level of information available for such a product, which is
10800very important for troubleshooting complex environments. Standard information
10801provided in logs include client ports, TCP/HTTP state timers, precise session
10802state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010803to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010804headers.
10805
10806In order to improve administrators reactivity, it offers a great transparency
10807about encountered problems, both internal and external, and it is possible to
10808send logs to different sources at the same time with different level filters :
10809
10810 - global process-level logs (system errors, start/stop, etc..)
10811 - per-instance system and internal errors (lack of resource, bugs, ...)
10812 - per-instance external troubles (servers up/down, max connections)
10813 - per-instance activity (client connections), either at the establishment or
10814 at the termination.
10815
10816The ability to distribute different levels of logs to different log servers
10817allow several production teams to interact and to fix their problems as soon
10818as possible. For example, the system team might monitor system-wide errors,
10819while the application team might be monitoring the up/down for their servers in
10820real time, and the security team might analyze the activity logs with one hour
10821delay.
10822
10823
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200108248.1. Log levels
10825---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010826
Simon Hormandf791f52011-05-29 15:01:10 +090010827TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010828source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090010829HTTP request, HTTP return code, number of bytes transmitted, conditions
10830in which the session ended, and even exchanged cookies values. For example
10831track a particular user's problems. All messages may be sent to up to two
10832syslog servers. Check the "log" keyword in section 4.2 for more information
10833about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010834
10835
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200108368.2. Log formats
10837----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010838
William Lallemand48940402012-01-30 16:47:22 +010010839HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090010840and will be detailed in the following sections. A few of them may vary
10841slightly with the configuration, due to indicators specific to certain
10842options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010843
10844 - the default format, which is very basic and very rarely used. It only
10845 provides very basic information about the incoming connection at the moment
10846 it is accepted : source IP:port, destination IP:port, and frontend-name.
10847 This mode will eventually disappear so it will not be described to great
10848 extents.
10849
10850 - the TCP format, which is more advanced. This format is enabled when "option
10851 tcplog" is set on the frontend. HAProxy will then usually wait for the
10852 connection to terminate before logging. This format provides much richer
10853 information, such as timers, connection counts, queue size, etc... This
10854 format is recommended for pure TCP proxies.
10855
10856 - the HTTP format, which is the most advanced for HTTP proxying. This format
10857 is enabled when "option httplog" is set on the frontend. It provides the
10858 same information as the TCP format with some HTTP-specific fields such as
10859 the request, the status code, and captures of headers and cookies. This
10860 format is recommended for HTTP proxies.
10861
Emeric Brun3a058f32009-06-30 18:26:00 +020010862 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
10863 fields arranged in the same order as the CLF format. In this mode, all
10864 timers, captures, flags, etc... appear one per field after the end of the
10865 common fields, in the same order they appear in the standard HTTP format.
10866
William Lallemand48940402012-01-30 16:47:22 +010010867 - the custom log format, allows you to make your own log line.
10868
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010869Next sections will go deeper into details for each of these formats. Format
10870specification will be performed on a "field" basis. Unless stated otherwise, a
10871field is a portion of text delimited by any number of spaces. Since syslog
10872servers are susceptible of inserting fields at the beginning of a line, it is
10873always assumed that the first field is the one containing the process name and
10874identifier.
10875
10876Note : Since log lines may be quite long, the log examples in sections below
10877 might be broken into multiple lines. The example log lines will be
10878 prefixed with 3 closing angle brackets ('>>>') and each time a log is
10879 broken into multiple lines, each non-final line will end with a
10880 backslash ('\') and the next line will start indented by two characters.
10881
10882
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200108838.2.1. Default log format
10884-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010885
10886This format is used when no specific option is set. The log is emitted as soon
10887as the connection is accepted. One should note that this currently is the only
10888format which logs the request's destination IP and ports.
10889
10890 Example :
10891 listen www
10892 mode http
10893 log global
10894 server srv1 127.0.0.1:8000
10895
10896 >>> Feb 6 12:12:09 localhost \
10897 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
10898 (www/HTTP)
10899
10900 Field Format Extract from the example above
10901 1 process_name '[' pid ']:' haproxy[14385]:
10902 2 'Connect from' Connect from
10903 3 source_ip ':' source_port 10.0.1.2:33312
10904 4 'to' to
10905 5 destination_ip ':' destination_port 10.0.3.31:8012
10906 6 '(' frontend_name '/' mode ')' (www/HTTP)
10907
10908Detailed fields description :
10909 - "source_ip" is the IP address of the client which initiated the connection.
10910 - "source_port" is the TCP port of the client which initiated the connection.
10911 - "destination_ip" is the IP address the client connected to.
10912 - "destination_port" is the TCP port the client connected to.
10913 - "frontend_name" is the name of the frontend (or listener) which received
10914 and processed the connection.
10915 - "mode is the mode the frontend is operating (TCP or HTTP).
10916
Willy Tarreauceb24bc2010-11-09 12:46:41 +010010917In case of a UNIX socket, the source and destination addresses are marked as
10918"unix:" and the ports reflect the internal ID of the socket which accepted the
10919connection (the same ID as reported in the stats).
10920
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010921It is advised not to use this deprecated format for newer installations as it
10922will eventually disappear.
10923
10924
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200109258.2.2. TCP log format
10926---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010927
10928The TCP format is used when "option tcplog" is specified in the frontend, and
10929is the recommended format for pure TCP proxies. It provides a lot of precious
10930information for troubleshooting. Since this format includes timers and byte
10931counts, the log is normally emitted at the end of the session. It can be
10932emitted earlier if "option logasap" is specified, which makes sense in most
10933environments with long sessions such as remote terminals. Sessions which match
10934the "monitor" rules are never logged. It is also possible not to emit logs for
10935sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020010936specifying "option dontlognull" in the frontend. Successful connections will
10937not be logged if "option dontlog-normal" is specified in the frontend. A few
10938fields may slightly vary depending on some configuration options, those are
10939marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010940
10941 Example :
10942 frontend fnt
10943 mode tcp
10944 option tcplog
10945 log global
10946 default_backend bck
10947
10948 backend bck
10949 server srv1 127.0.0.1:8000
10950
10951 >>> Feb 6 12:12:56 localhost \
10952 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
10953 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
10954
10955 Field Format Extract from the example above
10956 1 process_name '[' pid ']:' haproxy[14387]:
10957 2 client_ip ':' client_port 10.0.1.2:33313
10958 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
10959 4 frontend_name fnt
10960 5 backend_name '/' server_name bck/srv1
10961 6 Tw '/' Tc '/' Tt* 0/0/5007
10962 7 bytes_read* 212
10963 8 termination_state --
10964 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
10965 10 srv_queue '/' backend_queue 0/0
10966
10967Detailed fields description :
10968 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010010969 connection to haproxy. If the connection was accepted on a UNIX socket
10970 instead, the IP address would be replaced with the word "unix". Note that
10971 when the connection is accepted on a socket configured with "accept-proxy"
10972 and the PROXY protocol is correctly used, then the logs will reflect the
10973 forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010974
10975 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010010976 If the connection was accepted on a UNIX socket instead, the port would be
10977 replaced with the ID of the accepting socket, which is also reported in the
10978 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010010979
10980 - "accept_date" is the exact date when the connection was received by haproxy
10981 (which might be very slightly different from the date observed on the
10982 network if there was some queuing in the system's backlog). This is usually
10983 the same date which may appear in any upstream firewall's log.
10984
10985 - "frontend_name" is the name of the frontend (or listener) which received
10986 and processed the connection.
10987
10988 - "backend_name" is the name of the backend (or listener) which was selected
10989 to manage the connection to the server. This will be the same as the
10990 frontend if no switching rule has been applied, which is common for TCP
10991 applications.
10992
10993 - "server_name" is the name of the last server to which the connection was
10994 sent, which might differ from the first one if there were connection errors
10995 and a redispatch occurred. Note that this server belongs to the backend
10996 which processed the request. If the connection was aborted before reaching
10997 a server, "<NOSRV>" is indicated instead of a server name.
10998
10999 - "Tw" is the total time in milliseconds spent waiting in the various queues.
11000 It can be "-1" if the connection was aborted before reaching the queue.
11001 See "Timers" below for more details.
11002
11003 - "Tc" is the total time in milliseconds spent waiting for the connection to
11004 establish to the final server, including retries. It can be "-1" if the
11005 connection was aborted before a connection could be established. See
11006 "Timers" below for more details.
11007
11008 - "Tt" is the total time in milliseconds elapsed between the accept and the
11009 last close. It covers all possible processings. There is one exception, if
11010 "option logasap" was specified, then the time counting stops at the moment
11011 the log is emitted. In this case, a '+' sign is prepended before the value,
11012 indicating that the final one will be larger. See "Timers" below for more
11013 details.
11014
11015 - "bytes_read" is the total number of bytes transmitted from the server to
11016 the client when the log is emitted. If "option logasap" is specified, the
11017 this value will be prefixed with a '+' sign indicating that the final one
11018 may be larger. Please note that this value is a 64-bit counter, so log
11019 analysis tools must be able to handle it without overflowing.
11020
11021 - "termination_state" is the condition the session was in when the session
11022 ended. This indicates the session state, which side caused the end of
11023 session to happen, and for what reason (timeout, error, ...). The normal
11024 flags should be "--", indicating the session was closed by either end with
11025 no data remaining in buffers. See below "Session state at disconnection"
11026 for more details.
11027
11028 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040011029 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011030 limits have been reached. For instance, if actconn is close to 512 when
11031 multiple connection errors occur, chances are high that the system limits
11032 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011033 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011034
11035 - "feconn" is the total number of concurrent connections on the frontend when
11036 the session was logged. It is useful to estimate the amount of resource
11037 required to sustain high loads, and to detect when the frontend's "maxconn"
11038 has been reached. Most often when this value increases by huge jumps, it is
11039 because there is congestion on the backend servers, but sometimes it can be
11040 caused by a denial of service attack.
11041
11042 - "beconn" is the total number of concurrent connections handled by the
11043 backend when the session was logged. It includes the total number of
11044 concurrent connections active on servers as well as the number of
11045 connections pending in queues. It is useful to estimate the amount of
11046 additional servers needed to support high loads for a given application.
11047 Most often when this value increases by huge jumps, it is because there is
11048 congestion on the backend servers, but sometimes it can be caused by a
11049 denial of service attack.
11050
11051 - "srv_conn" is the total number of concurrent connections still active on
11052 the server when the session was logged. It can never exceed the server's
11053 configured "maxconn" parameter. If this value is very often close or equal
11054 to the server's "maxconn", it means that traffic regulation is involved a
11055 lot, meaning that either the server's maxconn value is too low, or that
11056 there aren't enough servers to process the load with an optimal response
11057 time. When only one of the server's "srv_conn" is high, it usually means
11058 that this server has some trouble causing the connections to take longer to
11059 be processed than on other servers.
11060
11061 - "retries" is the number of connection retries experienced by this session
11062 when trying to connect to the server. It must normally be zero, unless a
11063 server is being stopped at the same moment the connection was attempted.
11064 Frequent retries generally indicate either a network problem between
11065 haproxy and the server, or a misconfigured system backlog on the server
11066 preventing new connections from being queued. This field may optionally be
11067 prefixed with a '+' sign, indicating that the session has experienced a
11068 redispatch after the maximal retry count has been reached on the initial
11069 server. In this case, the server name appearing in the log is the one the
11070 connection was redispatched to, and not the first one, though both may
11071 sometimes be the same in case of hashing for instance. So as a general rule
11072 of thumb, when a '+' is present in front of the retry count, this count
11073 should not be attributed to the logged server.
