blob: be5b26c085fb4a4e28c817104cb2a45a922ead08 [file] [log] [blame]
Willy Tarreau6a06a402007-07-15 20:15:28 +02001 ----------------------
Willy Tarreau8317b282014-04-23 01:49:41 +02002 HAProxy
Willy Tarreau6a06a402007-07-15 20:15:28 +02003 Configuration Manual
4 ----------------------
Willy Tarreau15480d72014-06-19 21:10:58 +02005 version 1.6
Willy Tarreau6a06a402007-07-15 20:15:28 +02006 willy tarreau
Willy Tarreaue7ae6562015-09-28 23:46:27 +02007 2015/09/28
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
William Lallemandf9873ba2015-05-05 17:37:14 +0200422.2. Quoting and escaping
William Lallemandb2f07452015-05-12 14:27:13 +0200432.3. Environment variables
442.4. Time format
452.5. Examples
Willy Tarreauc57f0e22009-05-10 13:12:33 +020046
473. Global parameters
483.1. Process management and security
493.2. Performance tuning
503.3. Debugging
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100513.4. Userlists
Cyril Bontédc4d9032012-04-08 21:57:39 +0200523.5. Peers
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200533.6. Mailers
Willy Tarreauc57f0e22009-05-10 13:12:33 +020054
554. Proxies
564.1. Proxy keywords matrix
574.2. Alphabetically sorted keywords reference
58
Willy Tarreau086fbf52012-09-24 20:34:51 +0200595. Bind and Server options
605.1. Bind options
615.2. Server and default-server options
Baptiste Assmann1fa66662015-04-14 00:28:47 +0200625.3. Server DNS resolution
635.3.1. Global overview
645.3.2. The resolvers section
Willy Tarreauc57f0e22009-05-10 13:12:33 +020065
666. HTTP header manipulation
67
Willy Tarreau74ca5042013-06-11 23:12:07 +0200687. Using ACLs and fetching samples
697.1. ACL basics
707.1.1. Matching booleans
717.1.2. Matching integers
727.1.3. Matching strings
737.1.4. Matching regular expressions (regexes)
747.1.5. Matching arbitrary data blocks
757.1.6. Matching IPv4 and IPv6 addresses
767.2. Using ACLs to form conditions
777.3. Fetching samples
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200787.3.1. Converters
797.3.2. Fetching samples from internal states
807.3.3. Fetching samples at Layer 4
817.3.4. Fetching samples at Layer 5
827.3.5. Fetching samples from buffer contents (Layer 6)
837.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +0200847.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020085
868. Logging
878.1. Log levels
888.2. Log formats
898.2.1. Default log format
908.2.2. TCP log format
918.2.3. HTTP log format
William Lallemand48940402012-01-30 16:47:22 +0100928.2.4. Custom log format
Willy Tarreau5f51e1a2012-12-03 18:40:10 +0100938.2.5. Error log format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200948.3. Advanced logging options
958.3.1. Disabling logging of external tests
968.3.2. Logging before waiting for the session to terminate
978.3.3. Raising log level upon errors
988.3.4. Disabling logging of successful connections
998.4. Timing events
1008.5. Session state at disconnection
1018.6. Non-printable characters
1028.7. Capturing HTTP cookies
1038.8. Capturing HTTP headers
1048.9. Examples of logs
105
1069. Statistics and monitoring
1079.1. CSV format
1089.2. Unix Socket commands
109
110
1111. Quick reminder about HTTP
112----------------------------
113
114When haproxy is running in HTTP mode, both the request and the response are
115fully analyzed and indexed, thus it becomes possible to build matching criteria
116on almost anything found in the contents.
117
118However, it is important to understand how HTTP requests and responses are
119formed, and how HAProxy decomposes them. It will then become easier to write
120correct rules and to debug existing configurations.
121
122
1231.1. The HTTP transaction model
124-------------------------------
125
126The HTTP protocol is transaction-driven. This means that each request will lead
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100127to one and only one response. Traditionally, a TCP connection is established
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200128from the client to the server, a request is sent by the client on the
129connection, the server responds and the connection is closed. A new request
130will involve a new connection :
131
132 [CON1] [REQ1] ... [RESP1] [CLO1] [CON2] [REQ2] ... [RESP2] [CLO2] ...
133
134In this mode, called the "HTTP close" mode, there are as many connection
135establishments as there are HTTP transactions. Since the connection is closed
136by the server after the response, the client does not need to know the content
137length.
138
139Due to the transactional nature of the protocol, it was possible to improve it
140to avoid closing a connection between two subsequent transactions. In this mode
141however, it is mandatory that the server indicates the content length for each
142response so that the client does not wait indefinitely. For this, a special
143header is used: "Content-length". This mode is called the "keep-alive" mode :
144
145 [CON] [REQ1] ... [RESP1] [REQ2] ... [RESP2] [CLO] ...
146
147Its advantages are a reduced latency between transactions, and less processing
148power required on the server side. It is generally better than the close mode,
149but not always because the clients often limit their concurrent connections to
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200150a smaller value.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200151
152A last improvement in the communications is the pipelining mode. It still uses
153keep-alive, but the client does not wait for the first response to send the
154second request. This is useful for fetching large number of images composing a
155page :
156
157 [CON] [REQ1] [REQ2] ... [RESP1] [RESP2] [CLO] ...
158
159This can obviously have a tremendous benefit on performance because the network
160latency is eliminated between subsequent requests. Many HTTP agents do not
161correctly support pipelining since there is no way to associate a response with
162the corresponding request in HTTP. For this reason, it is mandatory for the
Cyril Bonté78caf842010-03-10 22:41:43 +0100163server to reply in the exact same order as the requests were received.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200164
Willy Tarreau70dffda2014-01-30 03:07:23 +0100165By default HAProxy operates in keep-alive mode with regards to persistent
166connections: for each connection it processes each request and response, and
167leaves the connection idle on both sides between the end of a response and the
168start of a new request.
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200169
Willy Tarreau70dffda2014-01-30 03:07:23 +0100170HAProxy supports 5 connection modes :
171 - keep alive : all requests and responses are processed (default)
172 - tunnel : only the first request and response are processed,
173 everything else is forwarded with no analysis.
174 - passive close : tunnel with "Connection: close" added in both directions.
175 - server close : the server-facing connection is closed after the response.
176 - forced close : the connection is actively closed after end of response.
177
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200178
1791.2. HTTP request
180-----------------
181
182First, let's consider this HTTP request :
183
184 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100185 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200186 1 GET /serv/login.php?lang=en&profile=2 HTTP/1.1
187 2 Host: www.mydomain.com
188 3 User-agent: my small browser
189 4 Accept: image/jpeg, image/gif
190 5 Accept: image/png
191
192
1931.2.1. The Request line
194-----------------------
195
196Line 1 is the "request line". It is always composed of 3 fields :
197
198 - a METHOD : GET
199 - a URI : /serv/login.php?lang=en&profile=2
200 - a version tag : HTTP/1.1
201
202All of them are delimited by what the standard calls LWS (linear white spaces),
203which are commonly spaces, but can also be tabs or line feeds/carriage returns
204followed by spaces/tabs. The method itself cannot contain any colon (':') and
205is limited to alphabetic letters. All those various combinations make it
206desirable that HAProxy performs the splitting itself rather than leaving it to
207the user to write a complex or inaccurate regular expression.
208
209The URI itself can have several forms :
210
211 - A "relative URI" :
212
213 /serv/login.php?lang=en&profile=2
214
215 It is a complete URL without the host part. This is generally what is
216 received by servers, reverse proxies and transparent proxies.
217
218 - An "absolute URI", also called a "URL" :
219
220 http://192.168.0.12:8080/serv/login.php?lang=en&profile=2
221
222 It is composed of a "scheme" (the protocol name followed by '://'), a host
223 name or address, optionally a colon (':') followed by a port number, then
224 a relative URI beginning at the first slash ('/') after the address part.
225 This is generally what proxies receive, but a server supporting HTTP/1.1
226 must accept this form too.
227
228 - a star ('*') : this form is only accepted in association with the OPTIONS
229 method and is not relayable. It is used to inquiry a next hop's
230 capabilities.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100231
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200232 - an address:port combination : 192.168.0.12:80
233 This is used with the CONNECT method, which is used to establish TCP
234 tunnels through HTTP proxies, generally for HTTPS, but sometimes for
235 other protocols too.
236
237In a relative URI, two sub-parts are identified. The part before the question
238mark is called the "path". It is typically the relative path to static objects
239on the server. The part after the question mark is called the "query string".
240It is mostly used with GET requests sent to dynamic scripts and is very
241specific to the language, framework or application in use.
242
243
2441.2.2. The request headers
245--------------------------
246
247The headers start at the second line. They are composed of a name at the
248beginning of the line, immediately followed by a colon (':'). Traditionally,
249an LWS is added after the colon but that's not required. Then come the values.
250Multiple identical headers may be folded into one single line, delimiting the
251values with commas, provided that their order is respected. This is commonly
252encountered in the "Cookie:" field. A header may span over multiple lines if
253the subsequent lines begin with an LWS. In the example in 1.2, lines 4 and 5
254define a total of 3 values for the "Accept:" header.
255
256Contrary to a common mis-conception, header names are not case-sensitive, and
257their values are not either if they refer to other header names (such as the
258"Connection:" header).
259
260The end of the headers is indicated by the first empty line. People often say
261that it's a double line feed, which is not exact, even if a double line feed
262is one valid form of empty line.
263
264Fortunately, HAProxy takes care of all these complex combinations when indexing
265headers, checking values and counting them, so there is no reason to worry
266about the way they could be written, but it is important not to accuse an
267application of being buggy if it does unusual, valid things.
268
269Important note:
270 As suggested by RFC2616, HAProxy normalizes headers by replacing line breaks
271 in the middle of headers by LWS in order to join multi-line headers. This
272 is necessary for proper analysis and helps less capable HTTP parsers to work
273 correctly and not to be fooled by such complex constructs.
274
275
2761.3. HTTP response
277------------------
278
279An HTTP response looks very much like an HTTP request. Both are called HTTP
280messages. Let's consider this HTTP response :
281
282 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100283 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200284 1 HTTP/1.1 200 OK
285 2 Content-length: 350
286 3 Content-Type: text/html
287
Willy Tarreau816b9792009-09-15 21:25:21 +0200288As a special case, HTTP supports so called "Informational responses" as status
289codes 1xx. These messages are special in that they don't convey any part of the
290response, they're just used as sort of a signaling message to ask a client to
Willy Tarreau5843d1a2010-02-01 15:13:32 +0100291continue to post its request for instance. In the case of a status 100 response
292the requested information will be carried by the next non-100 response message
293following the informational one. This implies that multiple responses may be
294sent to a single request, and that this only works when keep-alive is enabled
295(1xx messages are HTTP/1.1 only). HAProxy handles these messages and is able to
296correctly forward and skip them, and only process the next non-100 response. As
297such, these messages are neither logged nor transformed, unless explicitly
298state otherwise. Status 101 messages indicate that the protocol is changing
299over the same connection and that haproxy must switch to tunnel mode, just as
300if a CONNECT had occurred. Then the Upgrade header would contain additional
301information about the type of protocol the connection is switching to.
Willy Tarreau816b9792009-09-15 21:25:21 +0200302
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200303
3041.3.1. The Response line
305------------------------
306
307Line 1 is the "response line". It is always composed of 3 fields :
308
309 - a version tag : HTTP/1.1
310 - a status code : 200
311 - a reason : OK
312
313The status code is always 3-digit. The first digit indicates a general status :
Willy Tarreau816b9792009-09-15 21:25:21 +0200314 - 1xx = informational message to be skipped (eg: 100, 101)
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200315 - 2xx = OK, content is following (eg: 200, 206)
316 - 3xx = OK, no content following (eg: 302, 304)
317 - 4xx = error caused by the client (eg: 401, 403, 404)
318 - 5xx = error caused by the server (eg: 500, 502, 503)
319
320Please refer to RFC2616 for the detailed meaning of all such codes. The
Willy Tarreaud72758d2010-01-12 10:42:19 +0100321"reason" field is just a hint, but is not parsed by clients. Anything can be
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200322found there, but it's a common practice to respect the well-established
323messages. It can be composed of one or multiple words, such as "OK", "Found",
324or "Authentication Required".
325
326Haproxy may emit the following status codes by itself :
327
328 Code When / reason
329 200 access to stats page, and when replying to monitoring requests
330 301 when performing a redirection, depending on the configured code
331 302 when performing a redirection, depending on the configured code
332 303 when performing a redirection, depending on the configured code
Willy Tarreaub67fdc42013-03-29 19:28:11 +0100333 307 when performing a redirection, depending on the configured code
334 308 when performing a redirection, depending on the configured code
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200335 400 for an invalid or too large request
336 401 when an authentication is required to perform the action (when
337 accessing the stats page)
338 403 when a request is forbidden by a "block" ACL or "reqdeny" filter
339 408 when the request timeout strikes before the request is complete
340 500 when haproxy encounters an unrecoverable internal error, such as a
341 memory allocation failure, which should never happen
342 502 when the server returns an empty, invalid or incomplete response, or
343 when an "rspdeny" filter blocks the response.
344 503 when no server was available to handle the request, or in response to
345 monitoring requests which match the "monitor fail" condition
346 504 when the response timeout strikes before the server responds
347
348The error 4xx and 5xx codes above may be customized (see "errorloc" in section
3494.2).
350
351
3521.3.2. The response headers
353---------------------------
354
355Response headers work exactly like request headers, and as such, HAProxy uses
356the same parsing function for both. Please refer to paragraph 1.2.2 for more
357details.
358
359
3602. Configuring HAProxy
361----------------------
362
3632.1. Configuration file format
364------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +0200365
366HAProxy's configuration process involves 3 major sources of parameters :
367
368 - the arguments from the command-line, which always take precedence
369 - the "global" section, which sets process-wide parameters
370 - the proxies sections which can take form of "defaults", "listen",
371 "frontend" and "backend".
372
Willy Tarreau0ba27502007-12-24 16:55:16 +0100373The configuration file syntax consists in lines beginning with a keyword
374referenced in this manual, optionally followed by one or several parameters
William Lallemandf9873ba2015-05-05 17:37:14 +0200375delimited by spaces.
Willy Tarreau0ba27502007-12-24 16:55:16 +0100376
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200377
William Lallemandf9873ba2015-05-05 17:37:14 +02003782.2. Quoting and escaping
379-------------------------
380
381HAProxy's configuration introduces a quoting and escaping system similar to
382many programming languages. The configuration file supports 3 types: escaping
383with a backslash, weak quoting with double quotes, and strong quoting with
384single quotes.
385
386If spaces have to be entered in strings, then they must be escaped by preceding
387them by a backslash ('\') or by quoting them. Backslashes also have to be
388escaped by doubling or strong quoting them.
389
390Escaping is achieved by preceding a special character by a backslash ('\'):
391
392 \ to mark a space and differentiate it from a delimiter
393 \# to mark a hash and differentiate it from a comment
394 \\ to use a backslash
395 \' to use a single quote and differentiate it from strong quoting
396 \" to use a double quote and differentiate it from weak quoting
397
398Weak quoting is achieved by using double quotes (""). Weak quoting prevents
399the interpretation of:
400
401 space as a parameter separator
402 ' single quote as a strong quoting delimiter
403 # hash as a comment start
404
William Lallemandb2f07452015-05-12 14:27:13 +0200405Weak quoting permits the interpretation of variables, if you want to use a non
406-interpreted dollar within a double quoted string, you should escape it with a
407backslash ("\$"), it does not work outside weak quoting.