11074
11075 - "srv_queue" is the total number of requests which were processed before
11076 this one in the server queue. It is zero when the request has not gone
11077 through the server queue. It makes it possible to estimate the approximate
11078 server's response time by dividing the time spent in queue by the number of
11079 requests in the queue. It is worth noting that if a session experiences a
11080 redispatch and passes through two server queues, their positions will be
11081 cumulated. A request should not pass through both the server queue and the
11082 backend queue unless a redispatch occurs.
11083
11084 - "backend_queue" is the total number of requests which were processed before
11085 this one in the backend's global queue. It is zero when the request has not
11086 gone through the global queue. It makes it possible to estimate the average
11087 queue length, which easily translates into a number of missing servers when
11088 divided by a server's "maxconn" parameter. It is worth noting that if a
11089 session experiences a redispatch, it may pass twice in the backend's queue,
11090 and then both positions will be cumulated. A request should not pass
11091 through both the server queue and the backend queue unless a redispatch
11092 occurs.
11093
11094
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200110958.2.3. HTTP log format
11096----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011097
11098The HTTP format is the most complete and the best suited for HTTP proxies. It
11099is enabled by when "option httplog" is specified in the frontend. It provides
11100the same level of information as the TCP format with additional features which
11101are specific to the HTTP protocol. Just like the TCP format, the log is usually
11102emitted at the end of the session, unless "option logasap" is specified, which
11103generally only makes sense for download sites. A session which matches the
11104"monitor" rules will never logged. It is also possible not to log sessions for
11105which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020011106frontend. Successful connections will not be logged if "option dontlog-normal"
11107is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011108
11109Most fields are shared with the TCP log, some being different. A few fields may
11110slightly vary depending on some configuration options. Those ones are marked
11111with a star ('*') after the field name below.
11112
11113 Example :
11114 frontend http-in
11115 mode http
11116 option httplog
11117 log global
11118 default_backend bck
11119
11120 backend static
11121 server srv1 127.0.0.1:8000
11122
11123 >>> Feb 6 12:14:14 localhost \
11124 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
11125 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 +010011126 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011127
11128 Field Format Extract from the example above
11129 1 process_name '[' pid ']:' haproxy[14389]:
11130 2 client_ip ':' client_port 10.0.1.2:33317
11131 3 '[' accept_date ']' [06/Feb/2009:12:14:14.655]
11132 4 frontend_name http-in
11133 5 backend_name '/' server_name static/srv1
11134 6 Tq '/' Tw '/' Tc '/' Tr '/' Tt* 10/0/30/69/109
11135 7 status_code 200
11136 8 bytes_read* 2750
11137 9 captured_request_cookie -
11138 10 captured_response_cookie -
11139 11 termination_state ----
11140 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
11141 13 srv_queue '/' backend_queue 0/0
11142 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
11143 15 '{' captured_response_headers* '}' {}
11144 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010011145
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011146
11147Detailed fields description :
11148 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010011149 connection to haproxy. If the connection was accepted on a UNIX socket
11150 instead, the IP address would be replaced with the word "unix". Note that
11151 when the connection is accepted on a socket configured with "accept-proxy"
11152 and the PROXY protocol is correctly used, then the logs will reflect the
11153 forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011154
11155 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010011156 If the connection was accepted on a UNIX socket instead, the port would be
11157 replaced with the ID of the accepting socket, which is also reported in the
11158 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011159
11160 - "accept_date" is the exact date when the TCP connection was received by
11161 haproxy (which might be very slightly different from the date observed on
11162 the network if there was some queuing in the system's backlog). This is
11163 usually the same date which may appear in any upstream firewall's log. This
11164 does not depend on the fact that the client has sent the request or not.
11165
11166 - "frontend_name" is the name of the frontend (or listener) which received
11167 and processed the connection.
11168
11169 - "backend_name" is the name of the backend (or listener) which was selected
11170 to manage the connection to the server. This will be the same as the
11171 frontend if no switching rule has been applied.
11172
11173 - "server_name" is the name of the last server to which the connection was
11174 sent, which might differ from the first one if there were connection errors
11175 and a redispatch occurred. Note that this server belongs to the backend
11176 which processed the request. If the request was aborted before reaching a
11177 server, "<NOSRV>" is indicated instead of a server name. If the request was
11178 intercepted by the stats subsystem, "<STATS>" is indicated instead.
11179
11180 - "Tq" is the total time in milliseconds spent waiting for the client to send
11181 a full HTTP request, not counting data. It can be "-1" if the connection
11182 was aborted before a complete request could be received. It should always
11183 be very small because a request generally fits in one single packet. Large
11184 times here generally indicate network trouble between the client and
11185 haproxy. See "Timers" below for more details.
11186
11187 - "Tw" is the total time in milliseconds spent waiting in the various queues.
11188 It can be "-1" if the connection was aborted before reaching the queue.
11189 See "Timers" below for more details.
11190
11191 - "Tc" is the total time in milliseconds spent waiting for the connection to
11192 establish to the final server, including retries. It can be "-1" if the
11193 request was aborted before a connection could be established. See "Timers"
11194 below for more details.
11195
11196 - "Tr" is the total time in milliseconds spent waiting for the server to send
11197 a full HTTP response, not counting data. It can be "-1" if the request was
11198 aborted before a complete response could be received. It generally matches
11199 the server's processing time for the request, though it may be altered by
11200 the amount of data sent by the client to the server. Large times here on
11201 "GET" requests generally indicate an overloaded server. See "Timers" below
11202 for more details.
11203
11204 - "Tt" is the total time in milliseconds elapsed between the accept and the
11205 last close. It covers all possible processings. There is one exception, if
11206 "option logasap" was specified, then the time counting stops at the moment
11207 the log is emitted. In this case, a '+' sign is prepended before the value,
11208 indicating that the final one will be larger. See "Timers" below for more
11209 details.
11210
11211 - "status_code" is the HTTP status code returned to the client. This status
11212 is generally set by the server, but it might also be set by haproxy when
11213 the server cannot be reached or when its response is blocked by haproxy.
11214
11215 - "bytes_read" is the total number of bytes transmitted to the client when
11216 the log is emitted. This does include HTTP headers. If "option logasap" is
11217 specified, the this value will be prefixed with a '+' sign indicating that
11218 the final one may be larger. Please note that this value is a 64-bit
11219 counter, so log analysis tools must be able to handle it without
11220 overflowing.
11221
11222 - "captured_request_cookie" is an optional "name=value" entry indicating that
11223 the client had this cookie in the request. The cookie name and its maximum
11224 length are defined by the "capture cookie" statement in the frontend
11225 configuration. The field is a single dash ('-') when the option is not
11226 set. Only one cookie may be captured, it is generally used to track session
11227 ID exchanges between a client and a server to detect session crossing
11228 between clients due to application bugs. For more details, please consult
11229 the section "Capturing HTTP headers and cookies" below.
11230
11231 - "captured_response_cookie" is an optional "name=value" entry indicating
11232 that the server has returned a cookie with its response. The cookie name
11233 and its maximum length are defined by the "capture cookie" statement in the
11234 frontend configuration. The field is a single dash ('-') when the option is
11235 not set. Only one cookie may be captured, it is generally used to track
11236 session ID exchanges between a client and a server to detect session
11237 crossing between clients due to application bugs. For more details, please
11238 consult the section "Capturing HTTP headers and cookies" below.
11239
11240 - "termination_state" is the condition the session was in when the session
11241 ended. This indicates the session state, which side caused the end of
11242 session to happen, for what reason (timeout, error, ...), just like in TCP
11243 logs, and information about persistence operations on cookies in the last
11244 two characters. The normal flags should begin with "--", indicating the
11245 session was closed by either end with no data remaining in buffers. See
11246 below "Session state at disconnection" for more details.
11247
11248 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040011249 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011250 limits have been reached. For instance, if actconn is close to 512 or 1024
11251 when multiple connection errors occur, chances are high that the system
11252 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011253 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011254 system.
11255
11256 - "feconn" is the total number of concurrent connections on the frontend when
11257 the session was logged. It is useful to estimate the amount of resource
11258 required to sustain high loads, and to detect when the frontend's "maxconn"
11259 has been reached. Most often when this value increases by huge jumps, it is
11260 because there is congestion on the backend servers, but sometimes it can be
11261 caused by a denial of service attack.
11262
11263 - "beconn" is the total number of concurrent connections handled by the
11264 backend when the session was logged. It includes the total number of
11265 concurrent connections active on servers as well as the number of
11266 connections pending in queues. It is useful to estimate the amount of
11267 additional servers needed to support high loads for a given application.
11268 Most often when this value increases by huge jumps, it is because there is
11269 congestion on the backend servers, but sometimes it can be caused by a
11270 denial of service attack.
11271
11272 - "srv_conn" is the total number of concurrent connections still active on
11273 the server when the session was logged. It can never exceed the server's
11274 configured "maxconn" parameter. If this value is very often close or equal
11275 to the server's "maxconn", it means that traffic regulation is involved a
11276 lot, meaning that either the server's maxconn value is too low, or that
11277 there aren't enough servers to process the load with an optimal response
11278 time. When only one of the server's "srv_conn" is high, it usually means
11279 that this server has some trouble causing the requests to take longer to be
11280 processed than on other servers.
11281
11282 - "retries" is the number of connection retries experienced by this session
11283 when trying to connect to the server. It must normally be zero, unless a
11284 server is being stopped at the same moment the connection was attempted.
11285 Frequent retries generally indicate either a network problem between
11286 haproxy and the server, or a misconfigured system backlog on the server
11287 preventing new connections from being queued. This field may optionally be
11288 prefixed with a '+' sign, indicating that the session has experienced a
11289 redispatch after the maximal retry count has been reached on the initial
11290 server. In this case, the server name appearing in the log is the one the
11291 connection was redispatched to, and not the first one, though both may
11292 sometimes be the same in case of hashing for instance. So as a general rule
11293 of thumb, when a '+' is present in front of the retry count, this count
11294 should not be attributed to the logged server.
11295
11296 - "srv_queue" is the total number of requests which were processed before
11297 this one in the server queue. It is zero when the request has not gone
11298 through the server queue. It makes it possible to estimate the approximate
11299 server's response time by dividing the time spent in queue by the number of
11300 requests in the queue. It is worth noting that if a session experiences a
11301 redispatch and passes through two server queues, their positions will be
11302 cumulated. A request should not pass through both the server queue and the
11303 backend queue unless a redispatch occurs.
11304
11305 - "backend_queue" is the total number of requests which were processed before
11306 this one in the backend's global queue. It is zero when the request has not
11307 gone through the global queue. It makes it possible to estimate the average
11308 queue length, which easily translates into a number of missing servers when
11309 divided by a server's "maxconn" parameter. It is worth noting that if a
11310 session experiences a redispatch, it may pass twice in the backend's queue,
11311 and then both positions will be cumulated. A request should not pass
11312 through both the server queue and the backend queue unless a redispatch
11313 occurs.
11314
11315 - "captured_request_headers" is a list of headers captured in the request due
11316 to the presence of the "capture request header" statement in the frontend.
11317 Multiple headers can be captured, they will be delimited by a vertical bar
11318 ('|'). When no capture is enabled, the braces do not appear, causing a
11319 shift of remaining fields. It is important to note that this field may
11320 contain spaces, and that using it requires a smarter log parser than when
11321 it's not used. Please consult the section "Capturing HTTP headers and
11322 cookies" below for more details.