408
409Interpretation of escaping and special characters are not prevented by weak
William Lallemandf9873ba2015-05-05 17:37:14 +0200410quoting.
411
412Strong quoting is achieved by using single quotes (''). Inside single quotes,
413nothing is interpreted, it's the efficient way to quote regexes.
414
415Quoted and escaped strings are replaced in memory by their interpreted
416equivalent, it allows you to perform concatenation.
417
418 Example:
419 # those are equivalents:
420 log-format %{+Q}o\ %t\ %s\ %{-Q}r
421 log-format "%{+Q}o %t %s %{-Q}r"
422 log-format '%{+Q}o %t %s %{-Q}r'
423 log-format "%{+Q}o %t"' %s %{-Q}r'
424 log-format "%{+Q}o %t"' %s'\ %{-Q}r
425
426 # those are equivalents:
427 reqrep "^([^\ :]*)\ /static/(.*)" \1\ /\2
428 reqrep "^([^ :]*)\ /static/(.*)" '\1 /\2'
429 reqrep "^([^ :]*)\ /static/(.*)" "\1 /\2"
430 reqrep "^([^ :]*)\ /static/(.*)" "\1\ /\2"
431
432
William Lallemandb2f07452015-05-12 14:27:13 +02004332.3. Environment variables
434--------------------------
435
436HAProxy's configuration supports environment variables. Those variables are
437interpreted only within double quotes. Variables are expanded during the
438configuration parsing. Variable names must be preceded by a dollar ("$") and
439optionally enclosed with braces ("{}") similarly to what is done in Bourne
440shell. Variable names can contain alphanumerical characters or the character
441underscore ("_") but should not start with a digit.
442
443 Example:
444
445 bind "fd@${FD_APP1}"
446
447 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
448
449 user "$HAPROXY_USER"
450
451
4522.4. Time format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200453----------------
454
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100455Some parameters involve values representing time, such as timeouts. These
Willy Tarreau0ba27502007-12-24 16:55:16 +0100456values are generally expressed in milliseconds (unless explicitly stated
457otherwise) but may be expressed in any other unit by suffixing the unit to the
458numeric value. It is important to consider this because it will not be repeated
459for every keyword. Supported units are :
460
461 - us : microseconds. 1 microsecond = 1/1000000 second
462 - ms : milliseconds. 1 millisecond = 1/1000 second. This is the default.
463 - s : seconds. 1s = 1000ms
464 - m : minutes. 1m = 60s = 60000ms
465 - h : hours. 1h = 60m = 3600s = 3600000ms
466 - d : days. 1d = 24h = 1440m = 86400s = 86400000ms
467
468
William Lallemandf9873ba2015-05-05 17:37:14 +02004692.4. Examples
Patrick Mezard35da19c2010-06-12 17:02:47 +0200470-------------
471
472 # Simple configuration for an HTTP proxy listening on port 80 on all
473 # interfaces and forwarding requests to a single backend "servers" with a
474 # single server "server1" listening on 127.0.0.1:8000
475 global
476 daemon
477 maxconn 256
478
479 defaults
480 mode http
481 timeout connect 5000ms
482 timeout client 50000ms
483 timeout server 50000ms
484
485 frontend http-in
486 bind *:80
487 default_backend servers
488
489 backend servers
490 server server1 127.0.0.1:8000 maxconn 32
491
492
493 # The same configuration defined with a single listen block. Shorter but
494 # less expressive, especially in HTTP mode.
495 global
496 daemon
497 maxconn 256
498
499 defaults
500 mode http
501 timeout connect 5000ms
502 timeout client 50000ms
503 timeout server 50000ms
504
505 listen http-in
506 bind *:80
507 server server1 127.0.0.1:8000 maxconn 32
508
509
510Assuming haproxy is in $PATH, test these configurations in a shell with:
511
Willy Tarreauccb289d2010-12-11 20:19:38 +0100512 $ sudo haproxy -f configuration.conf -c
Patrick Mezard35da19c2010-06-12 17:02:47 +0200513
514
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005153. Global parameters
Willy Tarreau6a06a402007-07-15 20:15:28 +0200516--------------------
517
518Parameters in the "global" section are process-wide and often OS-specific. They
519are generally set once for all and do not need being changed once correct. Some
520of them have command-line equivalents.
521
522The following keywords are supported in the "global" section :
523
524 * Process management and security
Emeric Brunc8e8d122012-10-02 18:42:10 +0200525 - ca-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200526 - chroot
Emeric Brunc8e8d122012-10-02 18:42:10 +0200527 - crt-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200528 - daemon
Simon Horman98637e52014-06-20 12:30:16 +0900529 - external-check
Willy Tarreau6a06a402007-07-15 20:15:28 +0200530 - gid
531 - group
532 - log
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100533 - log-send-hostname
Willy Tarreau6a06a402007-07-15 20:15:28 +0200534 - nbproc
535 - pidfile
536 - uid
537 - ulimit-n
538 - user
Willy Tarreaufbee7132007-10-18 13:53:22 +0200539 - stats
Emeric Brun850efd52014-01-29 12:24:34 +0100540 - ssl-server-verify
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +0200541 - node
542 - description
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100543 - unix-bind
Thomas Holmesdb04f192015-05-18 13:21:39 +0100544 - 51degrees-data-file
545 - 51degrees-property-name-list
Dragan Dosen93b38d92015-06-29 16:43:25 +0200546 - 51degrees-property-separator
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200547 - 51degrees-cache-size
Willy Tarreaud72758d2010-01-12 10:42:19 +0100548
Willy Tarreau6a06a402007-07-15 20:15:28 +0200549 * Performance tuning
Willy Tarreau1746eec2014-04-25 10:46:47 +0200550 - max-spread-checks
Willy Tarreau6a06a402007-07-15 20:15:28 +0200551 - maxconn
Willy Tarreau81c25d02011-09-07 15:17:21 +0200552 - maxconnrate
William Lallemandd85f9172012-11-09 17:05:39 +0100553 - maxcomprate
William Lallemand072a2bf2012-11-20 17:01:01 +0100554 - maxcompcpuusage
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100555 - maxpipes
Willy Tarreau93e7c002013-10-07 18:51:07 +0200556 - maxsessrate
Willy Tarreau403edff2012-09-06 11:58:37 +0200557 - maxsslconn
Willy Tarreaue43d5322013-10-07 20:01:52 +0200558 - maxsslrate
Willy Tarreau6a06a402007-07-15 20:15:28 +0200559 - noepoll
560 - nokqueue
561 - nopoll
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100562 - nosplice
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300563 - nogetaddrinfo
Willy Tarreaufe255b72007-10-14 23:09:26 +0200564 - spread-checks
Baptiste Assmann5626f482015-08-23 10:00:10 +0200565 - server-state-base
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200566 - server-state-file
Willy Tarreau27a674e2009-08-17 07:23:33 +0200567 - tune.bufsize
Willy Tarreau43961d52010-10-04 20:39:20 +0200568 - tune.chksize
William Lallemandf3747832012-11-09 12:33:10 +0100569 - tune.comp.maxlevel
Willy Tarreau193b8c62012-11-22 00:17:38 +0100570 - tune.http.cookielen
Willy Tarreauac1932d2011-10-24 19:14:41 +0200571 - tune.http.maxhdr
Willy Tarreau7e312732014-02-12 16:35:14 +0100572 - tune.idletimer
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100573 - tune.lua.forced-yield
Willy Tarreau32f61e22015-03-18 17:54:59 +0100574 - tune.lua.maxmem
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100575 - tune.lua.session-timeout
576 - tune.lua.task-timeout
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +0200577 - tune.lua.service-timeout
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100578 - tune.maxaccept
579 - tune.maxpollevents
Willy Tarreau27a674e2009-08-17 07:23:33 +0200580 - tune.maxrewrite
Willy Tarreauf3045d22015-04-29 16:24:50 +0200581 - tune.pattern.cache-size
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200582 - tune.pipesize
Willy Tarreaue803de22010-01-21 17:43:04 +0100583 - tune.rcvbuf.client
584 - tune.rcvbuf.server
585 - tune.sndbuf.client
586 - tune.sndbuf.server
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100587 - tune.ssl.cachesize
Willy Tarreaubfd59462013-02-21 07:46:09 +0100588 - tune.ssl.lifetime
Emeric Brun8dc60392014-05-09 13:52:00 +0200589 - tune.ssl.force-private-cache
Willy Tarreaubfd59462013-02-21 07:46:09 +0100590 - tune.ssl.maxrecord
Remi Gacognef46cd6e2014-06-12 14:58:40 +0200591 - tune.ssl.default-dh-param
Christopher Faulet31af49d2015-06-09 17:29:50 +0200592 - tune.ssl.ssl-ctx-cache-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200593 - tune.vars.global-max-size
594 - tune.vars.reqres-max-size
595 - tune.vars.sess-max-size
596 - tune.vars.txn-max-size
William Lallemanda509e4c2012-11-07 16:54:34 +0100597 - tune.zlib.memlevel
598 - tune.zlib.windowsize
Willy Tarreaud72758d2010-01-12 10:42:19 +0100599
Willy Tarreau6a06a402007-07-15 20:15:28 +0200600 * Debugging
601 - debug
602 - quiet
Willy Tarreau6a06a402007-07-15 20:15:28 +0200603
604
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006053.1. Process management and security
Willy Tarreau6a06a402007-07-15 20:15:28 +0200606------------------------------------
607
Emeric Brunc8e8d122012-10-02 18:42:10 +0200608ca-base <dir>
609 Assigns a default directory to fetch SSL CA certificates and CRLs from when a
Emeric Brunfd33a262012-10-11 16:28:27 +0200610 relative path is used with "ca-file" or "crl-file" directives. Absolute
611 locations specified in "ca-file" and "crl-file" prevail and ignore "ca-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +0200612
Willy Tarreau6a06a402007-07-15 20:15:28 +0200613chroot <jail dir>
614 Changes current directory to <jail dir> and performs a chroot() there before
615 dropping privileges. This increases the security level in case an unknown
616 vulnerability would be exploited, since it would make it very hard for the
617 attacker to exploit the system. This only works when the process is started
618 with superuser privileges. It is important to ensure that <jail_dir> is both
619 empty and unwritable to anyone.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100620
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100621cpu-map <"all"|"odd"|"even"|process_num> <cpu-set>...
622 On Linux 2.6 and above, it is possible to bind a process to a specific CPU
623 set. This means that the process will never run on other CPUs. The "cpu-map"
624 directive specifies CPU sets for process sets. The first argument is the
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100625 process number to bind. This process must have a number between 1 and 32 or
626 64, depending on the machine's word size, and any process IDs above nbproc
627 are ignored. It is possible to specify all processes at once using "all",
628 only odd numbers using "odd" or even numbers using "even", just like with the
629 "bind-process" directive. The second and forthcoming arguments are CPU sets.
630 Each CPU set is either a unique number between 0 and 31 or 63 or a range with
631 two such numbers delimited by a dash ('-'). Multiple CPU numbers or ranges
632 may be specified, and the processes will be allowed to bind to all of them.
633 Obviously, multiple "cpu-map" directives may be specified. Each "cpu-map"
634 directive will replace the previous ones when they overlap.
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100635
Emeric Brunc8e8d122012-10-02 18:42:10 +0200636crt-base <dir>
637 Assigns a default directory to fetch SSL certificates from when a relative
638 path is used with "crtfile" directives. Absolute locations specified after
639 "crtfile" prevail and ignore "crt-base".
640
Willy Tarreau6a06a402007-07-15 20:15:28 +0200641daemon
642 Makes the process fork into background. This is the recommended mode of
643 operation. It is equivalent to the command line "-D" argument. It can be
644 disabled by the command line "-db" argument.
645
David Carlier8167f302015-06-01 13:50:06 +0200646deviceatlas-json-file <path>
647 Sets the path of the DeviceAtlas JSON data file to be loaded by the API.
648 The path must be a valid JSON data file and accessible by Haproxy process.
649
650deviceatlas-log-level <value>
651 Sets the level of informations returned by the API. This directive is
652 optional and set to 0 by default if not set.
653
654deviceatlas-separator <char>
655 Sets the character separator for the API properties results. This directive
656 is optional and set to | by default if not set.
657
David Carlier29b3ca32015-09-25 14:09:21 +0100658devicatlas-properties-cookie <name>
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200659 Sets the client cookie's name used for the detection if the DeviceAtlas
660 Client-side component was used during the request. This directive is optional
661 and set to DAPROPS by default if not set.
David Carlier29b3ca32015-09-25 14:09:21 +0100662
Simon Horman98637e52014-06-20 12:30:16 +0900663external-check
664 Allows the use of an external agent to perform health checks.
665 This is disabled by default as a security precaution.
666 See "option external-check".
667
Willy Tarreau6a06a402007-07-15 20:15:28 +0200668gid <number>
669 Changes the process' group ID to <number>. It is recommended that the group
670 ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
671 be started with a user belonging to this group, or with superuser privileges.
Michael Schererab012dd2013-01-12 18:35:19 +0100672 Note that if haproxy is started from a user having supplementary groups, it
673 will only be able to drop these groups if started with superuser privileges.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200674 See also "group" and "uid".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100675
Willy Tarreau6a06a402007-07-15 20:15:28 +0200676group <group name>
677 Similar to "gid" but uses the GID of group name <group name> from /etc/group.
678 See also "gid" and "user".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100679
Dragan Dosen7ad31542015-09-28 17:16:47 +0200680log <address> [len <length>] [format <format>] <facility> [max level [min level]]
Willy Tarreau6a06a402007-07-15 20:15:28 +0200681 Adds a global syslog server. Up to two global servers can be defined. They
682 will receive logs for startups and exits, as well as all logs from proxies
Robert Tsai81ae1952007-12-05 10:47:29 +0100683 configured with "log global".
684
685 <address> can be one of:
686
Willy Tarreau2769aa02007-12-27 18:26:09 +0100687 - An IPv4 address optionally followed by a colon and a UDP port. If
Robert Tsai81ae1952007-12-05 10:47:29 +0100688 no port is specified, 514 is used by default (the standard syslog
689 port).
690
David du Colombier24bb5f52011-03-17 10:40:23 +0100691 - An IPv6 address followed by a colon and optionally a UDP port. If
692 no port is specified, 514 is used by default (the standard syslog
693 port).
694
Robert Tsai81ae1952007-12-05 10:47:29 +0100695 - A filesystem path to a UNIX domain socket, keeping in mind
696 considerations for chroot (be sure the path is accessible inside
697 the chroot) and uid/gid (be sure the path is appropriately
698 writeable).
699
William Lallemandb2f07452015-05-12 14:27:13 +0200700 You may want to reference some environment variables in the address
701 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +0100702
Willy Tarreau18324f52014-06-27 18:10:07 +0200703 <length> is an optional maximum line length. Log lines larger than this value
704 will be truncated before being sent. The reason is that syslog
705 servers act differently on log line length. All servers support the
706 default value of 1024, but some servers simply drop larger lines
707 while others do log them. If a server supports long lines, it may
708 make sense to set this value here in order to avoid truncating long
709 lines. Similarly, if a server drops long lines, it is preferable to
710 truncate them before sending them. Accepted values are 80 to 65535
711 inclusive. The default value of 1024 is generally fine for all
712 standard usages. Some specific cases of long captures or
713 JSON-formated logs may require larger values.
714
Dragan Dosen7ad31542015-09-28 17:16:47 +0200715 <format> is the log format used when generating syslog messages. It may be
716 one of the following :
717
718 rfc3164 The RFC3164 syslog message format. This is the default.