11323
11324 - "captured_response_headers" is a list of headers captured in the response
11325 due to the presence of the "capture response header" statement in the
11326 frontend. Multiple headers can be captured, they will be delimited by a
11327 vertical bar ('|'). When no capture is enabled, the braces do not appear,
11328 causing a shift of remaining fields. It is important to note that this
11329 field may contain spaces, and that using it requires a smarter log parser
11330 than when it's not used. Please consult the section "Capturing HTTP headers
11331 and cookies" below for more details.
11332
11333 - "http_request" is the complete HTTP request line, including the method,
11334 request and HTTP version string. Non-printable characters are encoded (see
11335 below the section "Non-printable characters"). This is always the last
11336 field, and it is always delimited by quotes and is the only one which can
11337 contain quotes. If new fields are added to the log format, they will be
11338 added before this field. This field might be truncated if the request is
11339 huge and does not fit in the standard syslog buffer (1024 characters). This
11340 is the reason why this field must always remain the last one.
11341
11342
Cyril Bontédc4d9032012-04-08 21:57:39 +0200113438.2.4. Custom log format
11344------------------------
William Lallemand48940402012-01-30 16:47:22 +010011345
Willy Tarreau2beef582012-12-20 17:22:52 +010011346The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010011347mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010011348
11349HAproxy understands some log format variables. % precedes log format variables.
11350Variables can take arguments using braces ('{}'), and multiple arguments are
11351separated by commas within the braces. Flags may be added or removed by
11352prefixing them with a '+' or '-' sign.
11353
11354Special variable "%o" may be used to propagate its flags to all other
11355variables on the same format string. This is particularly handy with quoted
11356string formats ("Q").
11357
Willy Tarreauc8368452012-12-21 00:09:23 +010011358If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreau74ca5042013-06-11 23:12:07 +020011359as a pattern extraction rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010011360less common information such as the client's SSL certificate's DN, or to log
11361the key that would be used to store an entry into a stick table.
11362
William Lallemand48940402012-01-30 16:47:22 +010011363Note: spaces must be escaped. A space character is considered as a separator.
Willy Tarreau06d97f92013-12-02 17:45:48 +010011364In order to emit a verbatim '%', it must be preceeded by another '%' resulting
11365in '%%'. HAProxy will automatically merge consecutive separators.
William Lallemand48940402012-01-30 16:47:22 +010011366
11367Flags are :
11368 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040011369 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
William Lallemand48940402012-01-30 16:47:22 +010011370
11371 Example:
11372
11373 log-format %T\ %t\ Some\ Text
11374 log-format %{+Q}o\ %t\ %s\ %{-Q}r
11375
11376At the moment, the default HTTP format is defined this way :
11377
Willy Tarreau2beef582012-12-20 17:22:52 +010011378 log-format %ci:%cp\ [%t]\ %ft\ %b/%s\ %Tq/%Tw/%Tc/%Tr/%Tt\ %ST\ %B\ %CC\ \
11379 %CS\ %tsc\ %ac/%fc/%bc/%sc/%rc\ %sq/%bq\ %hr\ %hs\ %{+Q}r
William Lallemand48940402012-01-30 16:47:22 +010011380
William Lallemandbddd4fd2012-02-27 11:23:10 +010011381the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010011382
Willy Tarreau2beef582012-12-20 17:22:52 +010011383 log-format %{+Q}o\ %{-Q}ci\ -\ -\ [%T]\ %r\ %ST\ %B\ \"\"\ \"\"\ %cp\ \
Willy Tarreau773d65f2012-10-12 14:56:11 +020011384 %ms\ %ft\ %b\ %s\ \%Tq\ %Tw\ %Tc\ %Tr\ %Tt\ %tsc\ %ac\ %fc\ \
Willy Tarreau2beef582012-12-20 17:22:52 +010011385 %bc\ %sc\ %rc\ %sq\ %bq\ %CC\ %CS\ \%hrl\ %hsl
William Lallemand48940402012-01-30 16:47:22 +010011386
William Lallemandbddd4fd2012-02-27 11:23:10 +010011387and the default TCP format is defined this way :
11388
Willy Tarreau2beef582012-12-20 17:22:52 +010011389 log-format %ci:%cp\ [%t]\ %ft\ %b/%s\ %Tw/%Tc/%Tt\ %B\ %ts\ \
William Lallemandbddd4fd2012-02-27 11:23:10 +010011390 %ac/%fc/%bc/%sc/%rc\ %sq/%bq
11391
William Lallemand48940402012-01-30 16:47:22 +010011392Please refer to the table below for currently defined variables :
11393
William Lallemandbddd4fd2012-02-27 11:23:10 +010011394 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020011395 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010011396 +---+------+-----------------------------------------------+-------------+
11397 | | %o | special variable, apply flags on all next var | |
11398 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010011399 | | %B | bytes_read (from server to client) | numeric |
11400 | H | %CC | captured_request_cookie | string |
11401 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020011402 | | %H | hostname | string |
William Lallemanda73203e2012-03-12 12:48:57 +010011403 | | %ID | unique-id | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010011404 | H | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020011405 | | %T | gmt_date_time | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010011406 | | %Tc | Tc | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080011407 | | %Tl | local_date_time | date |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020011408 | H | %Tq | Tq | numeric |
11409 | H | %Tr | Tr | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020011410 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010011411 | | %Tt | Tt | numeric |
11412 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010011413 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010011414 | | %ac | actconn | numeric |
11415 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010011416 | | %bc | beconn (backend concurrent connections) | numeric |
11417 | | %bi | backend_source_ip (connecting address) | IP |
11418 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010011419 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010011420 | | %ci | client_ip (accepted address) | IP |
11421 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010011422 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010011423 | | %fc | feconn (frontend concurrent connections) | numeric |
11424 | | %fi | frontend_ip (accepting address) | IP |
11425 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020011426 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020011427 | H | %hr | captured_request_headers default style | string |
11428 | H | %hrl | captured_request_headers CLF style | string list |
11429 | H | %hs | captured_response_headers default style | string |
11430 | H | %hsl | captured_response_headers CLF style | string list |
William Lallemandbddd4fd2012-02-27 11:23:10 +010011431 | | %ms | accept date milliseconds | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020011432 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020011433 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010011434 | | %rc | retries | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010011435 | H | %rt | http_request_counter | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010011436 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010011437 | | %sc | srv_conn (server concurrent connections) | numeric |
11438 | | %si | server_IP (target address) | IP |
11439 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010011440 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020011441 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
11442 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010011443 | | %t | date_time (with millisecond resolution) | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010011444 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020011445 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010011446 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010011447
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020011448 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010011449
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010011450
114518.2.5. Error log format
11452-----------------------
11453
11454When an incoming connection fails due to an SSL handshake or an invalid PROXY
11455protocol header, haproxy will log the event using a shorter, fixed line format.
11456By default, logs are emitted at the LOG_INFO level, unless the option
11457"log-separate-errors" is set in the backend, in which case the LOG_ERR level
11458will be used. Connections on which no data are exchanged (eg: probes) are not
11459logged if the "dontlognull" option is set.
11460
11461The format looks like this :
11462
11463 >>> Dec 3 18:27:14 localhost \
11464 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
11465 Connection error during SSL handshake
11466
11467 Field Format Extract from the example above
11468 1 process_name '[' pid ']:' haproxy[6103]:
11469 2 client_ip ':' client_port 127.0.0.1:56059
11470 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
11471 4 frontend_name "/" bind_name ":" frt/f1:
11472 5 message Connection error during SSL handshake
11473
11474These fields just provide minimal information to help debugging connection
11475failures.
11476
11477
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200114788.3. Advanced logging options
11479-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011480
11481Some advanced logging options are often looked for but are not easy to find out
11482just by looking at the various options. Here is an entry point for the few
11483options which can enable better logging. Please refer to the keywords reference
11484for more information about their usage.
11485
11486
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200114878.3.1. Disabling logging of external tests
11488------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011489
11490It is quite common to have some monitoring tools perform health checks on
11491haproxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
11492commercial load-balancer, and sometimes it will simply be a more complete
11493monitoring system such as Nagios. When the tests are very frequent, users often
11494ask how to disable logging for those checks. There are three possibilities :
11495
11496 - if connections come from everywhere and are just TCP probes, it is often
11497 desired to simply disable logging of connections without data exchange, by
11498 setting "option dontlognull" in the frontend. It also disables logging of
11499 port scans, which may or may not be desired.
11500
11501 - if the connection come from a known source network, use "monitor-net" to
11502 declare this network as monitoring only. Any host in this network will then
11503 only be able to perform health checks, and their requests will not be
11504 logged. This is generally appropriate to designate a list of equipments
11505 such as other load-balancers.
11506
11507 - if the tests are performed on a known URI, use "monitor-uri" to declare
11508 this URI as dedicated to monitoring. Any host sending this request will
11509 only get the result of a health-check, and the request will not be logged.
11510
11511
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200115128.3.2. Logging before waiting for the session to terminate
11513----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011514
11515The problem with logging at end of connection is that you have no clue about
11516what is happening during very long sessions, such as remote terminal sessions
11517or large file downloads. This problem can be worked around by specifying
11518"option logasap" in the frontend. Haproxy will then log as soon as possible,
11519just before data transfer begins. This means that in case of TCP, it will still
11520log the connection status to the server, and in case of HTTP, it will log just
11521after processing the server headers. In this case, the number of bytes reported
11522is the number of header bytes sent to the client. In order to avoid confusion
11523with normal logs, the total time field and the number of bytes are prefixed
11524with a '+' sign which means that real numbers are certainly larger.
11525
11526
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200115278.3.3. Raising log level upon errors
11528------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020011529
11530Sometimes it is more convenient to separate normal traffic from errors logs,
11531for instance in order to ease error monitoring from log files. When the option
11532"log-separate-errors" is used, connections which experience errors, timeouts,
11533retries, redispatches or HTTP status codes 5xx will see their syslog level
11534raised from "info" to "err". This will help a syslog daemon store the log in
11535a separate file. It is very important to keep the errors in the normal traffic
11536file too, so that log ordering is not altered. You should also be careful if
11537you already have configured your syslog daemon to store all logs higher than
11538"notice" in an "admin" file, because the "err" level is higher than "notice".
11539
11540
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200115418.3.4. Disabling logging of successful connections
11542--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020011543
11544Although this may sound strange at first, some large sites have to deal with
11545multiple thousands of logs per second and are experiencing difficulties keeping
11546them intact for a long time or detecting errors within them. If the option
11547"dontlog-normal" is set on the frontend, all normal connections will not be
11548logged. In this regard, a normal connection is defined as one without any
11549error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
11550and a response with a status 5xx is not considered normal and will be logged
11551too. Of course, doing is is really discouraged as it will remove most of the
11552useful information from the logs. Do this only if you have no other
11553alternative.
11554
11555
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200115568.4. Timing events
11557------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011558
11559Timers provide a great help in troubleshooting network problems. All values are
11560reported in milliseconds (ms). These timers should be used in conjunction with
11561the session termination flags. In TCP mode with "option tcplog" set on the
11562frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
11563mode, 5 control points are reported under the form "Tq/Tw/Tc/Tr/Tt" :
11564
11565 - Tq: total time to get the client request (HTTP mode only). It's the time
11566 elapsed between the moment the client connection was accepted and the
11567 moment the proxy received the last HTTP header. The value "-1" indicates
11568 that the end of headers (empty line) has never been seen. This happens when
11569 the client closes prematurely or times out.
11570
11571 - Tw: total time spent in the queues waiting for a connection slot. It
11572 accounts for backend queue as well as the server queues, and depends on the
11573 queue size, and the time needed for the server to complete previous
11574 requests. The value "-1" means that the request was killed before reaching
11575 the queue, which is generally what happens with invalid or denied requests.
11576
11577 - Tc: total time to establish the TCP connection to the server. It's the time
11578 elapsed between the moment the proxy sent the connection request, and the
11579 moment it was acknowledged by the server, or between the TCP SYN packet and
11580 the matching SYN/ACK packet in return. The value "-1" means that the
11581 connection never established.
11582
11583 - Tr: server response time (HTTP mode only). It's the time elapsed between
11584 the moment the TCP connection was established to the server and the moment
11585 the server sent its complete response headers. It purely shows its request
11586 processing time, without the network overhead due to the data transmission.
11587 It is worth noting that when the client has data to send to the server, for
11588 instance during a POST request, the time already runs, and this can distort
11589 apparent response time. For this reason, it's generally wise not to trust
11590 too much this field for POST requests initiated from clients behind an
11591 untrusted network. A value of "-1" here means that the last the response
11592 header (empty line) was never seen, most likely because the server timeout
11593 stroke before the server managed to process the request.
11594
11595 - Tt: total session duration time, between the moment the proxy accepted it
11596 and the moment both ends were closed. The exception is when the "logasap"
11597 option is specified. In this case, it only equals (Tq+Tw+Tc+Tr), and is
11598 prefixed with a '+' sign. From this field, we can deduce "Td", the data
11599 transmission time, by substracting other timers when valid :
11600
11601 Td = Tt - (Tq + Tw + Tc + Tr)
11602
11603 Timers with "-1" values have to be excluded from this equation. In TCP
11604 mode, "Tq" and "Tr" have to be excluded too. Note that "Tt" can never be
11605 negative.
11606
11607These timers provide precious indications on trouble causes. Since the TCP
11608protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
11609that timers close to multiples of 3s are nearly always related to lost packets
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011610due to network problems (wires, negotiation, congestion). Moreover, if "Tt" is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011611close to a timeout value specified in the configuration, it often means that a
11612session has been aborted on timeout.
11613
11614Most common cases :
11615
11616 - If "Tq" is close to 3000, a packet has probably been lost between the
11617 client and the proxy. This is very rare on local networks but might happen
11618 when clients are on far remote networks and send large requests. It may
11619 happen that values larger than usual appear here without any network cause.
11620 Sometimes, during an attack or just after a resource starvation has ended,
11621 haproxy may accept thousands of connections in a few milliseconds. The time
11622 spent accepting these connections will inevitably slightly delay processing
11623 of other connections, and it can happen that request times in the order of
11624 a few tens of milliseconds are measured after a few thousands of new
Patrick Mezard105faca2010-06-12 17:02:46 +020011625 connections have been accepted at once. Setting "option http-server-close"
11626 may display larger request times since "Tq" also measures the time spent
11627 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011628
11629 - If "Tc" is close to 3000, a packet has probably been lost between the
11630 server and the proxy during the server connection phase. This value should
11631 always be very low, such as 1 ms on local networks and less than a few tens
11632 of ms on remote networks.
11633
Willy Tarreau55165fe2009-05-10 12:02:55 +020011634 - If "Tr" is nearly always lower than 3000 except some rare values which seem
11635 to be the average majored by 3000, there are probably some packets lost
11636 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011637
11638 - If "Tt" is large even for small byte counts, it generally is because
11639 neither the client nor the server decides to close the connection, for
11640 instance because both have agreed on a keep-alive connection mode. In order
11641 to solve this issue, it will be needed to specify "option httpclose" on
11642 either the frontend or the backend. If the problem persists, it means that
11643 the server ignores the "close" connection mode and expects the client to
11644 close. Then it will be required to use "option forceclose". Having the
11645 smallest possible 'Tt' is important when connection regulation is used with
11646 the "maxconn" option on the servers, since no new connection will be sent
11647 to the server until another one is released.
11648
11649Other noticeable HTTP log cases ('xx' means any value to be ignored) :
11650
11651 Tq/Tw/Tc/Tr/+Tt The "option logasap" is present on the frontend and the log
11652 was emitted before the data phase. All the timers are valid
11653 except "Tt" which is shorter than reality.
11654
11655 -1/xx/xx/xx/Tt The client was not able to send a complete request in time
11656 or it aborted too early. Check the session termination flags
11657 then "timeout http-request" and "timeout client" settings.
11658
11659 Tq/-1/xx/xx/Tt It was not possible to process the request, maybe because
11660 servers were out of order, because the request was invalid
11661 or forbidden by ACL rules. Check the session termination
11662 flags.
11663
11664 Tq/Tw/-1/xx/Tt The connection could not establish on the server. Either it
11665 actively refused it or it timed out after Tt-(Tq+Tw) ms.
11666 Check the session termination flags, then check the
11667 "timeout connect" setting. Note that the tarpit action might
11668 return similar-looking patterns, with "Tw" equal to the time
11669 the client connection was maintained open.
11670
11671 Tq/Tw/Tc/-1/Tt The server has accepted the connection but did not return
11672 a complete response in time, or it closed its connexion
11673 unexpectedly after Tt-(Tq+Tw+Tc) ms. Check the session
11674 termination flags, then check the "timeout server" setting.
11675
11676
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200116778.5. Session state at disconnection
11678-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011679
11680TCP and HTTP logs provide a session termination indicator in the
11681"termination_state" field, just before the number of active connections. It is
116822-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
11683each of which has a special meaning :
11684
11685 - On the first character, a code reporting the first event which caused the
11686 session to terminate :
11687
11688 C : the TCP session was unexpectedly aborted by the client.
11689
11690 S : the TCP session was unexpectedly aborted by the server, or the
11691 server explicitly refused it.
11692
11693 P : the session was prematurely aborted by the proxy, because of a
11694 connection limit enforcement, because a DENY filter was matched,
11695 because of a security check which detected and blocked a dangerous
11696 error in server response which might have caused information leak
Willy Tarreau570f2212013-06-10 16:42:09 +020011697 (eg: cacheable cookie).
11698
11699 L : the session was locally processed by haproxy and was not passed to
11700 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011701
11702 R : a resource on the proxy has been exhausted (memory, sockets, source
11703 ports, ...). Usually, this appears during the connection phase, and
11704 system logs should contain a copy of the precise error. If this
11705 happens, it must be considered as a very serious anomaly which
11706 should be fixed as soon as possible by any means.
11707
11708 I : an internal error was identified by the proxy during a self-check.
11709 This should NEVER happen, and you are encouraged to report any log
11710 containing this, because this would almost certainly be a bug. It
11711 would be wise to preventively restart the process after such an
11712 event too, in case it would be caused by memory corruption.
11713
Simon Horman752dc4a2011-06-21 14:34:59 +090011714 D : the session was killed by haproxy because the server was detected
11715 as down and was configured to kill all connections when going down.
11716
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070011717 U : the session was killed by haproxy on this backup server because an
11718 active server was detected as up and was configured to kill all
11719 backup connections when going up.
11720
Willy Tarreaua2a64e92011-09-07 23:01:56 +020011721 K : the session was actively killed by an admin operating on haproxy.
11722
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011723 c : the client-side timeout expired while waiting for the client to
11724 send or receive data.
11725
11726 s : the server-side timeout expired while waiting for the server to
11727 send or receive data.
11728
11729 - : normal session completion, both the client and the server closed
11730 with nothing left in the buffers.
11731
11732 - on the second character, the TCP or HTTP session state when it was closed :
11733
Willy Tarreauf7b30a92010-12-06 22:59:17 +010011734 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011735 (HTTP mode only). Nothing was sent to any server.
11736
11737 Q : the proxy was waiting in the QUEUE for a connection slot. This can
11738 only happen when servers have a 'maxconn' parameter set. It can
11739 also happen in the global queue after a redispatch consecutive to
11740 a failed attempt to connect to a dying server. If no redispatch is
11741 reported, then no connection attempt was made to any server.
11742
11743 C : the proxy was waiting for the CONNECTION to establish on the
11744 server. The server might at most have noticed a connection attempt.
11745
11746 H : the proxy was waiting for complete, valid response HEADERS from the
11747 server (HTTP only).
11748
11749 D : the session was in the DATA phase.
11750
11751 L : the proxy was still transmitting LAST data to the client while the
11752 server had already finished. This one is very rare as it can only
11753 happen when the client dies while receiving the last packets.
11754
11755 T : the request was tarpitted. It has been held open with the client
11756 during the whole "timeout tarpit" duration or until the client
11757 closed, both of which will be reported in the "Tw" timer.
11758
11759 - : normal session completion after end of data transfer.
11760
11761 - the third character tells whether the persistence cookie was provided by
11762 the client (only in HTTP mode) :
11763
11764 N : the client provided NO cookie. This is usually the case for new
11765 visitors, so counting the number of occurrences of this flag in the
11766 logs generally indicate a valid trend for the site frequentation.
11767
11768 I : the client provided an INVALID cookie matching no known server.
11769 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020011770 cookies between HTTP/HTTPS sites, persistence conditionally
11771 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011772
11773 D : the client provided a cookie designating a server which was DOWN,
11774 so either "option persist" was used and the client was sent to
11775 this server, or it was not set and the client was redispatched to
11776 another server.
11777
Willy Tarreau996a92c2010-10-13 19:30:47 +020011778 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011779 server.
11780
Willy Tarreau996a92c2010-10-13 19:30:47 +020011781 E : the client provided a valid cookie, but with a last date which was
11782 older than what is allowed by the "maxidle" cookie parameter, so
11783 the cookie is consider EXPIRED and is ignored. The request will be
11784 redispatched just as if there was no cookie.
11785
11786 O : the client provided a valid cookie, but with a first date which was
11787 older than what is allowed by the "maxlife" cookie parameter, so
11788 the cookie is consider too OLD and is ignored. The request will be
11789 redispatched just as if there was no cookie.
11790
Willy Tarreauc89ccb62012-04-05 21:18:22 +020011791 U : a cookie was present but was not used to select the server because
11792 some other server selection mechanism was used instead (typically a
11793 "use-server" rule).
11794
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011795 - : does not apply (no cookie set in configuration).
11796
11797 - the last character reports what operations were performed on the persistence
11798 cookie returned by the server (only in HTTP mode) :
11799
11800 N : NO cookie was provided by the server, and none was inserted either.
11801
11802 I : no cookie was provided by the server, and the proxy INSERTED one.
11803 Note that in "cookie insert" mode, if the server provides a cookie,
11804 it will still be overwritten and reported as "I" here.
11805
Willy Tarreau996a92c2010-10-13 19:30:47 +020011806 U : the proxy UPDATED the last date in the cookie that was presented by
11807 the client. This can only happen in insert mode with "maxidle". It
11808 happens everytime there is activity at a different date than the
11809 date indicated in the cookie. If any other change happens, such as
11810 a redispatch, then the cookie will be marked as inserted instead.
11811
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011812 P : a cookie was PROVIDED by the server and transmitted as-is.
11813
11814 R : the cookie provided by the server was REWRITTEN by the proxy, which
11815 happens in "cookie rewrite" or "cookie prefix" modes.
11816
11817 D : the cookie provided by the server was DELETED by the proxy.
11818
11819 - : does not apply (no cookie set in configuration).
11820
Willy Tarreau996a92c2010-10-13 19:30:47 +020011821The combination of the two first flags gives a lot of information about what
11822was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011823helpful to detect server saturation, network troubles, local system resource
11824starvation, attacks, etc...
11825
11826The most common termination flags combinations are indicated below. They are
11827alphabetically sorted, with the lowercase set just after the upper case for
11828easier finding and understanding.
11829
11830 Flags Reason
11831
11832 -- Normal termination.