719 (https://tools.ietf.org/html/rfc3164)
720
721 rfc5424 The RFC5424 syslog message format.
722 (https://tools.ietf.org/html/rfc5424)
723
Robert Tsai81ae1952007-12-05 10:47:29 +0100724 <facility> must be one of the 24 standard syslog facilities :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200725
726 kern user mail daemon auth syslog lpr news
727 uucp cron auth2 ftp ntp audit alert cron2
728 local0 local1 local2 local3 local4 local5 local6 local7
729
730 An optional level can be specified to filter outgoing messages. By default,
Willy Tarreauf7edefa2009-05-10 17:20:05 +0200731 all messages are sent. If a maximum level is specified, only messages with a
732 severity at least as important as this level will be sent. An optional minimum
733 level can be specified. If it is set, logs emitted with a more severe level
734 than this one will be capped to this level. This is used to avoid sending
735 "emerg" messages on all terminals on some default syslog configurations.
736 Eight levels are known :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200737
Cyril Bontédc4d9032012-04-08 21:57:39 +0200738 emerg alert crit err warning notice info debug
Willy Tarreau6a06a402007-07-15 20:15:28 +0200739
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100740log-send-hostname [<string>]
741 Sets the hostname field in the syslog header. If optional "string" parameter
742 is set the header is set to the string contents, otherwise uses the hostname
743 of the system. Generally used if one is not relaying logs through an
744 intermediate syslog server or for simply customizing the hostname printed in
745 the logs.
746
Kevinm48936af2010-12-22 16:08:21 +0000747log-tag <string>
748 Sets the tag field in the syslog header to this string. It defaults to the
749 program name as launched from the command line, which usually is "haproxy".
750 Sometimes it can be useful to differentiate between multiple processes
Willy Tarreau094af4e2015-01-07 15:03:42 +0100751 running on the same host. See also the per-proxy "log-tag" directive.
Kevinm48936af2010-12-22 16:08:21 +0000752
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100753lua-load <file>
754 This global directive loads and executes a Lua file. This directive can be
755 used multiple times.
756
Willy Tarreau6a06a402007-07-15 20:15:28 +0200757nbproc <number>
758 Creates <number> processes when going daemon. This requires the "daemon"
759 mode. By default, only one process is created, which is the recommended mode
760 of operation. For systems limited to small sets of file descriptors per
761 process, it may be needed to fork multiple daemons. USING MULTIPLE PROCESSES
762 IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. See also "daemon".
763
764pidfile <pidfile>
765 Writes pids of all daemons into file <pidfile>. This option is equivalent to
766 the "-p" command line argument. The file must be accessible to the user
767 starting the process. See also "daemon".
768
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100769stats bind-process [ all | odd | even | <number 1-64>[-<number 1-64>] ] ...
Willy Tarreau35b7b162012-10-22 23:17:18 +0200770 Limits the stats socket to a certain set of processes numbers. By default the
771 stats socket is bound to all processes, causing a warning to be emitted when
772 nbproc is greater than 1 because there is no way to select the target process
773 when connecting. However, by using this setting, it becomes possible to pin
774 the stats socket to a specific set of processes, typically the first one. The
775 warning will automatically be disabled when this setting is used, whatever
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100776 the number of processes used. The maximum process ID depends on the machine's
Willy Tarreauae302532014-05-07 19:22:24 +0200777 word size (32 or 64). A better option consists in using the "process" setting
778 of the "stats socket" line to force the process on each line.
Willy Tarreau35b7b162012-10-22 23:17:18 +0200779
Baptiste Assmann5626f482015-08-23 10:00:10 +0200780server-state-base <directory>
781 Specifies the directory prefix to be prepended in front of all servers state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +0200782 file names which do not start with a '/'. See also "server-state-file",
783 "load-server-state-from-file" and "server-state-file-name".
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200784
785server-state-file <file>
786 Specifies the path to the file containing state of servers. If the path starts
787 with a slash ('/'), it is considered absolute, otherwise it is considered
788 relative to the directory specified using "server-state-base" (if set) or to
789 the current directory. Before reloading HAProxy, it is possible to save the
790 servers' current state using the stats command "show servers state". The
791 output of this command must be written in the file pointed by <file>. When
792 starting up, before handling traffic, HAProxy will read, load and apply state
793 for each server found in the file and available in its current running
Baptiste Assmann01c6cc32015-08-23 11:45:29 +0200794 configuration. See also "server-state-base" and "show servers state",
795 "load-server-state-from-file" and "server-state-file-name"
Baptiste Assmann5626f482015-08-23 10:00:10 +0200796
Willy Tarreau610f04b2014-02-13 11:36:41 +0100797ssl-default-bind-ciphers <ciphers>
798 This setting is only available when support for OpenSSL was built in. It sets
799 the default string describing the list of cipher algorithms ("cipher suite")
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300800 that are negotiated during the SSL/TLS handshake for all "bind" lines which
Willy Tarreau610f04b2014-02-13 11:36:41 +0100801 do not explicitly define theirs. The format of the string is defined in
802 "man 1 ciphers" from OpenSSL man pages, and can be for instance a string such
803 as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes). Please check the
804 "bind" keyword for more information.
805
Emeric Brun2c86cbf2014-10-30 15:56:50 +0100806ssl-default-bind-options [<option>]...
807 This setting is only available when support for OpenSSL was built in. It sets
808 default ssl-options to force on all "bind" lines. Please check the "bind"
809 keyword to see available options.
810
811 Example:
812 global
813 ssl-default-bind-options no-sslv3 no-tls-tickets
814
Willy Tarreau610f04b2014-02-13 11:36:41 +0100815ssl-default-server-ciphers <ciphers>
816 This setting is only available when support for OpenSSL was built in. It
817 sets the default string describing the list of cipher algorithms that are
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300818 negotiated during the SSL/TLS handshake with the server, for all "server"
Willy Tarreau610f04b2014-02-13 11:36:41 +0100819 lines which do not explicitly define theirs. The format of the string is
820 defined in "man 1 ciphers". Please check the "server" keyword for more
821 information.
822
Emeric Brun2c86cbf2014-10-30 15:56:50 +0100823ssl-default-server-options [<option>]...
824 This setting is only available when support for OpenSSL was built in. It sets
825 default ssl-options to force on all "server" lines. Please check the "server"
826 keyword to see available options.
827
Remi Gacogne47783ef2015-05-29 15:53:22 +0200828ssl-dh-param-file <file>
829 This setting is only available when support for OpenSSL was built in. It sets
830 the default DH parameters that are used during the SSL/TLS handshake when
831 ephemeral Diffie-Hellman (DHE) key exchange is used, for all "bind" lines
832 which do not explicitely define theirs. It will be overridden by custom DH
833 parameters found in a bind certificate file if any. If custom DH parameters
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200834 are not specified either by using ssl-dh-param-file or by setting them
835 directly in the certificate file, pre-generated DH parameters of the size
836 specified by tune.ssl.default-dh-param will be used. Custom parameters are
837 known to be more secure and therefore their use is recommended.
Remi Gacogne47783ef2015-05-29 15:53:22 +0200838 Custom DH parameters may be generated by using the OpenSSL command
839 "openssl dhparam <size>", where size should be at least 2048, as 1024-bit DH
840 parameters should not be considered secure anymore.
841
Emeric Brun850efd52014-01-29 12:24:34 +0100842ssl-server-verify [none|required]
843 The default behavior for SSL verify on servers side. If specified to 'none',
844 servers certificates are not verified. The default is 'required' except if
845 forced using cmdline option '-dV'.
846
Willy Tarreauabb175f2012-09-24 12:43:26 +0200847stats socket [<address:port>|<path>] [param*]
848 Binds a UNIX socket to <path> or a TCPv4/v6 address to <address:port>.
849 Connections to this socket will return various statistics outputs and even
850 allow some commands to be issued to change some runtime settings. Please
851 consult section 9.2 "Unix Socket commands" for more details.
Willy Tarreau6162db22009-10-10 17:13:00 +0200852
Willy Tarreauabb175f2012-09-24 12:43:26 +0200853 All parameters supported by "bind" lines are supported, for instance to
854 restrict access to some users or their access rights. Please consult
855 section 5.1 for more information.
Willy Tarreaufbee7132007-10-18 13:53:22 +0200856
857stats timeout <timeout, in milliseconds>
858 The default timeout on the stats socket is set to 10 seconds. It is possible
859 to change this value with "stats timeout". The value must be passed in
Willy Tarreaubefdff12007-12-02 22:27:38 +0100860 milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.
Willy Tarreaufbee7132007-10-18 13:53:22 +0200861
862stats maxconn <connections>
863 By default, the stats socket is limited to 10 concurrent connections. It is
864 possible to change this value with "stats maxconn".
865
Willy Tarreau6a06a402007-07-15 20:15:28 +0200866uid <number>
867 Changes the process' user ID to <number>. It is recommended that the user ID
868 is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
869 be started with superuser privileges in order to be able to switch to another
870 one. See also "gid" and "user".
871
872ulimit-n <number>
873 Sets the maximum number of per-process file-descriptors to <number>. By
874 default, it is automatically computed, so it is recommended not to use this
875 option.
876
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100877unix-bind [ prefix <prefix> ] [ mode <mode> ] [ user <user> ] [ uid <uid> ]
878 [ group <group> ] [ gid <gid> ]
879
880 Fixes common settings to UNIX listening sockets declared in "bind" statements.
881 This is mainly used to simplify declaration of those UNIX sockets and reduce
882 the risk of errors, since those settings are most commonly required but are
883 also process-specific. The <prefix> setting can be used to force all socket
884 path to be relative to that directory. This might be needed to access another
885 component's chroot. Note that those paths are resolved before haproxy chroots
886 itself, so they are absolute. The <mode>, <user>, <uid>, <group> and <gid>
887 all have the same meaning as their homonyms used by the "bind" statement. If
888 both are specified, the "bind" statement has priority, meaning that the
889 "unix-bind" settings may be seen as process-wide default settings.
890
Willy Tarreau6a06a402007-07-15 20:15:28 +0200891user <user name>
892 Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.
893 See also "uid" and "group".
894
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +0200895node <name>
896 Only letters, digits, hyphen and underscore are allowed, like in DNS names.
897
898 This statement is useful in HA configurations where two or more processes or
899 servers share the same IP address. By setting a different node-name on all
900 nodes, it becomes easy to immediately spot what server is handling the
901 traffic.
902
903description <text>
904 Add a text that describes the instance.
905
906 Please note that it is required to escape certain characters (# for example)
907 and this text is inserted into a html page so you should avoid using
908 "<" and ">" characters.
909
Thomas Holmesdb04f192015-05-18 13:21:39 +010091051degrees-data-file <file path>
911 The path of the 51Degrees data file to provide device detection services. The
912 file should be unzipped and accessible by HAProxy with relevavnt permissions.
913
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200914 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +0100915 compiled with USE_51DEGREES.
916
91751degrees-property-name-list [<string>]
918 A list of 51Degrees property names to be load from the dataset. A full list
919 of names is available on the 51Degrees website:
920 https://51degrees.com/resources/property-dictionary
921
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200922 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +0100923 compiled with USE_51DEGREES.
924
Dragan Dosen93b38d92015-06-29 16:43:25 +020092551degrees-property-separator <char>
Thomas Holmesdb04f192015-05-18 13:21:39 +0100926 A char that will be appended to every property value in a response header
927 containing 51Degrees results. If not set that will be set as ','.
928
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200929 Please note that this option is only available when haproxy has been
930 compiled with USE_51DEGREES.
931
93251degrees-cache-size <number>
933 Sets the size of the 51Degrees converter cache to <number> entries. This
934 is an LRU cache which reminds previous device detections and their results.
935 By default, this cache is disabled.
936
937 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +0100938 compiled with USE_51DEGREES.
939
Willy Tarreau6a06a402007-07-15 20:15:28 +0200940
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009413.2. Performance tuning
Willy Tarreau6a06a402007-07-15 20:15:28 +0200942-----------------------
943
Willy Tarreau1746eec2014-04-25 10:46:47 +0200944max-spread-checks <delay in milliseconds>
945 By default, haproxy tries to spread the start of health checks across the
946 smallest health check interval of all the servers in a farm. The principle is
947 to avoid hammering services running on the same server. But when using large
948 check intervals (10 seconds or more), the last servers in the farm take some
949 time before starting to be tested, which can be a problem. This parameter is
950 used to enforce an upper bound on delay between the first and the last check,
951 even if the servers' check intervals are larger. When servers run with
952 shorter intervals, their intervals will be respected though.
953
Willy Tarreau6a06a402007-07-15 20:15:28 +0200954maxconn <number>
955 Sets the maximum per-process number of concurrent connections to <number>. It
956 is equivalent to the command-line argument "-n". Proxies will stop accepting
957 connections when this limit is reached. The "ulimit-n" parameter is
Willy Tarreau8274e102014-06-19 15:31:25 +0200958 automatically adjusted according to this value. See also "ulimit-n". Note:
959 the "select" poller cannot reliably use more than 1024 file descriptors on
960 some platforms. If your platform only supports select and reports "select
961 FAILED" on startup, you need to reduce maxconn until it works (slightly
Willy Tarreaud0256482015-01-15 21:45:22 +0100962 below 500 in general). If this value is not set, it will default to the value
963 set in DEFAULT_MAXCONN at build time (reported in haproxy -vv) if no memory
964 limit is enforced, or will be computed based on the memory limit, the buffer
965 size, memory allocated to compression, SSL cache size, and use or not of SSL
966 and the associated maxsslconn (which can also be automatic).
Willy Tarreau6a06a402007-07-15 20:15:28 +0200967
Willy Tarreau81c25d02011-09-07 15:17:21 +0200968maxconnrate <number>
969 Sets the maximum per-process number of connections per second to <number>.
970 Proxies will stop accepting connections when this limit is reached. It can be
971 used to limit the global capacity regardless of each frontend capacity. It is
972 important to note that this can only be used as a service protection measure,
973 as there will not necessarily be a fair share between frontends when the
974 limit is reached, so it's a good idea to also limit each frontend to some
975 value close to its expected share. Also, lowering tune.maxaccept can improve
976 fairness.
977
William Lallemandd85f9172012-11-09 17:05:39 +0100978maxcomprate <number>
979 Sets the maximum per-process input compression rate to <number> kilobytes
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300980 per second. For each session, if the maximum is reached, the compression
William Lallemandd85f9172012-11-09 17:05:39 +0100981 level will be decreased during the session. If the maximum is reached at the
982 beginning of a session, the session will not compress at all. If the maximum
983 is not reached, the compression level will be increased up to
984 tune.comp.maxlevel. A value of zero means there is no limit, this is the
985 default value.
986
William Lallemand072a2bf2012-11-20 17:01:01 +0100987maxcompcpuusage <number>
988 Sets the maximum CPU usage HAProxy can reach before stopping the compression
989 for new requests or decreasing the compression level of current requests.
990 It works like 'maxcomprate' but measures CPU usage instead of incoming data
991 bandwidth. The value is expressed in percent of the CPU used by haproxy. In
992 case of multiple processes (nbproc > 1), each process manages its individual
993 usage. A value of 100 disable the limit. The default value is 100. Setting
994 a lower value will prevent the compression work from slowing the whole
995 process down and from introducing high latencies.