11833
11834 CC The client aborted before the connection could be established to the
11835 server. This can happen when haproxy tries to connect to a recently
11836 dead (or unchecked) server, and the client aborts while haproxy is
11837 waiting for the server to respond or for "timeout connect" to expire.
11838
11839 CD The client unexpectedly aborted during data transfer. This can be
11840 caused by a browser crash, by an intermediate equipment between the
11841 client and haproxy which decided to actively break the connection,
11842 by network routing issues between the client and haproxy, or by a
11843 keep-alive session between the server and the client terminated first
11844 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010011845
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011846 cD The client did not send nor acknowledge any data for as long as the
11847 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020011848 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011849
11850 CH The client aborted while waiting for the server to start responding.
11851 It might be the server taking too long to respond or the client
11852 clicking the 'Stop' button too fast.
11853
11854 cH The "timeout client" stroke while waiting for client data during a
11855 POST request. This is sometimes caused by too large TCP MSS values
11856 for PPPoE networks which cannot transport full-sized packets. It can
11857 also happen when client timeout is smaller than server timeout and
11858 the server takes too long to respond.
11859
11860 CQ The client aborted while its session was queued, waiting for a server
11861 with enough empty slots to accept it. It might be that either all the
11862 servers were saturated or that the assigned server was taking too
11863 long a time to respond.
11864
11865 CR The client aborted before sending a full HTTP request. Most likely
11866 the request was typed by hand using a telnet client, and aborted
11867 too early. The HTTP status code is likely a 400 here. Sometimes this
11868 might also be caused by an IDS killing the connection between haproxy
11869 and the client.
11870
11871 cR The "timeout http-request" stroke before the client sent a full HTTP
11872 request. This is sometimes caused by too large TCP MSS values on the
11873 client side for PPPoE networks which cannot transport full-sized
11874 packets, or by clients sending requests by hand and not typing fast
11875 enough, or forgetting to enter the empty line at the end of the
11876 request. The HTTP status code is likely a 408 here.
11877
11878 CT The client aborted while its session was tarpitted. It is important to
11879 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020011880 wrong tarpit rules have been written. If a lot of them happen, it
11881 might make sense to lower the "timeout tarpit" value to something
11882 closer to the average reported "Tw" timer, in order not to consume
11883 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011884
Willy Tarreau570f2212013-06-10 16:42:09 +020011885 LR The request was intercepted and locally handled by haproxy. Generally
11886 it means that this was a redirect or a stats request.
11887
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011888 SC The server or an equipment between it and haproxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011889 the TCP connection (the proxy received a TCP RST or an ICMP message
11890 in return). Under some circumstances, it can also be the network
11891 stack telling the proxy that the server is unreachable (eg: no route,
11892 or no ARP response on local network). When this happens in HTTP mode,
11893 the status code is likely a 502 or 503 here.
11894
11895 sC The "timeout connect" stroke before a connection to the server could
11896 complete. When this happens in HTTP mode, the status code is likely a
11897 503 or 504 here.
11898
11899 SD The connection to the server died with an error during the data
11900 transfer. This usually means that haproxy has received an RST from
11901 the server or an ICMP message from an intermediate equipment while
11902 exchanging data with the server. This can be caused by a server crash
11903 or by a network issue on an intermediate equipment.
11904
11905 sD The server did not send nor acknowledge any data for as long as the
11906 "timeout server" setting during the data phase. This is often caused
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011907 by too short timeouts on L4 equipments before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011908 load-balancers, ...), as well as keep-alive sessions maintained
11909 between the client and the server expiring first on haproxy.
11910
11911 SH The server aborted before sending its full HTTP response headers, or
11912 it crashed while processing the request. Since a server aborting at
11913 this moment is very rare, it would be wise to inspect its logs to
11914 control whether it crashed and why. The logged request may indicate a
11915 small set of faulty requests, demonstrating bugs in the application.
11916 Sometimes this might also be caused by an IDS killing the connection
11917 between haproxy and the server.
11918
11919 sH The "timeout server" stroke before the server could return its
11920 response headers. This is the most common anomaly, indicating too
11921 long transactions, probably caused by server or database saturation.
11922 The immediate workaround consists in increasing the "timeout server"
11923 setting, but it is important to keep in mind that the user experience
11924 will suffer from these long response times. The only long term
11925 solution is to fix the application.
11926
11927 sQ The session spent too much time in queue and has been expired. See
11928 the "timeout queue" and "timeout connect" settings to find out how to
11929 fix this if it happens too often. If it often happens massively in
11930 short periods, it may indicate general problems on the affected
11931 servers due to I/O or database congestion, or saturation caused by
11932 external attacks.
11933
11934 PC The proxy refused to establish a connection to the server because the
11935 process' socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020011936 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011937 so that it does not happen anymore. This status is very rare and
11938 might happen when the global "ulimit-n" parameter is forced by hand.
11939
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010011940 PD The proxy blocked an incorrectly formatted chunked encoded message in
11941 a request or a response, after the server has emitted its headers. In
11942 most cases, this will indicate an invalid message from the server to
Willy Tarreauf3a3e132013-08-31 08:16:26 +020011943 the client. Haproxy supports chunk sizes of up to 2GB - 1 (2147483647
11944 bytes). Any larger size will be considered as an error.
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010011945
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011946 PH The proxy blocked the server's response, because it was invalid,
11947 incomplete, dangerous (cache control), or matched a security filter.
11948 In any case, an HTTP 502 error is sent to the client. One possible
11949 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010011950 containing unauthorized characters. It is also possible but quite
11951 rare, that the proxy blocked a chunked-encoding request from the
11952 client due to an invalid syntax, before the server responded. In this
11953 case, an HTTP 400 error is sent to the client and reported in the
11954 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010011955
11956 PR The proxy blocked the client's HTTP request, either because of an
11957 invalid HTTP syntax, in which case it returned an HTTP 400 error to
11958 the client, or because a deny filter matched, in which case it
11959 returned an HTTP 403 error.
11960
11961 PT The proxy blocked the client's request and has tarpitted its
11962 connection before returning it a 500 server error. Nothing was sent
11963 to the server. The connection was maintained open for as long as
11964 reported by the "Tw" timer field.
11965
11966 RC A local resource has been exhausted (memory, sockets, source ports)
11967 preventing the connection to the server from establishing. The error
11968 logs will tell precisely what was missing. This is very rare and can
11969 only be solved by proper system tuning.
11970
Willy Tarreau996a92c2010-10-13 19:30:47 +020011971The combination of the two last flags gives a lot of information about how
11972persistence was handled by the client, the server and by haproxy. This is very
11973important to troubleshoot disconnections, when users complain they have to
11974re-authenticate. The commonly encountered flags are :
11975
11976 -- Persistence cookie is not enabled.
11977
11978 NN No cookie was provided by the client, none was inserted in the
11979 response. For instance, this can be in insert mode with "postonly"
11980 set on a GET request.
11981
11982 II A cookie designating an invalid server was provided by the client,
11983 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040011984 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020011985 value can be presented by a client when no other server knows it.
11986
11987 NI No cookie was provided by the client, one was inserted in the
11988 response. This typically happens for first requests from every user
11989 in "insert" mode, which makes it an easy way to count real users.
11990
11991 VN A cookie was provided by the client, none was inserted in the
11992 response. This happens for most responses for which the client has
11993 already got a cookie.
11994
11995 VU A cookie was provided by the client, with a last visit date which is
11996 not completely up-to-date, so an updated cookie was provided in
11997 response. This can also happen if there was no date at all, or if
11998 there was a date but the "maxidle" parameter was not set, so that the
11999 cookie can be switched to unlimited time.
12000
12001 EI A cookie was provided by the client, with a last visit date which is
12002 too old for the "maxidle" parameter, so the cookie was ignored and a
12003 new cookie was inserted in the response.
12004
12005 OI A cookie was provided by the client, with a first visit date which is
12006 too old for the "maxlife" parameter, so the cookie was ignored and a
12007 new cookie was inserted in the response.
12008
12009 DI The server designated by the cookie was down, a new server was
12010 selected and a new cookie was emitted in the response.
12011
12012 VI The server designated by the cookie was not marked dead but could not
12013 be reached. A redispatch happened and selected another one, which was
12014 then advertised in the response.
12015
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012016
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200120178.6. Non-printable characters
12018-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012019
12020In order not to cause trouble to log analysis tools or terminals during log
12021consulting, non-printable characters are not sent as-is into log files, but are
12022converted to the two-digits hexadecimal representation of their ASCII code,
12023prefixed by the character '#'. The only characters that can be logged without
12024being escaped are comprised between 32 and 126 (inclusive). Obviously, the
12025escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
12026is the same for the character '"' which becomes "#22", as well as '{', '|' and
12027'}' when logging headers.
12028
12029Note that the space character (' ') is not encoded in headers, which can cause
12030issues for tools relying on space count to locate fields. A typical header
12031containing spaces is "User-Agent".
12032
12033Last, it has been observed that some syslog daemons such as syslog-ng escape
12034the quote ('"') with a backslash ('\'). The reverse operation can safely be
12035performed since no quote may appear anywhere else in the logs.
12036
12037
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200120388.7. Capturing HTTP cookies
12039---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012040
12041Cookie capture simplifies the tracking a complete user session. This can be
12042achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012043section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012044cookie will simultaneously be checked in the request ("Cookie:" header) and in
12045the response ("Set-Cookie:" header). The respective values will be reported in
12046the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012047locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012048not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
12049user switches to a new session for example, because the server will reassign it
12050a new cookie. It is also possible to detect if a server unexpectedly sets a
12051wrong cookie to a client, leading to session crossing.
12052
12053 Examples :
12054 # capture the first cookie whose name starts with "ASPSESSION"
12055 capture cookie ASPSESSION len 32
12056
12057 # capture the first cookie whose name is exactly "vgnvisitor"
12058 capture cookie vgnvisitor= len 32
12059
12060
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200120618.8. Capturing HTTP headers
12062---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012063
12064Header captures are useful to track unique request identifiers set by an upper
12065proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
12066the response, one can search for information about the response length, how the
12067server asked the cache to behave, or an object location during a redirection.
12068
12069Header captures are performed using the "capture request header" and "capture
12070response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012071section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012072
12073It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012074time. Non-existent headers are logged as empty strings, and if one header
12075appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012076are grouped within braces '{' and '}' in the same order as they were declared,
12077and delimited with a vertical bar '|' without any space. Response headers
12078follow the same representation, but are displayed after a space following the
12079request headers block. These blocks are displayed just before the HTTP request
12080in the logs.
12081
12082 Example :
12083 # This instance chains to the outgoing proxy
12084 listen proxy-out
12085 mode http
12086 option httplog
12087 option logasap
12088 log global
12089 server cache1 192.168.1.1:3128
12090
12091 # log the name of the virtual server
12092 capture request header Host len 20
12093
12094 # log the amount of data uploaded during a POST
12095 capture request header Content-Length len 10
12096
12097 # log the beginning of the referrer
12098 capture request header Referer len 20
12099
12100 # server name (useful for outgoing proxies only)
12101 capture response header Server len 20
12102
12103 # logging the content-length is useful with "option logasap"
12104 capture response header Content-Length len 10
12105
12106 # log the expected cache behaviour on the response
12107 capture response header Cache-Control len 8
12108
12109 # the Via header will report the next proxy's name
12110 capture response header Via len 20
12111
12112 # log the URL location during a redirection
12113 capture response header Location len 20
12114
12115 >>> Aug 9 20:26:09 localhost \
12116 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
12117 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
12118 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
12119 "GET http://fr.adserver.yahoo.com/"
12120
12121 >>> Aug 9 20:30:46 localhost \
12122 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
12123 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
12124 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010012125 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012126
12127 >>> Aug 9 20:30:46 localhost \
12128 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
12129 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
12130 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
12131 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010012132 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012133
12134
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200121358.9. Examples of logs
12136---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012137
12138These are real-world examples of logs accompanied with an explanation. Some of
12139them have been made up by hand. The syslog part has been removed for better
12140reading. Their sole purpose is to explain how to decipher them.