996
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100997maxpipes <number>
998 Sets the maximum per-process number of pipes to <number>. Currently, pipes
999 are only used by kernel-based tcp splicing. Since a pipe contains two file
1000 descriptors, the "ulimit-n" value will be increased accordingly. The default
1001 value is maxconn/4, which seems to be more than enough for most heavy usages.
1002 The splice code dynamically allocates and releases pipes, and can fall back
1003 to standard copy, so setting this value too low may only impact performance.
1004
Willy Tarreau93e7c002013-10-07 18:51:07 +02001005maxsessrate <number>
1006 Sets the maximum per-process number of sessions per second to <number>.
1007 Proxies will stop accepting connections when this limit is reached. It can be
1008 used to limit the global capacity regardless of each frontend capacity. It is
1009 important to note that this can only be used as a service protection measure,
1010 as there will not necessarily be a fair share between frontends when the
1011 limit is reached, so it's a good idea to also limit each frontend to some
1012 value close to its expected share. Also, lowering tune.maxaccept can improve
1013 fairness.
1014
Willy Tarreau403edff2012-09-06 11:58:37 +02001015maxsslconn <number>
1016 Sets the maximum per-process number of concurrent SSL connections to
1017 <number>. By default there is no SSL-specific limit, which means that the
1018 global maxconn setting will apply to all connections. Setting this limit
1019 avoids having openssl use too much memory and crash when malloc returns NULL
1020 (since it unfortunately does not reliably check for such conditions). Note
1021 that the limit applies both to incoming and outgoing connections, so one
1022 connection which is deciphered then ciphered accounts for 2 SSL connections.
Willy Tarreaud0256482015-01-15 21:45:22 +01001023 If this value is not set, but a memory limit is enforced, this value will be
1024 automatically computed based on the memory limit, maxconn, the buffer size,
1025 memory allocated to compression, SSL cache size, and use of SSL in either
1026 frontends, backends or both. If neither maxconn nor maxsslconn are specified
1027 when there is a memory limit, haproxy will automatically adjust these values
1028 so that 100% of the connections can be made over SSL with no risk, and will
1029 consider the sides where it is enabled (frontend, backend, both).
Willy Tarreau403edff2012-09-06 11:58:37 +02001030
Willy Tarreaue43d5322013-10-07 20:01:52 +02001031maxsslrate <number>
1032 Sets the maximum per-process number of SSL sessions per second to <number>.
1033 SSL listeners will stop accepting connections when this limit is reached. It
1034 can be used to limit the global SSL CPU usage regardless of each frontend
1035 capacity. It is important to note that this can only be used as a service
1036 protection measure, as there will not necessarily be a fair share between
1037 frontends when the limit is reached, so it's a good idea to also limit each
1038 frontend to some value close to its expected share. It is also important to
1039 note that the sessions are accounted before they enter the SSL stack and not
1040 after, which also protects the stack against bad handshakes. Also, lowering
1041 tune.maxaccept can improve fairness.
1042
William Lallemand9d5f5482012-11-07 16:12:57 +01001043maxzlibmem <number>
1044 Sets the maximum amount of RAM in megabytes per process usable by the zlib.
1045 When the maximum amount is reached, future sessions will not compress as long
1046 as RAM is unavailable. When sets to 0, there is no limit.
William Lallemande3a7d992012-11-20 11:25:20 +01001047 The default value is 0. The value is available in bytes on the UNIX socket
1048 with "show info" on the line "MaxZlibMemUsage", the memory used by zlib is
1049 "ZlibMemUsage" in bytes.
1050
Willy Tarreau6a06a402007-07-15 20:15:28 +02001051noepoll
1052 Disables the use of the "epoll" event polling system on Linux. It is
1053 equivalent to the command-line argument "-de". The next polling system
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001054 used will generally be "poll". See also "nopoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001055
1056nokqueue
1057 Disables the use of the "kqueue" event polling system on BSD. It is
1058 equivalent to the command-line argument "-dk". The next polling system
1059 used will generally be "poll". See also "nopoll".
1060
1061nopoll
1062 Disables the use of the "poll" event polling system. It is equivalent to the
1063 command-line argument "-dp". The next polling system used will be "select".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001064 It should never be needed to disable "poll" since it's available on all
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001065 platforms supported by HAProxy. See also "nokqueue" and "noepoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001066
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001067nosplice
1068 Disables the use of kernel tcp splicing between sockets on Linux. It is
1069 equivalent to the command line argument "-dS". Data will then be copied
1070 using conventional and more portable recv/send calls. Kernel tcp splicing is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001071 limited to some very recent instances of kernel 2.6. Most versions between
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001072 2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not
1073 be used. This option makes it easier to globally disable kernel splicing in
1074 case of doubt. See also "option splice-auto", "option splice-request" and
1075 "option splice-response".
1076
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001077nogetaddrinfo
1078 Disables the use of getaddrinfo(3) for name resolving. It is equivalent to
1079 the command line argument "-dG". Deprecated gethostbyname(3) will be used.
1080
Willy Tarreaufe255b72007-10-14 23:09:26 +02001081spread-checks <0..50, in percent>
Simon Hormand60d6912013-11-25 10:46:36 +09001082 Sometimes it is desirable to avoid sending agent and health checks to
1083 servers at exact intervals, for instance when many logical servers are
1084 located on the same physical server. With the help of this parameter, it
1085 becomes possible to add some randomness in the check interval between 0
1086 and +/- 50%. A value between 2 and 5 seems to show good results. The
1087 default value remains at 0.
Willy Tarreaufe255b72007-10-14 23:09:26 +02001088
Willy Tarreau33cb0652014-12-23 22:52:37 +01001089tune.buffers.limit <number>
1090 Sets a hard limit on the number of buffers which may be allocated per process.
1091 The default value is zero which means unlimited. The minimum non-zero value
1092 will always be greater than "tune.buffers.reserve" and should ideally always
1093 be about twice as large. Forcing this value can be particularly useful to
1094 limit the amount of memory a process may take, while retaining a sane
1095 behaviour. When this limit is reached, sessions which need a buffer wait for
1096 another one to be released by another session. Since buffers are dynamically
1097 allocated and released, the waiting time is very short and not perceptible
1098 provided that limits remain reasonable. In fact sometimes reducing the limit
1099 may even increase performance by increasing the CPU cache's efficiency. Tests
1100 have shown good results on average HTTP traffic with a limit to 1/10 of the
1101 expected global maxconn setting, which also significantly reduces memory
1102 usage. The memory savings come from the fact that a number of connections
1103 will not allocate 2*tune.bufsize. It is best not to touch this value unless
1104 advised to do so by an haproxy core developer.
1105
Willy Tarreau1058ae72014-12-23 22:40:40 +01001106tune.buffers.reserve <number>
1107 Sets the number of buffers which are pre-allocated and reserved for use only
1108 during memory shortage conditions resulting in failed memory allocations. The
1109 minimum value is 2 and is also the default. There is no reason a user would
1110 want to change this value, it's mostly aimed at haproxy core developers.
1111
Willy Tarreau27a674e2009-08-17 07:23:33 +02001112tune.bufsize <number>
1113 Sets the buffer size to this size (in bytes). Lower values allow more
1114 sessions to coexist in the same amount of RAM, and higher values allow some
1115 applications with very large cookies to work. The default value is 16384 and
1116 can be changed at build time. It is strongly recommended not to change this
1117 from the default value, as very low values will break some services such as
1118 statistics, and values larger than default size will increase memory usage,
1119 possibly causing the system to run out of memory. At least the global maxconn
1120 parameter should be decreased by the same factor as this one is increased.
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04001121 If HTTP request is larger than (tune.bufsize - tune.maxrewrite), haproxy will
1122 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
1123 than this size, haproxy will return HTTP 502 (Bad Gateway).
Willy Tarreau27a674e2009-08-17 07:23:33 +02001124
Willy Tarreau43961d52010-10-04 20:39:20 +02001125tune.chksize <number>
1126 Sets the check buffer size to this size (in bytes). Higher values may help
1127 find string or regex patterns in very large pages, though doing so may imply
1128 more memory and CPU usage. The default value is 16384 and can be changed at
1129 build time. It is not recommended to change this value, but to use better
1130 checks whenever possible.
1131
William Lallemandf3747832012-11-09 12:33:10 +01001132tune.comp.maxlevel <number>
1133 Sets the maximum compression level. The compression level affects CPU
1134 usage during compression. This value affects CPU usage during compression.
1135 Each session using compression initializes the compression algorithm with
1136 this value. The default value is 1.
1137
Willy Tarreau193b8c62012-11-22 00:17:38 +01001138tune.http.cookielen <number>
1139 Sets the maximum length of captured cookies. This is the maximum value that
1140 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
1141 will automatically be truncated to this one. It is important not to set too
1142 high a value because all cookie captures still allocate this size whatever
1143 their configured value (they share a same pool). This value is per request
1144 per response, so the memory allocated is twice this value per connection.
1145 When not specified, the limit is set to 63 characters. It is recommended not
1146 to change this value.
1147
Willy Tarreauac1932d2011-10-24 19:14:41 +02001148tune.http.maxhdr <number>
1149 Sets the maximum number of headers in a request. When a request comes with a
1150 number of headers greater than this value (including the first line), it is
1151 rejected with a "400 Bad Request" status code. Similarly, too large responses
1152 are blocked with "502 Bad Gateway". The default value is 101, which is enough
1153 for all usages, considering that the widely deployed Apache server uses the
1154 same limit. It can be useful to push this limit further to temporarily allow
1155 a buggy application to work by the time it gets fixed. Keep in mind that each
1156 new header consumes 32bits of memory for each session, so don't push this
1157 limit too high.
1158
Willy Tarreau7e312732014-02-12 16:35:14 +01001159tune.idletimer <timeout>
1160 Sets the duration after which haproxy will consider that an empty buffer is
1161 probably associated with an idle stream. This is used to optimally adjust
1162 some packet sizes while forwarding large and small data alternatively. The
1163 decision to use splice() or to send large buffers in SSL is modulated by this
1164 parameter. The value is in milliseconds between 0 and 65535. A value of zero
1165 means that haproxy will not try to detect idle streams. The default is 1000,
1166 which seems to correctly detect end user pauses (eg: read a page before
1167 clicking). There should be not reason for changing this value. Please check
1168 tune.ssl.maxrecord below.
1169
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001170tune.lua.forced-yield <number>
1171 This directive forces the Lua engine to execute a yield each <number> of
1172 instructions executed. This permits interruptng a long script and allows the
1173 HAProxy scheduler to process other tasks like accepting connections or
1174 forwarding traffic. The default value is 10000 instructions. If HAProxy often
1175 executes some Lua code but more reactivity is required, this value can be
1176 lowered. If the Lua code is quite long and its result is absolutely required
1177 to process the data, the <number> can be increased.
1178
Willy Tarreau32f61e22015-03-18 17:54:59 +01001179tune.lua.maxmem
1180 Sets the maximum amount of RAM in megabytes per process usable by Lua. By
1181 default it is zero which means unlimited. It is important to set a limit to
1182 ensure that a bug in a script will not result in the system running out of
1183 memory.
1184
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001185tune.lua.session-timeout <timeout>
1186 This is the execution timeout for the Lua sessions. This is useful for
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001187 preventing infinite loops or spending too much time in Lua. This timeout
1188 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
1189 not taked in account. The default timeout is 4s.
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001190
1191tune.lua.task-timeout <timeout>
1192 Purpose is the same as "tune.lua.session-timeout", but this timeout is
1193 dedicated to the tasks. By default, this timeout isn't set because a task may
1194 remain alive during of the lifetime of HAProxy. For example, a task used to
1195 check servers.
1196
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001197tune.lua.service-timeout <timeout>
1198 This is the execution timeout for the Lua services. This is useful for
1199 preventing infinite loops or spending too much time in Lua. This timeout
1200 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
1201 not taked in account. The default timeout is 4s.
1202
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001203tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +01001204 Sets the maximum number of consecutive connections a process may accept in a
1205 row before switching to other work. In single process mode, higher numbers
1206 give better performance at high connection rates. However in multi-process
1207 modes, keeping a bit of fairness between processes generally is better to
1208 increase performance. This value applies individually to each listener, so
1209 that the number of processes a listener is bound to is taken into account.
1210 This value defaults to 64. In multi-process mode, it is divided by twice
1211 the number of processes the listener is bound to. Setting this value to -1
1212 completely disables the limitation. It should normally not be needed to tweak
1213 this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001214
1215tune.maxpollevents <number>
1216 Sets the maximum amount of events that can be processed at once in a call to
1217 the polling system. The default value is adapted to the operating system. It
1218 has been noticed that reducing it below 200 tends to slightly decrease
1219 latency at the expense of network bandwidth, and increasing it above 200
1220 tends to trade latency for slightly increased bandwidth.
1221
Willy Tarreau27a674e2009-08-17 07:23:33 +02001222tune.maxrewrite <number>
1223 Sets the reserved buffer space to this size in bytes. The reserved space is
1224 used for header rewriting or appending. The first reads on sockets will never
1225 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
1226 bufsize, though that does not make much sense since there are rarely large
1227 numbers of headers to add. Setting it too high prevents processing of large
1228 requests or responses. Setting it too low prevents addition of new headers
1229 to already large requests or to POST requests. It is generally wise to set it
1230 to about 1024. It is automatically readjusted to half of bufsize if it is
1231 larger than that. This means you don't have to worry about it when changing
1232 bufsize.
1233
Willy Tarreauf3045d22015-04-29 16:24:50 +02001234tune.pattern.cache-size <number>
1235 Sets the size of the pattern lookup cache to <number> entries. This is an LRU
1236 cache which reminds previous lookups and their results. It is used by ACLs
1237 and maps on slow pattern lookups, namely the ones using the "sub", "reg",
1238 "dir", "dom", "end", "bin" match methods as well as the case-insensitive
1239 strings. It applies to pattern expressions which means that it will be able
1240 to memorize the result of a lookup among all the patterns specified on a
1241 configuration line (including all those loaded from files). It automatically
1242 invalidates entries which are updated using HTTP actions or on the CLI. The
1243 default cache size is set to 10000 entries, which limits its footprint to
1244 about 5 MB on 32-bit systems and 8 MB on 64-bit systems. There is a very low
1245 risk of collision in this cache, which is in the order of the size of the
1246 cache divided by 2^64. Typically, at 10000 requests per second with the
1247 default cache size of 10000 entries, there's 1% chance that a brute force
1248 attack could cause a single collision after 60 years, or 0.1% after 6 years.
1249 This is considered much lower than the risk of a memory corruption caused by
1250 aging components. If this is not acceptable, the cache can be disabled by
1251 setting this parameter to 0.
1252
Willy Tarreaubd9a0a72011-10-23 21:14:29 +02001253tune.pipesize <number>
1254 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
1255 are the default size for the system. But sometimes when using TCP splicing,
1256 it can improve performance to increase pipe sizes, especially if it is
1257 suspected that pipes are not filled and that many calls to splice() are
1258 performed. This has an impact on the kernel's memory footprint, so this must
1259 not be changed if impacts are not understood.
1260
Willy Tarreaue803de22010-01-21 17:43:04 +01001261tune.rcvbuf.client <number>
1262tune.rcvbuf.server <number>
1263 Forces the kernel socket receive buffer size on the client or the server side
1264 to the specified value in bytes. This value applies to all TCP/HTTP frontends
1265 and backends. It should normally never be set, and the default size (0) lets
1266 the kernel autotune this value depending on the amount of available memory.