12141
12142 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
12143 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
12144 "HEAD / HTTP/1.0"
12145
12146 => long request (6.5s) entered by hand through 'telnet'. The server replied
12147 in 147 ms, and the session ended normally ('----')
12148
12149 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
12150 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
12151 0/9 "HEAD / HTTP/1.0"
12152
12153 => Idem, but the request was queued in the global queue behind 9 other
12154 requests, and waited there for 1230 ms.
12155
12156 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
12157 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
12158 "GET /image.iso HTTP/1.0"
12159
12160 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012161 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012162 14 ms, 243 bytes of headers were sent to the client, and total time from
12163 accept to first data byte is 30 ms.
12164
12165 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
12166 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
12167 "GET /cgi-bin/bug.cgi? HTTP/1.0"
12168
12169 => the proxy blocked a server response either because of an "rspdeny" or
12170 "rspideny" filter, or because the response was improperly formatted and
Willy Tarreau3c92c5f2011-08-28 09:45:47 +020012171 not HTTP-compliant, or because it blocked sensitive information which
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012172 risked being cached. In this case, the response is replaced with a "502
12173 bad gateway". The flags ("PH--") tell us that it was haproxy who decided
12174 to return the 502 and not the server.
12175
12176 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010012177 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 +010012178
12179 => the client never completed its request and aborted itself ("C---") after
12180 8.5s, while the proxy was waiting for the request headers ("-R--").
12181 Nothing was sent to any server.
12182
12183 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
12184 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
12185
12186 => The client never completed its request, which was aborted by the
12187 time-out ("c---") after 50s, while the proxy was waiting for the request
12188 headers ("-R--"). Nothing was sent to any server, but the proxy could
12189 send a 408 return code to the client.
12190
12191 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
12192 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
12193
12194 => This log was produced with "option tcplog". The client timed out after
12195 5 seconds ("c----").
12196
12197 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
12198 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010012199 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012200
12201 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012202 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012203 (config says 'retries 3'), and no redispatch (otherwise we would have
12204 seen "/+3"). Status code 503 was returned to the client. There were 115
12205 connections on this server, 202 connections on this proxy, and 205 on
12206 the global process. It is possible that the server refused the
12207 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010012208
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010012209
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200122109. Statistics and monitoring
12211----------------------------
12212
12213It is possible to query HAProxy about its status. The most commonly used
12214mechanism is the HTTP statistics page. This page also exposes an alternative
12215CSV output format for monitoring tools. The same format is provided on the
12216Unix socket.
12217
12218
122199.1. CSV format
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010012220---------------
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +010012221
Willy Tarreau7f062c42009-03-05 18:43:00 +010012222The statistics may be consulted either from the unix socket or from the HTTP
12223page. Both means provide a CSV format whose fields follow.
12224
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +010012225 0. pxname: proxy name
12226 1. svname: service name (FRONTEND for frontend, BACKEND for backend, any name
12227 for server)
12228 2. qcur: current queued requests
12229 3. qmax: max queued requests
12230 4. scur: current sessions
12231 5. smax: max sessions
12232 6. slim: sessions limit
12233 7. stot: total sessions
12234 8. bin: bytes in
12235 9. bout: bytes out
12236 10. dreq: denied requests
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010012237 11. dresp: denied responses
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +010012238 12. ereq: request errors
12239 13. econ: connection errors
Willy Tarreauae526782010-03-04 20:34:23 +010012240 14. eresp: response errors (among which srv_abrt)
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +010012241 15. wretr: retries (warning)
12242 16. wredis: redispatches (warning)
Cyril Bonté0dae5852010-02-03 00:26:28 +010012243 17. status: status (UP/DOWN/NOLB/MAINT/MAINT(via)...)
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +010012244 18. weight: server weight (server), total weight (backend)
12245 19. act: server is active (server), number of active servers (backend)
12246 20. bck: server is backup (server), number of backup servers (backend)
12247 21. chkfail: number of failed checks
12248 22. chkdown: number of UP->DOWN transitions
12249 23. lastchg: last status change (in seconds)
12250 24. downtime: total downtime (in seconds)
12251 25. qlimit: queue limit
12252 26. pid: process id (0 for first instance, 1 for second, ...)
12253 27. iid: unique proxy id
12254 28. sid: service id (unique inside a proxy)
12255 29. throttle: warm up status
12256 30. lbtot: total number of times a server was selected
12257 31. tracked: id of proxy/server if tracking is enabled
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +020012258 32. type (0=frontend, 1=backend, 2=server, 3=socket)
Krzysztof Piotr Oledzkidb57c6b2009-08-31 21:23:27 +020012259 33. rate: number of sessions per second over last elapsed second
12260 34. rate_lim: limit on new sessions per second
12261 35. rate_max: max number of new sessions per second
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +020012262 36. check_status: status of last health check, one of:
Cyril Bontéf0c60612010-02-06 14:44:47 +010012263 UNK -> unknown
12264 INI -> initializing
12265 SOCKERR -> socket error
12266 L4OK -> check passed on layer 4, no upper layers testing enabled
12267 L4TMOUT -> layer 1-4 timeout
12268 L4CON -> layer 1-4 connection problem, for example
12269 "Connection refused" (tcp rst) or "No route to host" (icmp)
12270 L6OK -> check passed on layer 6
12271 L6TOUT -> layer 6 (SSL) timeout
12272 L6RSP -> layer 6 invalid response - protocol error
12273 L7OK -> check passed on layer 7
12274 L7OKC -> check conditionally passed on layer 7, for example 404 with
12275 disable-on-404
12276 L7TOUT -> layer 7 (HTTP/SMTP) timeout
12277 L7RSP -> layer 7 invalid response - protocol error
12278 L7STS -> layer 7 response error, for example HTTP 5xx
Krzysztof Piotr Oledzki09605412009-09-23 22:09:24 +020012279 37. check_code: layer5-7 code, if available
12280 38. check_duration: time in ms took to finish last health check
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012281 39. hrsp_1xx: http responses with 1xx code
12282 40. hrsp_2xx: http responses with 2xx code
12283 41. hrsp_3xx: http responses with 3xx code
12284 42. hrsp_4xx: http responses with 4xx code
12285 43. hrsp_5xx: http responses with 5xx code
12286 44. hrsp_other: http responses with other codes (protocol error)
Willy Tarreaud63335a2010-02-26 12:56:52 +010012287 45. hanafail: failed health checks details
12288 46. req_rate: HTTP requests per second over last elapsed second
12289 47. req_rate_max: max number of HTTP requests per second observed
12290 48. req_tot: total number of HTTP requests received
Willy Tarreauae526782010-03-04 20:34:23 +010012291 49. cli_abrt: number of data transfers aborted by the client
12292 50. srv_abrt: number of data transfers aborted by the server (inc. in eresp)
Willy Tarreau55058a72012-11-21 08:27:21 +010012293 51. comp_in: number of HTTP response bytes fed to the compressor
12294 52. comp_out: number of HTTP response bytes emitted by the compressor
12295 53. comp_byp: number of bytes that bypassed the HTTP compressor (CPU/BW limit)
Willy Tarreau11d4ec82012-11-26 00:49:03 +010012296 54. comp_rsp: number of HTTP responses that were compressed
Willy Tarreau844e3c52008-01-11 16:28:18 +010012297
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010012298
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200122999.2. Unix Socket commands
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010012300-------------------------
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010012301
Willy Tarreau468f4932013-08-01 16:50:16 +020012302The stats socket is not enabled by default. In order to enable it, it is
12303necessary to add one line in the global section of the haproxy configuration.
12304A second line is recommended to set a larger timeout, always appreciated when
12305issuing commands by hand :
12306
12307 global
12308 stats socket /var/run/haproxy.sock mode 600 level admin
12309 stats timeout 2m
12310
12311It is also possible to add multiple instances of the stats socket by repeating
12312the line, and make them listen to a TCP port instead of a UNIX socket. This is
12313never done by default because this is dangerous, but can be handy in some
12314situations :
12315
12316 global
12317 stats socket /var/run/haproxy.sock mode 600 level admin
12318 stats socket ipv4@192.168.0.1:9999 level admin
12319 stats timeout 2m
12320
12321To access the socket, an external utility such as "socat" is required. Socat is a
12322swiss-army knife to connect anything to anything. We use it to connect terminals
12323to the socket, or a couple of stdin/stdout pipes to it for scripts. The two main
12324syntaxes we'll use are the following :
12325
12326 # socat /var/run/haproxy.sock stdio
12327 # socat /var/run/haproxy.sock readline
12328
12329The first one is used with scripts. It is possible to send the output of a
12330script to haproxy, and pass haproxy's output to another script. That's useful
12331for retrieving counters or attack traces for example.
12332
12333The second one is only useful for issuing commands by hand. It has the benefit
12334that the terminal is handled by the readline library which supports line
12335editing and history, which is very convenient when issuing repeated commands
12336(eg: watch a counter).
12337
12338The socket supports two operation modes :
12339 - interactive
12340 - non-interactive
12341
12342The non-interactive mode is the default when socat connects to the socket. In
12343this mode, a single line may be sent. It is processed as a whole, responses are
12344sent back, and the connection closes after the end of the response. This is the
12345mode that scripts and monitoring tools use. It is possible to send multiple
12346commands in this mode, they need to be delimited by a semi-colon (';'). For
12347example :
12348
12349 # echo "show info;show stat;show table" | socat /var/run/haproxy stdio
12350
12351The interactive mode displays a prompt ('>') and waits for commands to be
12352entered on the line, then processes them, and displays the prompt again to wait
12353for a new command. This mode is entered via the "prompt" command which must be
12354sent on the first line in non-interactive mode. The mode is a flip switch, if
12355"prompt" is sent in interactive mode, it is disabled and the connection closes
12356after processing the last command of the same line.
12357
12358For this reason, when debugging by hand, it's quite common to start with the
12359"prompt" command :
12360
12361 # socat /var/run/haproxy readline
12362 prompt
12363 > show info
12364 ...
12365 >
12366
12367Since multiple commands may be issued at once, haproxy uses the empty line as a
12368delimiter to mark an end of output for each command, and takes care of ensuring
12369that no command can emit an empty line on output. A script can thus easily
12370parse the output even when multiple commands were pipelined on a single line.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010012371
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020012372It is important to understand that when multiple haproxy processes are started
12373on the same sockets, any process may pick up the request and will output its
12374own stats.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010012375
Willy Tarreau468f4932013-08-01 16:50:16 +020012376The list of commands currently supported on the stats socket is provided below.
12377If an unknown command is sent, haproxy displays the usage message which reminds
12378all supported commands. Some commands support a more complex syntax, generally
12379it will explain what part of the command is invalid when this happens.
12380
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010012381add map <map> <key> <value>
12382 Add an entry into the map <map> to associate the value <value> to the key
12383 <key>. This command does not verify if the entry already exists. It is
12384 mainly used to fill a map after a clear operation.