1267 However it can sometimes help to set it to very low values (eg: 4096) in
1268 order to save kernel memory by preventing it from buffering too large amounts
1269 of received data. Lower values will significantly increase CPU usage though.
1270
1271tune.sndbuf.client <number>
1272tune.sndbuf.server <number>
1273 Forces the kernel socket send buffer size on the client or the server side to
1274 the specified value in bytes. This value applies to all TCP/HTTP frontends
1275 and backends. It should normally never be set, and the default size (0) lets
1276 the kernel autotune this value depending on the amount of available memory.
1277 However it can sometimes help to set it to very low values (eg: 4096) in
1278 order to save kernel memory by preventing it from buffering too large amounts
1279 of received data. Lower values will significantly increase CPU usage though.
1280 Another use case is to prevent write timeouts with extremely slow clients due
1281 to the kernel waiting for a large part of the buffer to be read before
1282 notifying haproxy again.
1283
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001284tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +01001285 Sets the size of the global SSL session cache, in a number of blocks. A block
1286 is large enough to contain an encoded session without peer certificate.
1287 An encoded session with peer certificate is stored in multiple blocks
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001288 depending on the size of the peer certificate. A block uses approximately
Emeric Brunaf9619d2012-11-28 18:47:52 +01001289 200 bytes of memory. The default value may be forced at build time, otherwise
1290 defaults to 20000. When the cache is full, the most idle entries are purged
1291 and reassigned. Higher values reduce the occurrence of such a purge, hence
1292 the number of CPU-intensive SSL handshakes by ensuring that all users keep
1293 their session as long as possible. All entries are pre-allocated upon startup
Emeric Brun22890a12012-12-28 14:41:32 +01001294 and are shared between all processes if "nbproc" is greater than 1. Setting
1295 this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001296
Emeric Brun8dc60392014-05-09 13:52:00 +02001297tune.ssl.force-private-cache
1298 This boolean disables SSL session cache sharing between all processes. It
1299 should normally not be used since it will force many renegotiations due to
1300 clients hitting a random process. But it may be required on some operating
1301 systems where none of the SSL cache synchronization method may be used. In
1302 this case, adding a first layer of hash-based load balancing before the SSL
1303 layer might limit the impact of the lack of session sharing.
1304
Emeric Brun4f65bff2012-11-16 15:11:00 +01001305tune.ssl.lifetime <timeout>
1306 Sets how long a cached SSL session may remain valid. This time is expressed
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001307 in seconds and defaults to 300 (5 min). It is important to understand that it
Emeric Brun4f65bff2012-11-16 15:11:00 +01001308 does not guarantee that sessions will last that long, because if the cache is
1309 full, the longest idle sessions will be purged despite their configured
1310 lifetime. The real usefulness of this setting is to prevent sessions from
1311 being used for too long.
1312
Willy Tarreaubfd59462013-02-21 07:46:09 +01001313tune.ssl.maxrecord <number>
1314 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
1315 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
1316 data only once it has received a full record. With large records, it means
1317 that clients might have to download up to 16kB of data before starting to
1318 process them. Limiting the value can improve page load times on browsers
1319 located over high latency or low bandwidth networks. It is suggested to find
1320 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
1321 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
1322 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
1323 2859 gave good results during tests. Use "strace -e trace=write" to find the
Willy Tarreau7e312732014-02-12 16:35:14 +01001324 best value. Haproxy will automatically switch to this setting after an idle
1325 stream has been detected (see tune.idletimer above).
Willy Tarreaubfd59462013-02-21 07:46:09 +01001326
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001327tune.ssl.default-dh-param <number>
1328 Sets the maximum size of the Diffie-Hellman parameters used for generating
1329 the ephemeral/temporary Diffie-Hellman key in case of DHE key exchange. The
1330 final size will try to match the size of the server's RSA (or DSA) key (e.g,
1331 a 2048 bits temporary DH key for a 2048 bits RSA key), but will not exceed
1332 this maximum value. Default value if 1024. Only 1024 or higher values are
1333 allowed. Higher values will increase the CPU load, and values greater than
1334 1024 bits are not supported by Java 7 and earlier clients. This value is not
Remi Gacogne47783ef2015-05-29 15:53:22 +02001335 used if static Diffie-Hellman parameters are supplied either directly
1336 in the certificate file or by using the ssl-dh-param-file parameter.
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001337
Christopher Faulet31af49d2015-06-09 17:29:50 +02001338tune.ssl.ssl-ctx-cache-size <number>
1339 Sets the size of the cache used to store generated certificates to <number>
1340 entries. This is a LRU cache. Because generating a SSL certificate
1341 dynamically is expensive, they are cached. The default cache size is set to
1342 1000 entries.
1343
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001344tune.vars.global-max-size <size>
1345tune.vars.reqres-max-size <size>
1346tune.vars.sess-max-size <size>
1347tune.vars.txn-max-size <size>
1348 These four tunes helps to manage the allowed amount of memory used by the
1349 variables system. "global" limits the memory for all the systems. "sess" limit
1350 the memory by session, "txn" limits the memory by transaction and "reqres"
1351 limits the memory for each request or response processing. during the
1352 accounting, "sess" embbed "txn" and "txn" embed "reqres".
1353
1354 By example, we considers that "tune.vars.sess-max-size" is fixed to 100,
1355 "tune.vars.txn-max-size" is fixed to 100, "tune.vars.reqres-max-size" is
1356 also fixed to 100. If we create a variable "txn.var" that contains 100 bytes,
1357 we cannot create any more variable in the other contexts.
1358
William Lallemanda509e4c2012-11-07 16:54:34 +01001359tune.zlib.memlevel <number>
1360 Sets the memLevel parameter in zlib initialization for each session. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001361 defines how much memory should be allocated for the internal compression
William Lallemanda509e4c2012-11-07 16:54:34 +01001362 state. A value of 1 uses minimum memory but is slow and reduces compression
1363 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
1364 between 1 and 9. The default value is 8.
1365
1366tune.zlib.windowsize <number>
1367 Sets the window size (the size of the history buffer) as a parameter of the
1368 zlib initialization for each session. Larger values of this parameter result
1369 in better compression at the expense of memory usage. Can be a value between
1370 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001371
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013723.3. Debugging
1373--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02001374
1375debug
1376 Enables debug mode which dumps to stdout all exchanges, and disables forking
1377 into background. It is the equivalent of the command-line argument "-d". It
1378 should never be used in a production configuration since it may prevent full
1379 system startup.
1380
1381quiet
1382 Do not display any message during startup. It is equivalent to the command-
1383 line argument "-q".
1384
Emeric Brunf099e792010-09-27 12:05:28 +02001385
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010013863.4. Userlists
1387--------------
1388It is possible to control access to frontend/backend/listen sections or to
1389http stats by allowing only authenticated and authorized users. To do this,
1390it is required to create at least one userlist and to define users.
1391
1392userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +01001393 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001394 used to store authentication & authorization data for independent customers.
1395
1396group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +01001397 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001398 attach users to this group by using a comma separated list of names
1399 proceeded by "users" keyword.
1400
Cyril Bontéf0c60612010-02-06 14:44:47 +01001401user <username> [password|insecure-password <password>]
1402 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001403 Adds user <username> to the current userlist. Both secure (encrypted) and
1404 insecure (unencrypted) passwords can be used. Encrypted passwords are
Cyril Bonté78caf842010-03-10 22:41:43 +01001405 evaluated using the crypt(3) function so depending of the system's
1406 capabilities, different algorithms are supported. For example modern Glibc
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001407 based Linux system supports MD5, SHA-256, SHA-512 and of course classic,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001408 DES-based method of encrypting passwords.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001409
1410
1411 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +01001412 userlist L1
1413 group G1 users tiger,scott
1414 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001415
Cyril Bontéf0c60612010-02-06 14:44:47 +01001416 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
1417 user scott insecure-password elgato
1418 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001419
Cyril Bontéf0c60612010-02-06 14:44:47 +01001420 userlist L2
1421 group G1
1422 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001423
Cyril Bontéf0c60612010-02-06 14:44:47 +01001424 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
1425 user scott insecure-password elgato groups G1,G2
1426 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001427
1428 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001429
Emeric Brunf099e792010-09-27 12:05:28 +02001430
14313.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02001432----------
Emeric Brun94900952015-06-11 18:25:54 +02001433It is possible to propagate entries of any data-types in stick-tables between
1434several haproxy instances over TCP connections in a multi-master fashion. Each
1435instance pushes its local updates and insertions to remote peers. The pushed
1436values overwrite remote ones without aggregation. Interrupted exchanges are
1437automatically detected and recovered from the last known point.
1438In addition, during a soft restart, the old process connects to the new one
1439using such a TCP connection to push all its entries before the new process
1440tries to connect to other peers. That ensures very fast replication during a
1441reload, it typically takes a fraction of a second even for large tables.
1442Note that Server IDs are used to identify servers remotely, so it is important
1443that configurations look similar or at least that the same IDs are forced on
1444each server on all participants.
Emeric Brunf099e792010-09-27 12:05:28 +02001445
1446peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001447 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02001448 which is referenced by one or more stick-tables.
1449
Willy Tarreau77e4bd12015-05-01 20:02:17 +02001450disabled
1451 Disables a peers section. It disables both listening and any synchronization
1452 related to this section. This is provided to disable synchronization of stick
1453 tables without having to comment out all "peers" references.
1454
1455enable
1456 This re-enables a disabled peers section which was previously disabled.
1457
Emeric Brunf099e792010-09-27 12:05:28 +02001458peer <peername> <ip>:<port>
1459 Defines a peer inside a peers section.
1460 If <peername> is set to the local peer name (by default hostname, or forced
1461 using "-L" command line option), haproxy will listen for incoming remote peer
1462 connection on <ip>:<port>. Otherwise, <ip>:<port> defines where to connect to
1463 to join the remote peer, and <peername> is used at the protocol level to
1464 identify and validate the remote peer on the server side.
1465
1466 During a soft restart, local peer <ip>:<port> is used by the old instance to
1467 connect the new one and initiate a complete replication (teaching process).
1468
1469 It is strongly recommended to have the exact same peers declaration on all
1470 peers and to only rely on the "-L" command line argument to change the local
1471 peer name. This makes it easier to maintain coherent configuration files
1472 across all peers.
1473
William Lallemandb2f07452015-05-12 14:27:13 +02001474 You may want to reference some environment variables in the address
1475 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01001476
Cyril Bontédc4d9032012-04-08 21:57:39 +02001477 Example:
Emeric Brunf099e792010-09-27 12:05:28 +02001478 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001479 peer haproxy1 192.168.0.1:1024
1480 peer haproxy2 192.168.0.2:1024
1481 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02001482
1483 backend mybackend
1484 mode tcp
1485 balance roundrobin
1486 stick-table type ip size 20k peers mypeers
1487 stick on src
1488
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001489 server srv1 192.168.0.30:80
1490 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02001491
1492
Simon Horman51a1cf62015-02-03 13:00:44 +090014933.6. Mailers
1494------------
1495It is possible to send email alerts when the state of servers changes.
1496If configured email alerts are sent to each mailer that is configured
1497in a mailers section. Email is sent to mailers using SMTP.
1498
Pieter Baauw386a1272015-08-16 15:26:24 +02001499mailers <mailersect>
Simon Horman51a1cf62015-02-03 13:00:44 +09001500 Creates a new mailer list with the name <mailersect>. It is an
1501 independent section which is referenced by one or more proxies.
1502
1503mailer <mailername> <ip>:<port>
1504 Defines a mailer inside a mailers section.
1505
1506 Example:
1507 mailers mymailers
1508 mailer smtp1 192.168.0.1:587
1509 mailer smtp2 192.168.0.2:587
1510
1511 backend mybackend
1512 mode tcp
1513 balance roundrobin
1514
1515 email-alert mailers mymailers
1516 email-alert from test1@horms.org
1517 email-alert to test2@horms.org
1518
1519 server srv1 192.168.0.30:80
1520 server srv2 192.168.0.31:80
1521
1522
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015234. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02001524----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001525
Willy Tarreau6a06a402007-07-15 20:15:28 +02001526Proxy configuration can be located in a set of sections :
William Lallemand6e62fb62015-04-28 16:55:23 +02001527 - defaults [<name>]
Willy Tarreau6a06a402007-07-15 20:15:28 +02001528 - frontend <name>
1529 - backend <name>
1530 - listen <name>
1531
1532A "defaults" section sets default parameters for all other sections following
1533its declaration. Those default parameters are reset by the next "defaults"
1534section. See below for the list of parameters which can be set in a "defaults"
Willy Tarreau0ba27502007-12-24 16:55:16 +01001535section. The name is optional but its use is encouraged for better readability.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001536
1537A "frontend" section describes a set of listening sockets accepting client
1538connections.
1539
1540A "backend" section describes a set of servers to which the proxy will connect
1541to forward incoming connections.
1542
1543A "listen" section defines a complete proxy with its frontend and backend
1544parts combined in one section. It is generally useful for TCP-only traffic.
1545
Willy Tarreau0ba27502007-12-24 16:55:16 +01001546All proxy names must be formed from upper and lower case letters, digits,
1547'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
1548case-sensitive, which means that "www" and "WWW" are two different proxies.
1549
1550Historically, all proxy names could overlap, it just caused troubles in the
1551logs. Since the introduction of content switching, it is mandatory that two
1552proxies with overlapping capabilities (frontend/backend) have different names.
1553However, it is still permitted that a frontend and a backend share the same
1554name, as this configuration seems to be commonly encountered.
1555
1556Right now, two major proxy modes are supported : "tcp", also known as layer 4,
1557and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001558bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01001559protocol, and can interact with it by allowing, blocking, switching, adding,
1560modifying, or removing arbitrary contents in requests or responses, based on
1561arbitrary criteria.
1562
Willy Tarreau70dffda2014-01-30 03:07:23 +01001563In HTTP mode, the processing applied to requests and responses flowing over
1564a connection depends in the combination of the frontend's HTTP options and
1565the backend's. HAProxy supports 5 connection modes :
1566
1567 - KAL : keep alive ("option http-keep-alive") which is the default mode : all
1568 requests and responses are processed, and connections remain open but idle
1569 between responses and new requests.
1570
1571 - TUN: tunnel ("option http-tunnel") : this was the default mode for versions
1572 1.0 to 1.5-dev21 : only the first request and response are processed, and
1573 everything else is forwarded with no analysis at all. This mode should not
1574 be used as it creates lots of trouble with logging and HTTP processing.
1575
1576 - PCL: passive close ("option httpclose") : exactly the same as tunnel mode,
1577 but with "Connection: close" appended in both directions to try to make
1578 both ends close after the first request/response exchange.
1579
1580 - SCL: server close ("option http-server-close") : the server-facing
1581 connection is closed after the end of the response is received, but the
1582 client-facing connection remains open.
1583
1584 - FCL: forced close ("option forceclose") : the connection is actively closed
1585 after the end of the response.
1586
1587The effective mode that will be applied to a connection passing through a
1588frontend and a backend can be determined by both proxy modes according to the
1589following matrix, but in short, the modes are symmetric, keep-alive is the
1590weakest option and force close is the strongest.
1591
1592 Backend mode
1593
1594 | KAL | TUN | PCL | SCL | FCL
1595 ----+-----+-----+-----+-----+----
1596 KAL | KAL | TUN | PCL | SCL | FCL
1597 ----+-----+-----+-----+-----+----
1598 TUN | TUN | TUN | PCL | SCL | FCL
1599 Frontend ----+-----+-----+-----+-----+----
1600 mode PCL | PCL | PCL | PCL | FCL | FCL
1601 ----+-----+-----+-----+-----+----
1602 SCL | SCL | SCL | FCL | SCL | FCL
1603 ----+-----+-----+-----+-----+----
1604 FCL | FCL | FCL | FCL | FCL | FCL
1605
Willy Tarreau0ba27502007-12-24 16:55:16 +01001606
Willy Tarreau70dffda2014-01-30 03:07:23 +01001607
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016084.1. Proxy keywords matrix
1609--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001610
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001611The following list of keywords is supported. Most of them may only be used in a
1612limited set of section types. Some of them are marked as "deprecated" because
1613they are inherited from an old syntax which may be confusing or functionally
1614limited, and there are new recommended keywords to replace them. Keywords
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001615marked with "(*)" can be optionally inverted using the "no" prefix, eg. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001616option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02001617and must be disabled for a specific instance. Such options may also be prefixed
1618with "default" in order to restore default settings regardless of what has been
1619specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001620
Willy Tarreau6a06a402007-07-15 20:15:28 +02001621
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001622 keyword defaults frontend listen backend
1623------------------------------------+----------+----------+---------+---------
1624acl - X X X
Willy Tarreau294d0f02015-08-10 19:40:12 +02001625appsession - - - -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001626backlog X X X -
1627balance X - X X
1628bind - X X -
1629bind-process X X X X
1630block - X X X
1631capture cookie - X X -
1632capture request header - X X -
1633capture response header - X X -
1634clitimeout (deprecated) X X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02001635compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001636contimeout (deprecated) X - X X
1637cookie X - X X
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02001638declare capture - X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001639default-server X - X X
1640default_backend X X X -
1641description - X X X
1642disabled X X X X
1643dispatch - - X X
Simon Horman51a1cf62015-02-03 13:00:44 +09001644email-alert from X X X X
Simon Horman64e34162015-02-06 11:11:57 +09001645email-alert level X X X X
Simon Horman51a1cf62015-02-03 13:00:44 +09001646email-alert mailers X X X X
1647email-alert myhostname X X X X
1648email-alert to X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001649enabled X X X X
1650errorfile X X X X
1651errorloc X X X X
1652errorloc302 X X X X
1653-- keyword -------------------------- defaults - frontend - listen -- backend -
1654errorloc303 X X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001655force-persist - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001656fullconn X - X X
1657grace X X X X
1658hash-type X - X X
1659http-check disable-on-404 X - X X
Willy Tarreaubd741542010-03-16 18:46:54 +01001660http-check expect - - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02001661http-check send-state X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001662http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02001663http-response - X X X
Willy Tarreau30631952015-08-06 15:05:24 +02001664http-reuse X - X X
Baptiste Assmann2c42ef52013-10-09 21:57:02 +02001665http-send-name-header - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001666id - X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001667ignore-persist - X X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001668load-server-state-from-file X - X X
William Lallemand0f99e342011-10-12 17:50:54 +02001669log (*) X X X X
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01001670log-format X X X -
Dragan Dosen7ad31542015-09-28 17:16:47 +02001671log-format-sd X X X -
Willy Tarreau094af4e2015-01-07 15:03:42 +01001672log-tag X X X X
Willy Tarreauc35362a2014-04-25 13:58:37 +02001673max-keep-alive-queue X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001674maxconn X X X -
1675mode X X X X
1676monitor fail - X X -
1677monitor-net X X X -
1678monitor-uri X X X -
1679option abortonclose (*) X - X X
1680option accept-invalid-http-request (*) X X X -
1681option accept-invalid-http-response (*) X - X X
1682option allbackups (*) X - X X
1683option checkcache (*) X - X X
1684option clitcpka (*) X X X -
1685option contstats (*) X X X -
1686option dontlog-normal (*) X X X -
1687option dontlognull (*) X X X -
1688option forceclose (*) X X X X
1689-- keyword -------------------------- defaults - frontend - listen -- backend -
1690option forwardfor X X X X
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02001691option http-buffer-request (*) X X X X
Willy Tarreau82649f92015-05-01 22:40:51 +02001692option http-ignore-probes (*) X X X -
Willy Tarreau16bfb022010-01-16 19:48:41 +01001693option http-keep-alive (*) X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02001694option http-no-delay (*) X X X X
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02001695option http-pretend-keepalive (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001696option http-server-close (*) X X X X
Willy Tarreau02bce8b2014-01-30 00:15:28 +01001697option http-tunnel (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001698option http-use-proxy-header (*) X X X -
1699option httpchk X - X X
1700option httpclose (*) X X X X
1701option httplog X X X X
1702option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001703option independent-streams (*) X X X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02001704option ldap-check X - X X
Simon Horman98637e52014-06-20 12:30:16 +09001705option external-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001706option log-health-checks (*) X - X X
1707option log-separate-errors (*) X X X -
1708option logasap (*) X X X -
1709option mysql-check X - X X
Rauf Kuliyev38b41562011-01-04 15:14:13 +01001710option pgsql-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001711option nolinger (*) X X X X
1712option originalto X X X X
1713option persist (*) X - X X
1714option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02001715option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001716option smtpchk X - X X
1717option socket-stats (*) X X X -
1718option splice-auto (*) X X X X
1719option splice-request (*) X X X X
1720option splice-response (*) X X X X
1721option srvtcpka (*) X - X X
1722option ssl-hello-chk X - X X
1723-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreaued179852013-12-16 01:07:00 +01001724option tcp-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001725option tcp-smart-accept (*) X X X -
1726option tcp-smart-connect (*) X - X X
1727option tcpka X X X X
1728option tcplog X X X X
1729option transparent (*) X - X X
Simon Horman98637e52014-06-20 12:30:16 +09001730external-check command X - X X
1731external-check path X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001732persist rdp-cookie X - X X
1733rate-limit sessions X X X -
1734redirect - X X X
1735redisp (deprecated) X - X X
1736redispatch (deprecated) X - X X
1737reqadd - X X X
1738reqallow - X X X
1739reqdel - X X X
1740reqdeny - X X X
1741reqiallow - X X X
1742reqidel - X X X
1743reqideny - X X X
1744reqipass - X X X
1745reqirep - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001746reqitarpit - X X X
1747reqpass - X X X
1748reqrep - X X X
1749-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001750reqtarpit - X X X
1751retries X - X X
1752rspadd - X X X
1753rspdel - X X X
1754rspdeny - X X X
1755rspidel - X X X
1756rspideny - X X X
1757rspirep - X X X
1758rsprep - X X X
1759server - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001760server-state-file-name X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001761source X - X X
1762srvtimeout (deprecated) X - X X
Cyril Bonté66c327d2010-10-12 00:14:37 +02001763stats admin - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001764stats auth X - X X
1765stats enable X - X X
1766stats hide-version X - X X
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02001767stats http-request - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001768stats realm X - X X
1769stats refresh X - X X
1770stats scope X - X X
1771stats show-desc X - X X
1772stats show-legends X - X X
1773stats show-node X - X X
1774stats uri X - X X
1775-- keyword -------------------------- defaults - frontend - listen -- backend -
1776stick match - - X X
1777stick on - - X X
1778stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02001779stick store-response - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001780stick-table - - X X
Willy Tarreau938c7fe2014-04-25 14:21:39 +02001781tcp-check connect - - X X
1782tcp-check expect - - X X
1783tcp-check send - - X X
1784tcp-check send-binary - - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02001785tcp-request connection - X X -
1786tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02001787tcp-request inspect-delay - X X X
Emeric Brun0a3b67f2010-09-24 15:34:53 +02001788tcp-response content - - X X
1789tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001790timeout check X - X X
1791timeout client X X X -
Willy Tarreau05cdd962014-05-10 14:30:07 +02001792timeout client-fin X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001793timeout clitimeout (deprecated) X X X -
1794timeout connect X - X X
1795timeout contimeout (deprecated) X - X X
1796timeout http-keep-alive X X X X
1797timeout http-request X X X X
1798timeout queue X - X X
1799timeout server X - X X
Willy Tarreau05cdd962014-05-10 14:30:07 +02001800timeout server-fin X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001801timeout srvtimeout (deprecated) X - X X
1802timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02001803timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001804transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01001805unique-id-format X X X -
1806unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001807use_backend - X X -
Willy Tarreau4a5cade2012-04-05 21:09:48 +02001808use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001809------------------------------------+----------+----------+---------+---------
1810 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02001811
Willy Tarreau0ba27502007-12-24 16:55:16 +01001812
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018134.2. Alphabetically sorted keywords reference
1814---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001815
1816This section provides a description of each keyword and its usage.
1817
1818
1819acl <aclname> <criterion> [flags] [operator] <value> ...
1820 Declare or complete an access list.
1821 May be used in sections : defaults | frontend | listen | backend
1822 no | yes | yes | yes
1823 Example:
1824 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
1825 acl invalid_src src_port 0:1023
1826 acl local_dst hdr(host) -i localhost
1827
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001828 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001829
1830
Cyril Bontéb21570a2009-11-29 20:04:48 +01001831appsession <cookie> len <length> timeout <holdtime>
1832 [request-learn] [prefix] [mode <path-parameters|query-string>]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001833 Define session stickiness on an existing application cookie.
1834 May be used in sections : defaults | frontend | listen | backend
1835 no | no | yes | yes
1836 Arguments :
1837 <cookie> this is the name of the cookie used by the application and which
1838 HAProxy will have to learn for each new session.
1839
Cyril Bontéb21570a2009-11-29 20:04:48 +01001840 <length> this is the max number of characters that will be memorized and
Willy Tarreau0ba27502007-12-24 16:55:16 +01001841 checked in each cookie value.
1842
1843 <holdtime> this is the time after which the cookie will be removed from
1844 memory if unused. If no unit is specified, this time is in
1845 milliseconds.
1846
Cyril Bontébf47aeb2009-10-15 00:15:40 +02001847 request-learn
1848 If this option is specified, then haproxy will be able to learn
1849 the cookie found in the request in case the server does not
1850 specify any in response. This is typically what happens with
1851 PHPSESSID cookies, or when haproxy's session expires before
1852 the application's session and the correct server is selected.
1853 It is recommended to specify this option to improve reliability.
1854
Cyril Bontéb21570a2009-11-29 20:04:48 +01001855 prefix When this option is specified, haproxy will match on the cookie
1856 prefix (or URL parameter prefix). The appsession value is the
1857 data following this prefix.
1858
1859 Example :
1860 appsession ASPSESSIONID len 64 timeout 3h prefix
1861
1862 This will match the cookie ASPSESSIONIDXXXX=XXXXX,
1863 the appsession value will be XXXX=XXXXX.
1864
1865 mode This option allows to change the URL parser mode.
1866 2 modes are currently supported :
1867 - path-parameters :
1868 The parser looks for the appsession in the path parameters
1869 part (each parameter is separated by a semi-colon), which is
1870 convenient for JSESSIONID for example.
1871 This is the default mode if the option is not set.
1872 - query-string :
1873 In this mode, the parser will look for the appsession in the
1874 query string.
1875
Willy Tarreau294d0f02015-08-10 19:40:12 +02001876 As of version 1.6, appsessions was removed. It is more flexible and more
1877 convenient to use stick-tables instead, and stick-tables support multi-master
1878 replication and data conservation across reloads, which appsessions did not.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001879
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01001880 See also : "cookie", "capture cookie", "balance", "stick", "stick-table",
1881 "ignore-persist", "nbproc" and "bind-process".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001882
1883
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01001884backlog <conns>
1885 Give hints to the system about the approximate listen backlog desired size
1886 May be used in sections : defaults | frontend | listen | backend
1887 yes | yes | yes | no
1888 Arguments :
1889 <conns> is the number of pending connections. Depending on the operating
1890 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02001891 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01001892
1893 In order to protect against SYN flood attacks, one solution is to increase
1894 the system's SYN backlog size. Depending on the system, sometimes it is just
1895 tunable via a system parameter, sometimes it is not adjustable at all, and
1896 sometimes the system relies on hints given by the application at the time of
1897 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
1898 to the listen() syscall. On systems which can make use of this value, it can
1899 sometimes be useful to be able to specify a different value, hence this
1900 backlog parameter.
1901
1902 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
1903 used as a hint and the system accepts up to the smallest greater power of
1904 two, and never more than some limits (usually 32768).
1905
1906 See also : "maxconn" and the target operating system's tuning guide.
1907
1908
Willy Tarreau0ba27502007-12-24 16:55:16 +01001909balance <algorithm> [ <arguments> ]
Willy Tarreau226071e2014-04-10 11:55:45 +02001910balance url_param <param> [check_post]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001911 Define the load balancing algorithm to be used in a backend.
1912 May be used in sections : defaults | frontend | listen | backend
1913 yes | no | yes | yes
1914 Arguments :
1915 <algorithm> is the algorithm used to select a server when doing load
1916 balancing. This only applies when no persistence information
1917 is available, or when a connection is redispatched to another
1918 server. <algorithm> may be one of the following :
1919
1920 roundrobin Each server is used in turns, according to their weights.
1921 This is the smoothest and fairest algorithm when the server's
1922 processing time remains equally distributed. This algorithm
1923 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02001924 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08001925 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02001926 large farms, when a server becomes up after having been down
1927 for a very short time, it may sometimes take a few hundreds
1928 requests for it to be re-integrated into the farm and start
1929 receiving traffic. This is normal, though very rare. It is
1930 indicated here in case you would have the chance to observe
1931 it, so that you don't worry.
1932
1933 static-rr Each server is used in turns, according to their weights.
1934 This algorithm is as similar to roundrobin except that it is
1935 static, which means that changing a server's weight on the
1936 fly will have no effect. On the other hand, it has no design
1937 limitation on the number of servers, and when a server goes
1938 up, it is always immediately reintroduced into the farm, once
1939 the full map is recomputed. It also uses slightly less CPU to
1940 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01001941
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01001942 leastconn The server with the lowest number of connections receives the
1943 connection. Round-robin is performed within groups of servers
1944 of the same load to ensure that all servers will be used. Use
1945 of this algorithm is recommended where very long sessions are
1946 expected, such as LDAP, SQL, TSE, etc... but is not very well
1947 suited for protocols using short sessions such as HTTP. This
1948 algorithm is dynamic, which means that server weights may be
1949 adjusted on the fly for slow starts for instance.
1950
Willy Tarreauf09c6602012-02-13 17:12:08 +01001951 first The first server with available connection slots receives the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001952 connection. The servers are chosen from the lowest numeric
Willy Tarreauf09c6602012-02-13 17:12:08 +01001953 identifier to the highest (see server parameter "id"), which
1954 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02001955 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01001956 not make sense to use this algorithm without setting maxconn.
1957 The purpose of this algorithm is to always use the smallest
1958 number of servers so that extra servers can be powered off
1959 during non-intensive hours. This algorithm ignores the server
1960 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02001961 or IMAP than HTTP, though it can be useful there too. In
1962 order to use this algorithm efficiently, it is recommended
1963 that a cloud controller regularly checks server usage to turn
1964 them off when unused, and regularly checks backend queue to
1965 turn new servers on when the queue inflates. Alternatively,
1966 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01001967
Willy Tarreau0ba27502007-12-24 16:55:16 +01001968 source The source IP address is hashed and divided by the total
1969 weight of the running servers to designate which server will
1970 receive the request. This ensures that the same client IP
1971 address will always reach the same server as long as no
1972 server goes down or up. If the hash result changes due to the
1973 number of running servers changing, many clients will be
1974 directed to a different server. This algorithm is generally
1975 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001976 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01001977 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001978 static by default, which means that changing a server's
1979 weight on the fly will have no effect, but this can be
1980 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001981
Oskar Stolc8dc41842012-05-19 10:19:54 +01001982 uri This algorithm hashes either the left part of the URI (before
1983 the question mark) or the whole URI (if the "whole" parameter
1984 is present) and divides the hash value by the total weight of
1985 the running servers. The result designates which server will
1986 receive the request. This ensures that the same URI will
1987 always be directed to the same server as long as no server
1988 goes up or down. This is used with proxy caches and
1989 anti-virus proxies in order to maximize the cache hit rate.