12385
Willy Tarreaud63335a2010-02-26 12:56:52 +010012386clear counters
12387 Clear the max values of the statistics counters in each proxy (frontend &
12388 backend) and in each server. The cumulated counters are not affected. This
12389 can be used to get clean counters after an incident, without having to
12390 restart nor to clear traffic counters. This command is restricted and can
12391 only be issued on sockets configured for levels "operator" or "admin".
12392
12393clear counters all
12394 Clear all statistics counters in each proxy (frontend & backend) and in each
12395 server. This has the same effect as restarting. This command is restricted
12396 and can only be issued on sockets configured for level "admin".
12397
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010012398clear map <map>
12399 Remove all entries from the map <map>.
12400
Simon Hormanc88b8872011-06-15 15:18:49 +090012401clear table <table> [ data.<type> <operator> <value> ] | [ key <key> ]
12402 Remove entries from the stick-table <table>.
12403
12404 This is typically used to unblock some users complaining they have been
12405 abusively denied access to a service, but this can also be used to clear some
12406 stickiness entries matching a server that is going to be replaced (see "show
12407 table" below for details). Note that sometimes, removal of an entry will be
12408 refused because it is currently tracked by a session. Retrying a few seconds
12409 later after the session ends is usual enough.
12410
12411 In the case where no options arguments are given all entries will be removed.
12412
12413 When the "data." form is used entries matching a filter applied using the
12414 stored data (see "stick-table" in section 4.2) are removed. A stored data
12415 type must be specified in <type>, and this data type must be stored in the
12416 table otherwise an error is reported. The data is compared according to
12417 <operator> with the 64-bit integer <value>. Operators are the same as with
12418 the ACLs :
12419
12420 - eq : match entries whose data is equal to this value
12421 - ne : match entries whose data is not equal to this value
12422 - le : match entries whose data is less than or equal to this value
12423 - ge : match entries whose data is greater than or equal to this value
12424 - lt : match entries whose data is less than this value
12425 - gt : match entries whose data is greater than this value
12426
12427 When the key form is used the entry <key> is removed. The key must be of the
Simon Horman619e3cc2011-06-15 15:18:52 +090012428 same type as the table, which currently is limited to IPv4, IPv6, integer and
12429 string.
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012430
12431 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020012432 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Emeric Brun7c6b82e2010-09-24 16:34:28 +020012433 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020012434 >>> 0x80e6a4c: key=127.0.0.1 use=0 exp=3594729 gpc0=0 conn_rate(30000)=1 \
12435 bytes_out_rate(60000)=187
12436 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
12437 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012438
12439 $ echo "clear table http_proxy key 127.0.0.1" | socat stdio /tmp/sock1
12440
12441 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Emeric Brun7c6b82e2010-09-24 16:34:28 +020012442 >>> # table: http_proxy, type: ip, size:204800, used:1
Willy Tarreau62a36c42010-08-17 15:53:10 +020012443 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
12444 bytes_out_rate(60000)=191
Simon Hormanc88b8872011-06-15 15:18:49 +090012445 $ echo "clear table http_proxy data.gpc0 eq 1" | socat stdio /tmp/sock1
12446 $ echo "show table http_proxy" | socat stdio /tmp/sock1
12447 >>> # table: http_proxy, type: ip, size:204800, used:1
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012448
Simon Horman671b6f02013-11-25 10:46:39 +090012449enable agent <backend>/<server>
12450 Mark the auxiliary agent check as temporarily stopped.
12451
12452 In the case where an agent check is being run as a auxiliary check, due
12453 to the agent-check parameter of a server directive, new checks are only
12454 initialised when the agent is in the enabled. Thus, disable agent will
12455 prevent any new agent checks from begin initiated until the agent
12456 re-enabled using enable agent.
12457
12458 When an agent is disabled the processing of an auxiliary agent check that
12459 was initiated while the agent was set as enabled is as follows: All
12460 results that would alter the weight, specifically "drain" or a weight
12461 returned by the agent, are ignored. The processing of agent check is
12462 otherwise unchanged.
12463
12464 The motivation for this feature is to allow the weight changing effects
12465 of the agent checks to be paused to allow the weight of a server to be
12466 configured using set weight without being overridden by the agent.
12467
12468 This command is restricted and can only be issued on sockets configured for
12469 level "admin".
12470
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010012471del map <map> <key>
12472 Delete all the map entries from the map <map> corresponding to the key <key>.
12473
Willy Tarreau532a4502011-09-07 22:37:44 +020012474disable frontend <frontend>
12475 Mark the frontend as temporarily stopped. This corresponds to the mode which
12476 is used during a soft restart : the frontend releases the port but can be
12477 enabled again if needed. This should be used with care as some non-Linux OSes
12478 are unable to enable it back. This is intended to be used in environments
12479 where stopping a proxy is not even imaginable but a misconfigured proxy must
12480 be fixed. That way it's possible to release the port and bind it into another
12481 process to restore operations. The frontend will appear with status "STOP"
12482 on the stats page.
12483
12484 The frontend may be specified either by its name or by its numeric ID,
12485 prefixed with a sharp ('#').
12486
12487 This command is restricted and can only be issued on sockets configured for
12488 level "admin".
12489
Willy Tarreaud63335a2010-02-26 12:56:52 +010012490disable server <backend>/<server>
12491 Mark the server DOWN for maintenance. In this mode, no more checks will be
12492 performed on the server until it leaves maintenance.
12493 If the server is tracked by other servers, those servers will be set to DOWN
12494 during the maintenance.
12495
12496 In the statistics page, a server DOWN for maintenance will appear with a
12497 "MAINT" status, its tracking servers with the "MAINT(via)" one.
12498
12499 Both the backend and the server may be specified either by their name or by
Willy Tarreauf5f31922011-08-02 11:32:07 +020012500 their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010012501
12502 This command is restricted and can only be issued on sockets configured for
12503 level "admin".
12504
Simon Horman671b6f02013-11-25 10:46:39 +090012505enable agent <backend>/<server>
12506 Resume auxiliary agent check that was temporarily stopped.
12507
12508 See "disable agent" for details of the effect of temporarily starting
12509 and stopping an auxiliary agent.
12510
12511 This command is restricted and can only be issued on sockets configured for
12512 level "admin".
12513
Willy Tarreau532a4502011-09-07 22:37:44 +020012514enable frontend <frontend>
12515 Resume a frontend which was temporarily stopped. It is possible that some of
12516 the listening ports won't be able to bind anymore (eg: if another process
12517 took them since the 'disable frontend' operation). If this happens, an error
12518 is displayed. Some operating systems might not be able to resume a frontend
12519 which was disabled.
12520
12521 The frontend may be specified either by its name or by its numeric ID,
12522 prefixed with a sharp ('#').
12523
12524 This command is restricted and can only be issued on sockets configured for
12525 level "admin".
12526
Willy Tarreaud63335a2010-02-26 12:56:52 +010012527enable server <backend>/<server>
12528 If the server was previously marked as DOWN for maintenance, this marks the
12529 server UP and checks are re-enabled.
12530
12531 Both the backend and the server may be specified either by their name or by
Willy Tarreauf5f31922011-08-02 11:32:07 +020012532 their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010012533
12534 This command is restricted and can only be issued on sockets configured for
12535 level "admin".
12536
12537get weight <backend>/<server>
12538 Report the current weight and the initial weight of server <server> in
12539 backend <backend> or an error if either doesn't exist. The initial weight is
12540 the one that appears in the configuration file. Both are normally equal
12541 unless the current weight has been changed. Both the backend and the server
12542 may be specified either by their name or by their numeric ID, prefixed with a
Willy Tarreauf5f31922011-08-02 11:32:07 +020012543 sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010012544
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020012545help
12546 Print the list of known keywords and their basic usage. The same help screen
12547 is also displayed for unknown commands.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010012548
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020012549prompt
12550 Toggle the prompt at the beginning of the line and enter or leave interactive
12551 mode. In interactive mode, the connection is not closed after a command
12552 completes. Instead, the prompt will appear again, indicating the user that
12553 the interpreter is waiting for a new command. The prompt consists in a right
12554 angle bracket followed by a space "> ". This mode is particularly convenient
12555 when one wants to periodically check information such as stats or errors.
12556 It is also a good idea to enter interactive mode before issuing a "help"
12557 command.
12558
12559quit
12560 Close the connection when in interactive mode.
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010012561
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010012562set map <map> <key> <value>
12563 Modify the value corresponding to each key <key> in a map <map>. The new value
12564 is <value>.
12565
Willy Tarreau2a0f4d22011-08-02 11:49:05 +020012566set maxconn frontend <frontend> <value>
Willy Tarreau3c7a79d2012-09-26 21:07:15 +020012567 Dynamically change the specified frontend's maxconn setting. Any positive
12568 value is allowed including zero, but setting values larger than the global
12569 maxconn does not make much sense. If the limit is increased and connections
12570 were pending, they will immediately be accepted. If it is lowered to a value
12571 below the current number of connections, new connections acceptation will be
Willy Tarreau2a0f4d22011-08-02 11:49:05 +020012572 delayed until the threshold is reached. The frontend might be specified by
12573 either its name or its numeric ID prefixed with a sharp ('#').
12574
Willy Tarreau91886b62011-09-07 14:38:31 +020012575set maxconn global <maxconn>
12576 Dynamically change the global maxconn setting within the range defined by the
12577 initial global maxconn setting. If it is increased and connections were
12578 pending, they will immediately be accepted. If it is lowered to a value below
12579 the current number of connections, new connections acceptation will be
12580 delayed until the threshold is reached. A value of zero restores the initial
12581 setting.
12582
Willy Tarreauf5b22872011-09-07 16:13:44 +020012583set rate-limit connections global <value>
12584 Change the process-wide connection rate limit, which is set by the global
12585 'maxconnrate' setting. A value of zero disables the limitation. This limit
12586 applies to all frontends and the change has an immediate effect. The value
12587 is passed in number of connections per second.
12588
William Lallemandd85f9172012-11-09 17:05:39 +010012589set rate-limit http-compression global <value>
12590 Change the maximum input compression rate, which is set by the global
12591 'maxcomprate' setting. A value of zero disables the limitation. The value is
William Lallemand096f5542012-11-19 17:26:05 +010012592 passed in number of kilobytes per second. The value is available in the "show
12593 info" on the line "CompressBpsRateLim" in bytes.
William Lallemandd85f9172012-11-09 17:05:39 +010012594
Willy Tarreau47060b62013-08-01 21:11:42 +020012595set table <table> key <key> [data.<data_type> <value>]*
Willy Tarreau654694e2012-06-07 01:03:16 +020012596 Create or update a stick-table entry in the table. If the key is not present,
12597 an entry is inserted. See stick-table in section 4.2 to find all possible
12598 values for <data_type>. The most likely use consists in dynamically entering
12599 entries for source IP addresses, with a flag in gpc0 to dynamically block an
Willy Tarreau47060b62013-08-01 21:11:42 +020012600 IP address or affect its quality of service. It is possible to pass multiple
12601 data_types in a single call.
Willy Tarreau654694e2012-06-07 01:03:16 +020012602
Willy Tarreaud63335a2010-02-26 12:56:52 +010012603set timeout cli <delay>
12604 Change the CLI interface timeout for current connection. This can be useful
12605 during long debugging sessions where the user needs to constantly inspect
12606 some indicators without being disconnected. The delay is passed in seconds.
12607
12608set weight <backend>/<server> <weight>[%]
12609 Change a server's weight to the value passed in argument. If the value ends
12610 with the '%' sign, then the new weight will be relative to the initially
Simon Horman58b5d292013-02-12 10:45:52 +090012611 configured weight. Absolute weights are permitted between 0 and 256.