1990 Note that this algorithm may only be used in an HTTP backend.
1991 This algorithm is static by default, which means that
1992 changing a server's weight on the fly will have no effect,
1993 but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001994
Oskar Stolc8dc41842012-05-19 10:19:54 +01001995 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02001996 "depth", both followed by a positive integer number. These
1997 options may be helpful when it is needed to balance servers
1998 based on the beginning of the URI only. The "len" parameter
1999 indicates that the algorithm should only consider that many
2000 characters at the beginning of the URI to compute the hash.
2001 Note that having "len" set to 1 rarely makes sense since most
2002 URIs start with a leading "/".
2003
2004 The "depth" parameter indicates the maximum directory depth
2005 to be used to compute the hash. One level is counted for each
2006 slash in the request. If both parameters are specified, the
2007 evaluation stops when either is reached.
2008
Willy Tarreau0ba27502007-12-24 16:55:16 +01002009 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002010 the query string of each HTTP GET request.
2011
2012 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02002013 request entity will be searched for the parameter argument,
2014 when it is not found in a query string after a question mark
Willy Tarreau226071e2014-04-10 11:55:45 +02002015 ('?') in the URL. The message body will only start to be
2016 analyzed once either the advertised amount of data has been
2017 received or the request buffer is full. In the unlikely event
2018 that chunked encoding is used, only the first chunk is
Cyril Bontédc4d9032012-04-08 21:57:39 +02002019 scanned. Parameter values separated by a chunk boundary, may
Willy Tarreau226071e2014-04-10 11:55:45 +02002020 be randomly balanced if at all. This keyword used to support
2021 an optional <max_wait> parameter which is now ignored.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002022
2023 If the parameter is found followed by an equal sign ('=') and
2024 a value, then the value is hashed and divided by the total
2025 weight of the running servers. The result designates which
2026 server will receive the request.
2027
2028 This is used to track user identifiers in requests and ensure
2029 that a same user ID will always be sent to the same server as
2030 long as no server goes up or down. If no value is found or if
2031 the parameter is not found, then a round robin algorithm is
2032 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002033 backend. This algorithm is static by default, which means
2034 that changing a server's weight on the fly will have no
2035 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002036
Cyril Bontédc4d9032012-04-08 21:57:39 +02002037 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
2038 request. Just as with the equivalent ACL 'hdr()' function,
2039 the header name in parenthesis is not case sensitive. If the
2040 header is absent or if it does not contain any value, the
2041 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01002042
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002043 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01002044 reducing the hash algorithm to the main domain part with some
2045 specific headers such as 'Host'. For instance, in the Host
2046 value "haproxy.1wt.eu", only "1wt" will be considered.
2047
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002048 This algorithm is static by default, which means that
2049 changing a server's weight on the fly will have no effect,
2050 but this can be changed using "hash-type".
2051
Emeric Brun736aa232009-06-30 17:56:00 +02002052 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02002053 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02002054 The RDP cookie <name> (or "mstshash" if omitted) will be
2055 looked up and hashed for each incoming TCP request. Just as
2056 with the equivalent ACL 'req_rdp_cookie()' function, the name
2057 is not case-sensitive. This mechanism is useful as a degraded
2058 persistence mode, as it makes it possible to always send the
2059 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002060 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02002061 used instead.
2062
2063 Note that for this to work, the frontend must ensure that an
2064 RDP cookie is already present in the request buffer. For this
2065 you must use 'tcp-request content accept' rule combined with
2066 a 'req_rdp_cookie_cnt' ACL.
2067
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002068 This algorithm is static by default, which means that
2069 changing a server's weight on the fly will have no effect,
2070 but this can be changed using "hash-type".
2071
Cyril Bontédc4d9032012-04-08 21:57:39 +02002072 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09002073
Willy Tarreau0ba27502007-12-24 16:55:16 +01002074 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002075 algorithms. Right now, only "url_param" and "uri" support an
2076 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002077
Willy Tarreau3cd9af22009-03-15 14:06:41 +01002078 The load balancing algorithm of a backend is set to roundrobin when no other
2079 algorithm, mode nor option have been set. The algorithm may only be set once
2080 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002081
2082 Examples :
2083 balance roundrobin
2084 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002085 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01002086 balance hdr(User-Agent)
2087 balance hdr(host)
2088 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002089
2090 Note: the following caveats and limitations on using the "check_post"
2091 extension with "url_param" must be considered :
2092
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002093 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002094 to determine if the parameters will be found in the body or entity which
2095 may contain binary data. Therefore another method may be required to
2096 restrict consideration of POST requests that have no URL parameters in
2097 the body. (see acl reqideny http_end)
2098
2099 - using a <max_wait> value larger than the request buffer size does not
2100 make sense and is useless. The buffer size is set at build time, and
2101 defaults to 16 kB.
2102
2103 - Content-Encoding is not supported, the parameter search will probably
2104 fail; and load balancing will fall back to Round Robin.
2105
2106 - Expect: 100-continue is not supported, load balancing will fall back to
2107 Round Robin.
2108
2109 - Transfer-Encoding (RFC2616 3.6.1) is only supported in the first chunk.
2110 If the entire parameter value is not present in the first chunk, the
2111 selection of server is undefined (actually, defined by how little
2112 actually appeared in the first chunk).
2113
2114 - This feature does not support generation of a 100, 411 or 501 response.
2115
2116 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002117 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002118 white space or control characters are found, indicating the end of what
2119 might be a URL parameter list. This is probably not a concern with SGML
2120 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002121
Willy Tarreau294d0f02015-08-10 19:40:12 +02002122 See also : "dispatch", "cookie", "transparent", "hash-type" and "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002123
2124
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002125bind [<address>]:<port_range> [, ...] [param*]
2126bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002127 Define one or several listening addresses and/or ports in a frontend.
2128 May be used in sections : defaults | frontend | listen | backend
2129 no | yes | yes | no
2130 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002131 <address> is optional and can be a host name, an IPv4 address, an IPv6
2132 address, or '*'. It designates the address the frontend will
2133 listen on. If unset, all IPv4 addresses of the system will be
2134 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01002135 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01002136 Optionally, an address family prefix may be used before the
2137 address to force the family regardless of the address format,
2138 which can be useful to specify a path to a unix socket with
2139 no slash ('/'). Currently supported prefixes are :
2140 - 'ipv4@' -> address is always IPv4
2141 - 'ipv6@' -> address is always IPv6
2142 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau70f72e02014-07-08 00:37:50 +02002143 - 'abns@' -> address is in abstract namespace (Linux only).
2144 Note: since abstract sockets are not "rebindable", they
2145 do not cope well with multi-process mode during
2146 soft-restart, so it is better to avoid them if
2147 nbproc is greater than 1. The effect is that if the
2148 new process fails to start, only one of the old ones
2149 will be able to rebind to the socket.
Willy Tarreau40aa0702013-03-10 23:51:38 +01002150 - 'fd@<n>' -> use file descriptor <n> inherited from the
2151 parent. The fd must be bound and may or may not already
2152 be listening.
William Lallemandb2f07452015-05-12 14:27:13 +02002153 You may want to reference some environment variables in the
2154 address parameter, see section 2.3 about environment
2155 variables.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002156
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002157 <port_range> is either a unique TCP port, or a port range for which the
2158 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002159 above. The port is mandatory for TCP listeners. Note that in
2160 the case of an IPv6 address, the port is always the number
2161 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002162 - a numerical port (ex: '80')
2163 - a dash-delimited ports range explicitly stating the lower
2164 and upper bounds (ex: '2000-2100') which are included in
2165 the range.
2166
2167 Particular care must be taken against port ranges, because
2168 every <address:port> couple consumes one socket (= a file
2169 descriptor), so it's easy to consume lots of descriptors
2170 with a simple range, and to run out of sockets. Also, each
2171 <address:port> couple must be used only once among all
2172 instances running on a same system. Please note that binding
2173 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002174 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002175 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002176
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002177 <path> is a UNIX socket path beginning with a slash ('/'). This is
2178 alternative to the TCP listening port. Haproxy will then
2179 receive UNIX connections on the socket located at this place.
2180 The path must begin with a slash and by default is absolute.
2181 It can be relative to the prefix defined by "unix-bind" in
2182 the global section. Note that the total length of the prefix
2183 followed by the socket path cannot exceed some system limits
2184 for UNIX sockets, which commonly are set to 107 characters.
2185
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002186 <param*> is a list of parameters common to all sockets declared on the
2187 same line. These numerous parameters depend on OS and build
2188 options and have a complete section dedicated to them. Please
2189 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002190
Willy Tarreau0ba27502007-12-24 16:55:16 +01002191 It is possible to specify a list of address:port combinations delimited by
2192 commas. The frontend will then listen on all of these addresses. There is no
2193 fixed limit to the number of addresses and ports which can be listened on in
2194 a frontend, as well as there is no limit to the number of "bind" statements
2195 in a frontend.
2196
2197 Example :
2198 listen http_proxy
2199 bind :80,:443
2200 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002201 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01002202
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002203 listen http_https_proxy
2204 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02002205 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002206
Willy Tarreau24709282013-03-10 21:32:12 +01002207 listen http_https_proxy_explicit
2208 bind ipv6@:80
2209 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
2210 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
2211
Willy Tarreaudad36a32013-03-11 01:20:04 +01002212 listen external_bind_app1
William Lallemandb2f07452015-05-12 14:27:13 +02002213 bind "fd@${FD_APP1}"
Willy Tarreaudad36a32013-03-11 01:20:04 +01002214
Willy Tarreau55dcaf62015-09-27 15:03:15 +02002215 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
2216 sun_path length is used for the address length. Some other programs
2217 such as socat use the string length only by default. Pass the option
2218 ",unix-tightsocklen=0" to any abstract socket definition in socat to
2219 make it compatible with HAProxy's.
2220
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002221 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002222 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002223
2224
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002225bind-process [ all | odd | even | <number 1-64>[-<number 1-64>] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002226 Limit visibility of an instance to a certain set of processes numbers.
2227 May be used in sections : defaults | frontend | listen | backend
2228 yes | yes | yes | yes
2229 Arguments :
2230 all All process will see this instance. This is the default. It
2231 may be used to override a default value.
2232
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002233 odd This instance will be enabled on processes 1,3,5,...63. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002234 option may be combined with other numbers.
2235
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002236 even This instance will be enabled on processes 2,4,6,...64. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002237 option may be combined with other numbers. Do not use it
2238 with less than 2 processes otherwise some instances might be
2239 missing from all processes.
2240
Willy Tarreau110ecc12012-11-15 17:50:01 +01002241 number The instance will be enabled on this process number or range,
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002242 whose values must all be between 1 and 32 or 64 depending on
Willy Tarreau102df612014-05-07 23:56:38 +02002243 the machine's word size. If a proxy is bound to process
2244 numbers greater than the configured global.nbproc, it will
2245 either be forced to process #1 if a single process was
2246 specified, or to all processes otherwise.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002247
2248 This keyword limits binding of certain instances to certain processes. This
2249 is useful in order not to have too many processes listening to the same
2250 ports. For instance, on a dual-core machine, it might make sense to set
2251 'nbproc 2' in the global section, then distributes the listeners among 'odd'
2252 and 'even' instances.
2253
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002254 At the moment, it is not possible to reference more than 32 or 64 processes
2255 using this keyword, but this should be more than enough for most setups.
2256 Please note that 'all' really means all processes regardless of the machine's
2257 word size, and is not limited to the first 32 or 64.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002258
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002259 Each "bind" line may further be limited to a subset of the proxy's processes,
2260 please consult the "process" bind keyword in section 5.1.
2261
Willy Tarreaub369a042014-09-16 13:21:03 +02002262 When a frontend has no explicit "bind-process" line, it tries to bind to all
2263 the processes referenced by its "bind" lines. That means that frontends can
2264 easily adapt to their listeners' processes.
2265
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002266 If some backends are referenced by frontends bound to other processes, the
2267 backend automatically inherits the frontend's processes.
2268
2269 Example :
2270 listen app_ip1
2271 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002272 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002273
2274 listen app_ip2
2275 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002276 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002277
2278 listen management
2279 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002280 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002281
Willy Tarreau110ecc12012-11-15 17:50:01 +01002282 listen management
2283 bind 10.0.0.4:80
2284 bind-process 1-4
2285
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002286 See also : "nbproc" in global section, and "process" in section 5.1.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002287
2288
Willy Tarreau0ba27502007-12-24 16:55:16 +01002289block { if | unless } <condition>
2290 Block a layer 7 request if/unless a condition is matched
2291 May be used in sections : defaults | frontend | listen | backend
2292 no | yes | yes | yes
2293
2294 The HTTP request will be blocked very early in the layer 7 processing
2295 if/unless <condition> is matched. A 403 error will be returned if the request
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002296 is blocked. The condition has to reference ACLs (see section 7). This is
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02002297 typically used to deny access to certain sensitive resources if some
Willy Tarreau0ba27502007-12-24 16:55:16 +01002298 conditions are met or not met. There is no fixed limit to the number of
2299 "block" statements per instance.
2300
2301 Example:
2302 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2303 acl invalid_src src_port 0:1023
2304 acl local_dst hdr(host) -i localhost
2305 block if invalid_src || local_dst
2306
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002307 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002308
2309
2310capture cookie <name> len <length>
2311 Capture and log a cookie in the request and in the response.
2312 May be used in sections : defaults | frontend | listen | backend
2313 no | yes | yes | no
2314 Arguments :
2315 <name> is the beginning of the name of the cookie to capture. In order
2316 to match the exact name, simply suffix the name with an equal
2317 sign ('='). The full name will appear in the logs, which is
2318 useful with application servers which adjust both the cookie name
2319 and value (eg: ASPSESSIONXXXXX).
2320
2321 <length> is the maximum number of characters to report in the logs, which
2322 include the cookie name, the equal sign and the value, all in the
2323 standard "name=value" form. The string will be truncated on the
2324 right if it exceeds <length>.
2325
2326 Only the first cookie is captured. Both the "cookie" request headers and the
2327 "set-cookie" response headers are monitored. This is particularly useful to
2328 check for application bugs causing session crossing or stealing between
2329 users, because generally the user's cookies can only change on a login page.
2330
2331 When the cookie was not presented by the client, the associated log column
2332 will report "-". When a request does not cause a cookie to be assigned by the
2333 server, a "-" is reported in the response column.
2334
2335 The capture is performed in the frontend only because it is necessary that
2336 the log format does not change for a given frontend depending on the
2337 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01002338 "capture cookie" statement in a frontend. The maximum capture length is set
2339 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
2340 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002341
2342 Example:
2343 capture cookie ASPSESSION len 32
2344
2345 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002346 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002347
2348
2349capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01002350 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002351 May be used in sections : defaults | frontend | listen | backend
2352 no | yes | yes | no
2353 Arguments :
2354 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002355 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01002356 appear in the requests, with the first letter of each word in
2357 upper case. The header name will not appear in the logs, only the
2358 value is reported, but the position in the logs is respected.