12612 Relative weights must be positive with the resulting absolute weight is
12613 capped at 256. Servers which are part of a farm running a static
12614 load-balancing algorithm have stricter limitations because the weight
12615 cannot change once set. Thus for these servers, the only accepted values
12616 are 0 and 100% (or 0 and the initial weight). Changes take effect
12617 immediately, though certain LB algorithms require a certain amount of
12618 requests to consider changes. A typical usage of this command is to
12619 disable a server during an update by setting its weight to zero, then to
12620 enable it again after the update by setting it back to 100%. This command
12621 is restricted and can only be issued on sockets configured for level
12622 "admin". Both the backend and the server may be specified either by their
12623 name or by their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010012624
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010012625show errors [<iid>]
12626 Dump last known request and response errors collected by frontends and
12627 backends. If <iid> is specified, the limit the dump to errors concerning
Willy Tarreau6162db22009-10-10 17:13:00 +020012628 either frontend or backend whose ID is <iid>. This command is restricted
12629 and can only be issued on sockets configured for levels "operator" or
12630 "admin".
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010012631
12632 The errors which may be collected are the last request and response errors
12633 caused by protocol violations, often due to invalid characters in header
12634 names. The report precisely indicates what exact character violated the
12635 protocol. Other important information such as the exact date the error was
12636 detected, frontend and backend names, the server name (when known), the
12637 internal session ID and the source address which has initiated the session
12638 are reported too.
12639
12640 All characters are returned, and non-printable characters are encoded. The
12641 most common ones (\t = 9, \n = 10, \r = 13 and \e = 27) are encoded as one
12642 letter following a backslash. The backslash itself is encoded as '\\' to
12643 avoid confusion. Other non-printable characters are encoded '\xNN' where
12644 NN is the two-digits hexadecimal representation of the character's ASCII
12645 code.
12646
12647 Lines are prefixed with the position of their first character, starting at 0
12648 for the beginning of the buffer. At most one input line is printed per line,
12649 and large lines will be broken into multiple consecutive output lines so that
12650 the output never goes beyond 79 characters wide. It is easy to detect if a
12651 line was broken, because it will not end with '\n' and the next line's offset
12652 will be followed by a '+' sign, indicating it is a continuation of previous
12653 line.
12654
12655 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020012656 $ echo "show errors" | socat stdio /tmp/sock1
12657 >>> [04/Mar/2009:15:46:56.081] backend http-in (#2) : invalid response
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010012658 src 127.0.0.1, session #54, frontend fe-eth0 (#1), server s2 (#1)
12659 response length 213 bytes, error at position 23:
12660
12661 00000 HTTP/1.0 200 OK\r\n
12662 00017 header/bizarre:blah\r\n
12663 00038 Location: blah\r\n
12664 00054 Long-line: this is a very long line which should b
12665 00104+ e broken into multiple lines on the output buffer,
12666 00154+ otherwise it would be too large to print in a ter
12667 00204+ minal\r\n
12668 00211 \r\n
12669
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012670 In the example above, we see that the backend "http-in" which has internal
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010012671 ID 2 has blocked an invalid response from its server s2 which has internal
12672 ID 1. The request was on session 54 initiated by source 127.0.0.1 and
12673 received by frontend fe-eth0 whose ID is 1. The total response length was
12674 213 bytes when the error was detected, and the error was at byte 23. This
12675 is the slash ('/') in header name "header/bizarre", which is not a valid
12676 HTTP character for a header name.
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010012677
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020012678show info
12679 Dump info about haproxy status on current process.
12680
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010012681show map [<map>]
12682 Dump info about map converters. Without argument, the list of all available
12683 maps is returned. If a <map> is specified, its contents are dumped.
12684
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020012685show sess
12686 Dump all known sessions. Avoid doing this on slow connections as this can
Willy Tarreau6162db22009-10-10 17:13:00 +020012687 be huge. This command is restricted and can only be issued on sockets
12688 configured for levels "operator" or "admin".
12689
Willy Tarreau66dc20a2010-03-05 17:53:32 +010012690show sess <id>
12691 Display a lot of internal information about the specified session identifier.
12692 This identifier is the first field at the beginning of the lines in the dumps
12693 of "show sess" (it corresponds to the session pointer). Those information are
12694 useless to most users but may be used by haproxy developers to troubleshoot a
12695 complex bug. The output format is intentionally not documented so that it can
Willy Tarreau76153662012-11-26 01:16:39 +010012696 freely evolve depending on demands. The special id "all" dumps the states of
12697 all sessions, which can be avoided as much as possible as it is highly CPU
12698 intensive and can take a lot of time.
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020012699
12700show stat [<iid> <type> <sid>]
12701 Dump statistics in the CSV format. By passing <id>, <type> and <sid>, it is
12702 possible to dump only selected items :
12703 - <iid> is a proxy ID, -1 to dump everything
12704 - <type> selects the type of dumpable objects : 1 for frontends, 2 for
12705 backends, 4 for servers, -1 for everything. These values can be ORed,
12706 for example:
12707 1 + 2 = 3 -> frontend + backend.
12708 1 + 2 + 4 = 7 -> frontend + backend + server.
12709 - <sid> is a server ID, -1 to dump everything from the selected proxy.
12710
12711 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020012712 $ echo "show info;show stat" | socat stdio unix-connect:/tmp/sock1
12713 >>> Name: HAProxy
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020012714 Version: 1.4-dev2-49
12715 Release_date: 2009/09/23
12716 Nbproc: 1
12717 Process_num: 1
12718 (...)
12719
12720 # pxname,svname,qcur,qmax,scur,smax,slim,stot,bin,bout,dreq, (...)
12721 stats,FRONTEND,,,0,0,1000,0,0,0,0,0,0,,,,,OPEN,,,,,,,,,1,1,0, (...)
12722 stats,BACKEND,0,0,0,0,1000,0,0,0,0,0,,0,0,0,0,UP,0,0,0,,0,250,(...)
12723 (...)
12724 www1,BACKEND,0,0,0,0,1000,0,0,0,0,0,,0,0,0,0,UP,1,1,0,,0,250, (...)
12725
12726 $
12727
12728 Here, two commands have been issued at once. That way it's easy to find
12729 which process the stats apply to in multi-process mode. Notice the empty
12730 line after the information output which marks the end of the first block.
12731 A similar empty line appears at the end of the second block (stats) so that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012732 the reader knows the output has not been truncated.
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020012733
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012734show table
12735 Dump general information on all known stick-tables. Their name is returned
12736 (the name of the proxy which holds them), their type (currently zero, always
12737 IP), their size in maximum possible number of entries, and the number of
12738 entries currently in use.
12739
12740 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020012741 $ echo "show table" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090012742 >>> # table: front_pub, type: ip, size:204800, used:171454
12743 >>> # table: back_rdp, type: ip, size:204800, used:0
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012744
Simon Horman17bce342011-06-15 15:18:47 +090012745show table <name> [ data.<type> <operator> <value> ] | [ key <key> ]
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012746 Dump contents of stick-table <name>. In this mode, a first line of generic
12747 information about the table is reported as with "show table", then all
12748 entries are dumped. Since this can be quite heavy, it is possible to specify
Simon Horman17bce342011-06-15 15:18:47 +090012749 a filter in order to specify what entries to display.
12750
12751 When the "data." form is used the filter applies to the stored data (see
12752 "stick-table" in section 4.2). A stored data type must be specified
12753 in <type>, and this data type must be stored in the table otherwise an
12754 error is reported. The data is compared according to <operator> with the
12755 64-bit integer <value>. Operators are the same as with the ACLs :
12756
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012757 - eq : match entries whose data is equal to this value
12758 - ne : match entries whose data is not equal to this value
12759 - le : match entries whose data is less than or equal to this value
12760 - ge : match entries whose data is greater than or equal to this value
12761 - lt : match entries whose data is less than this value
12762 - gt : match entries whose data is greater than this value
12763
Simon Hormanc88b8872011-06-15 15:18:49 +090012764
12765 When the key form is used the entry <key> is shown. The key must be of the
Simon Horman619e3cc2011-06-15 15:18:52 +090012766 same type as the table, which currently is limited to IPv4, IPv6, integer,
12767 and string.
Simon Horman17bce342011-06-15 15:18:47 +090012768
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012769 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020012770 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090012771 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020012772 >>> 0x80e6a4c: key=127.0.0.1 use=0 exp=3594729 gpc0=0 conn_rate(30000)=1 \
12773 bytes_out_rate(60000)=187
12774 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
12775 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012776
Willy Tarreau62a36c42010-08-17 15:53:10 +020012777 $ echo "show table http_proxy data.gpc0 gt 0" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090012778 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020012779 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
12780 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012781
Willy Tarreau62a36c42010-08-17 15:53:10 +020012782 $ echo "show table http_proxy data.conn_rate gt 5" | \
12783 socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090012784 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020012785 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
12786 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012787
Simon Horman17bce342011-06-15 15:18:47 +090012788 $ echo "show table http_proxy key 127.0.0.2" | \
12789 socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090012790 >>> # table: http_proxy, type: ip, size:204800, used:2
Simon Horman17bce342011-06-15 15:18:47 +090012791 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
12792 bytes_out_rate(60000)=191
12793
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012794 When the data criterion applies to a dynamic value dependent on time such as
12795 a bytes rate, the value is dynamically computed during the evaluation of the
12796 entry in order to decide whether it has to be dumped or not. This means that
12797 such a filter could match for some time then not match anymore because as
12798 time goes, the average event rate drops.
12799
12800 It is possible to use this to extract lists of IP addresses abusing the
12801 service, in order to monitor them or even blacklist them in a firewall.
12802 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020012803 $ echo "show table http_proxy data.gpc0 gt 0" \
12804 | socat stdio /tmp/sock1 \
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020012805 | fgrep 'key=' | cut -d' ' -f2 | cut -d= -f2 > abusers-ip.txt
12806 ( or | awk '/key/{ print a[split($2,a,"=")]; }' )
Krzysztof Piotr Oledzki719e7262009-10-04 15:02:46 +020012807
Willy Tarreau532a4502011-09-07 22:37:44 +020012808shutdown frontend <frontend>
12809 Completely delete the specified frontend. All the ports it was bound to will
12810 be released. It will not be possible to enable the frontend anymore after
12811 this operation. This is intended to be used in environments where stopping a
12812 proxy is not even imaginable but a misconfigured proxy must be fixed. That
12813 way it's possible to release the port and bind it into another process to
12814 restore operations. The frontend will not appear at all on the stats page
12815 once it is terminated.
12816
12817 The frontend may be specified either by its name or by its numeric ID,
12818 prefixed with a sharp ('#').
12819
12820 This command is restricted and can only be issued on sockets configured for
12821 level "admin".
12822
Willy Tarreaua295edc2011-09-07 23:21:03 +020012823shutdown session <id>
12824 Immediately terminate the session matching the specified session identifier.
12825 This identifier is the first field at the beginning of the lines in the dumps
12826 of "show sess" (it corresponds to the session pointer). This can be used to
12827 terminate a long-running session without waiting for a timeout or when an
12828 endless transfer is ongoing. Such terminated sessions are reported with a 'K'
12829 flag in the logs.
12830
Willy Tarreau52b2d222011-09-07 23:48:48 +020012831shutdown sessions <backend>/<server>
12832 Immediately terminate all the sessions attached to the specified server. This
12833 can be used to terminate long-running sessions after a server is put into
12834 maintenance mode, for instance. Such terminated sessions are reported with a
12835 'K' flag in the logs.
12836
Willy Tarreau0ba27502007-12-24 16:55:16 +010012837/*
12838 * Local variables:
12839 * fill-column: 79
12840 * End:
12841 */