2359
2360 <length> is the maximum number of characters to extract from the value and
2361 report in the logs. The string will be truncated on the right if
2362 it exceeds <length>.
2363
Willy Tarreau4460d032012-11-21 23:37:37 +01002364 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01002365 value will be added to the logs between braces ('{}'). If multiple headers
2366 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002367 in the same order they were declared in the configuration. Non-existent
2368 headers will be logged just as an empty string. Common uses for request
2369 header captures include the "Host" field in virtual hosting environments, the
2370 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002371 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002372 environments to find where the request came from.
2373
2374 Note that when capturing headers such as "User-agent", some spaces may be
2375 logged, making the log analysis more difficult. Thus be careful about what
2376 you log if you know your log parser is not smart enough to rely on the
2377 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002378
Willy Tarreau0900abb2012-11-22 00:21:46 +01002379 There is no limit to the number of captured request headers nor to their
2380 length, though it is wise to keep them low to limit memory usage per session.
2381 In order to keep log format consistent for a same frontend, header captures
2382 can only be declared in a frontend. It is not possible to specify a capture
2383 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002384
2385 Example:
2386 capture request header Host len 15
2387 capture request header X-Forwarded-For len 15
2388 capture request header Referrer len 15
2389
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002390 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01002391 about logging.
2392
2393
2394capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01002395 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002396 May be used in sections : defaults | frontend | listen | backend
2397 no | yes | yes | no
2398 Arguments :
2399 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002400 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01002401 appear in the response, with the first letter of each word in
2402 upper case. The header name will not appear in the logs, only the
2403 value is reported, but the position in the logs is respected.
2404
2405 <length> is the maximum number of characters to extract from the value and
2406 report in the logs. The string will be truncated on the right if
2407 it exceeds <length>.
2408
Willy Tarreau4460d032012-11-21 23:37:37 +01002409 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01002410 result will be added to the logs between braces ('{}') after the captured
2411 request headers. If multiple headers are captured, they will be delimited by
2412 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002413 the configuration. Non-existent headers will be logged just as an empty
2414 string. Common uses for response header captures include the "Content-length"
2415 header which indicates how many bytes are expected to be returned, the
2416 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002417
Willy Tarreau0900abb2012-11-22 00:21:46 +01002418 There is no limit to the number of captured response headers nor to their
2419 length, though it is wise to keep them low to limit memory usage per session.
2420 In order to keep log format consistent for a same frontend, header captures
2421 can only be declared in a frontend. It is not possible to specify a capture
2422 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002423
2424 Example:
2425 capture response header Content-length len 9
2426 capture response header Location len 15
2427
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002428 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01002429 about logging.
2430
2431
Cyril Bontéf0c60612010-02-06 14:44:47 +01002432clitimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002433 Set the maximum inactivity time on the client side.
2434 May be used in sections : defaults | frontend | listen | backend
2435 yes | yes | yes | no
2436 Arguments :
2437 <timeout> is the timeout value is specified in milliseconds by default, but
2438 can be in any other unit if the number is suffixed by the unit,
2439 as explained at the top of this document.
2440
2441 The inactivity timeout applies when the client is expected to acknowledge or
2442 send data. In HTTP mode, this timeout is particularly important to consider
2443 during the first phase, when the client sends the request, and during the
2444 response while it is reading data sent by the server. The value is specified
2445 in milliseconds by default, but can be in any other unit if the number is
2446 suffixed by the unit, as specified at the top of this document. In TCP mode
2447 (and to a lesser extent, in HTTP mode), it is highly recommended that the
2448 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002449 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01002450 losses by specifying timeouts that are slightly above multiples of 3 seconds
2451 (eg: 4 or 5 seconds).
2452
2453 This parameter is specific to frontends, but can be specified once for all in
2454 "defaults" sections. This is in fact one of the easiest solutions not to
2455 forget about it. An unspecified timeout results in an infinite timeout, which
2456 is not recommended. Such a usage is accepted and works but reports a warning
2457 during startup because it may results in accumulation of expired sessions in
2458 the system if the system's timeouts are not configured either.
2459
2460 This parameter is provided for compatibility but is currently deprecated.
2461 Please use "timeout client" instead.
2462
Willy Tarreau036fae02008-01-06 13:24:40 +01002463 See also : "timeout client", "timeout http-request", "timeout server", and
2464 "srvtimeout".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002465
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002466compression algo <algorithm> ...
2467compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02002468compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02002469 Enable HTTP compression.
2470 May be used in sections : defaults | frontend | listen | backend
2471 yes | yes | yes | yes
2472 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002473 algo is followed by the list of supported compression algorithms.
2474 type is followed by the list of MIME types that will be compressed.
2475 offload makes haproxy work as a compression offloader only (see notes).
2476
2477 The currently supported algorithms are :
Willy Tarreauc91840a2015-03-28 17:00:39 +01002478 identity this is mostly for debugging, and it was useful for developing
2479 the compression feature. Identity does not apply any change on
2480 data.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002481
Willy Tarreauc91840a2015-03-28 17:00:39 +01002482 gzip applies gzip compression. This setting is only available when
2483 support for zlib was built in.
2484
2485 deflate same as "gzip", but with deflate algorithm and zlib format.
2486 Note that this algorithm has ambiguous support on many
2487 browsers and no support at all from recent ones. It is
2488 strongly recommended not to use it for anything else than
2489 experimentation. This setting is only available when support
2490 for zlib was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002491
Willy Tarreauc91840a2015-03-28 17:00:39 +01002492 raw-deflate same as "deflate" without the zlib wrapper, and used as an
2493 alternative when the browser wants "deflate". All major
2494 browsers understand it and despite violating the standards,
2495 it is known to work better than "deflate", at least on MSIE
2496 and some versions of Safari. Do not use it in conjunction
2497 with "deflate", use either one or the other since both react
2498 to the same Accept-Encoding token. This setting is only
2499 available when support for zlib was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002500
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04002501 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002502 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04002503 If backend servers support HTTP compression, these directives
2504 will be no-op: haproxy will see the compressed response and will not
2505 compress again. If backend servers do not support HTTP compression and
2506 there is Accept-Encoding header in request, haproxy will compress the
2507 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02002508
2509 The "offload" setting makes haproxy remove the Accept-Encoding header to
2510 prevent backend servers from compressing responses. It is strongly
2511 recommended not to do this because this means that all the compression work
2512 will be done on the single point where haproxy is located. However in some
2513 deployment scenarios, haproxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04002514 with broken HTTP compression implementation which can't be turned off.
2515 In that case haproxy can be used to prevent that gateway from emitting
2516 invalid payloads. In this case, simply removing the header in the
2517 configuration does not work because it applies before the header is parsed,
2518 so that prevents haproxy from compressing. The "offload" setting should
Willy Tarreauffea9fd2014-07-12 16:37:02 +02002519 then be used for such scenarios. Note: for now, the "offload" setting is
2520 ignored when set in a defaults section.
William Lallemand82fe75c2012-10-23 10:25:10 +02002521
William Lallemand05097442012-11-20 12:14:28 +01002522 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002523 * the request does not advertise a supported compression algorithm in the
2524 "Accept-Encoding" header
2525 * the response message is not HTTP/1.1
William Lallemandd3002612012-11-26 14:34:47 +01002526 * HTTP status code is not 200
William Lallemand8bb4e342013-12-10 17:28:48 +01002527 * response header "Transfer-Encoding" contains "chunked" (Temporary
2528 Workaround)
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002529 * response contain neither a "Content-Length" header nor a
2530 "Transfer-Encoding" whose last value is "chunked"
2531 * response contains a "Content-Type" header whose first value starts with
2532 "multipart"
2533 * the response contains the "no-transform" value in the "Cache-control"
2534 header
2535 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
2536 and later
2537 * The response contains a "Content-Encoding" header, indicating that the
2538 response is already compressed (see compression offload)
William Lallemand05097442012-11-20 12:14:28 +01002539
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002540 Note: The compression does not rewrite Etag headers, and does not emit the
2541 Warning header.
William Lallemand05097442012-11-20 12:14:28 +01002542
William Lallemand82fe75c2012-10-23 10:25:10 +02002543 Examples :
2544 compression algo gzip
2545 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01002546
Cyril Bontéf0c60612010-02-06 14:44:47 +01002547contimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002548 Set the maximum time to wait for a connection attempt to a server to succeed.
2549 May be used in sections : defaults | frontend | listen | backend
2550 yes | no | yes | yes
2551 Arguments :
2552 <timeout> is the timeout value is specified in milliseconds by default, but
2553 can be in any other unit if the number is suffixed by the unit,
2554 as explained at the top of this document.
2555
2556 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002557 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01002558 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01002559 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
2560 connect timeout also presets the queue timeout to the same value if this one
2561 has not been specified. Historically, the contimeout was also used to set the
2562 tarpit timeout in a listen section, which is not possible in a pure frontend.
2563
2564 This parameter is specific to backends, but can be specified once for all in
2565 "defaults" sections. This is in fact one of the easiest solutions not to
2566 forget about it. An unspecified timeout results in an infinite timeout, which
2567 is not recommended. Such a usage is accepted and works but reports a warning
2568 during startup because it may results in accumulation of failed sessions in
2569 the system if the system's timeouts are not configured either.
2570
2571 This parameter is provided for backwards compatibility but is currently
2572 deprecated. Please use "timeout connect", "timeout queue" or "timeout tarpit"
2573 instead.
2574
2575 See also : "timeout connect", "timeout queue", "timeout tarpit",
2576 "timeout server", "contimeout".
2577
2578
Willy Tarreau55165fe2009-05-10 12:02:55 +02002579cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02002580 [ postonly ] [ preserve ] [ httponly ] [ secure ]
2581 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002582 Enable cookie-based persistence in a backend.
2583 May be used in sections : defaults | frontend | listen | backend
2584 yes | no | yes | yes
2585 Arguments :
2586 <name> is the name of the cookie which will be monitored, modified or
2587 inserted in order to bring persistence. This cookie is sent to
2588 the client via a "Set-Cookie" header in the response, and is
2589 brought back by the client in a "Cookie" header in all requests.
2590 Special care should be taken to choose a name which does not
2591 conflict with any likely application cookie. Also, if the same
2592 backends are subject to be used by the same clients (eg:
2593 HTTP/HTTPS), care should be taken to use different cookie names
2594 between all backends if persistence between them is not desired.
2595
2596 rewrite This keyword indicates that the cookie will be provided by the
2597 server and that haproxy will have to modify its value to set the
2598 server's identifier in it. This mode is handy when the management
2599 of complex combinations of "Set-cookie" and "Cache-control"
2600 headers is left to the application. The application can then
2601 decide whether or not it is appropriate to emit a persistence
2602 cookie. Since all responses should be monitored, this mode only
2603 works in HTTP close mode. Unless the application behaviour is
2604 very complex and/or broken, it is advised not to start with this
2605 mode for new deployments. This keyword is incompatible with
2606 "insert" and "prefix".
2607
2608 insert This keyword indicates that the persistence cookie will have to
Willy Tarreaua79094d2010-08-31 22:54:15 +02002609 be inserted by haproxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002610
Willy Tarreaua79094d2010-08-31 22:54:15 +02002611 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002612 server. When used without the "preserve" option, if the server
2613 emits a cookie with the same name, it will be remove before
2614 processing. For this reason, this mode can be used to upgrade
2615 existing configurations running in the "rewrite" mode. The cookie
2616 will only be a session cookie and will not be stored on the
2617 client's disk. By default, unless the "indirect" option is added,
2618 the server will see the cookies emitted by the client. Due to
2619 caching effects, it is generally wise to add the "nocache" or
2620 "postonly" keywords (see below). The "insert" keyword is not
2621 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002622
2623 prefix This keyword indicates that instead of relying on a dedicated
2624 cookie for the persistence, an existing one will be completed.
2625 This may be needed in some specific environments where the client
2626 does not support more than one single cookie and the application
2627 already needs it. In this case, whenever the server sets a cookie
2628 named <name>, it will be prefixed with the server's identifier
2629 and a delimiter. The prefix will be removed from all client
2630 requests so that the server still finds the cookie it emitted.
2631 Since all requests and responses are subject to being modified,
2632 this mode requires the HTTP close mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02002633 not compatible with "rewrite" and "insert". Note: it is highly
2634 recommended not to use "indirect" with "prefix", otherwise server
2635 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002636
Willy Tarreaua79094d2010-08-31 22:54:15 +02002637 indirect When this option is specified, no cookie will be emitted to a
2638 client which already has a valid one for the server which has
2639 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002640 it will be removed, unless the "preserve" option is also set. In
2641 "insert" mode, this will additionally remove cookies from the
2642 requests transmitted to the server, making the persistence
2643 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02002644 Note: it is highly recommended not to use "indirect" with
2645 "prefix", otherwise server cookie updates would not be sent to
2646 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002647
2648 nocache This option is recommended in conjunction with the insert mode
2649 when there is a cache between the client and HAProxy, as it
2650 ensures that a cacheable response will be tagged non-cacheable if
2651 a cookie needs to be inserted. This is important because if all
2652 persistence cookies are added on a cacheable home page for
2653 instance, then all customers will then fetch the page from an
2654 outer cache and will all share the same persistence cookie,
2655 leading to one server receiving much more traffic than others.
2656 See also the "insert" and "postonly" options.
2657
2658 postonly This option ensures that cookie insertion will only be performed
2659 on responses to POST requests. It is an alternative to the
2660 "nocache" option, because POST responses are not cacheable, so
2661 this ensures that the persistence cookie will never get cached.
2662 Since most sites do not need any sort of persistence before the
2663 first POST which generally is a login request, this is a very
2664 efficient method to optimize caching without risking to find a
2665 persistence cookie in the cache.
2666 See also the "insert" and "nocache" options.
2667
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002668 preserve This option may only be used with "insert" and/or "indirect". It
2669 allows the server to emit the persistence cookie itself. In this
2670 case, if a cookie is found in the response, haproxy will leave it
2671 untouched. This is useful in order to end persistence after a
2672 logout request for instance. For this, the server just has to
2673 emit a cookie with an invalid value (eg: empty) or with a date in
2674 the past. By combining this mechanism with the "disable-on-404"
2675 check option, it is possible to perform a completely graceful
2676 shutdown because users will definitely leave the server after
2677 they logout.
2678
Willy Tarreau4992dd22012-05-31 21:02:17 +02002679 httponly This option tells haproxy to add an "HttpOnly" cookie attribute
2680 when a cookie is inserted. This attribute is used so that a
2681 user agent doesn't share the cookie with non-HTTP components.
2682 Please check RFC6265 for more information on this attribute.
2683
2684 secure This option tells haproxy to add a "Secure" cookie attribute when
2685 a cookie is inserted. This attribute is used so that a user agent
2686 never emits this cookie over non-secure channels, which means
2687 that a cookie learned with this flag will be presented only over
2688 SSL/TLS connections. Please check RFC6265 for more information on
2689 this attribute.
2690
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002691 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002692 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01002693 name. If the domain begins with a dot, the browser is allowed to
2694 use it for any host ending with that name. It is also possible to
2695 specify several domain names by invoking this option multiple
2696 times. Some browsers might have small limits on the number of
2697 domains, so be careful when doing that. For the record, sending
2698 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002699
Willy Tarreau996a92c2010-10-13 19:30:47 +02002700 maxidle This option allows inserted cookies to be ignored after some idle
2701 time. It only works with insert-mode cookies. When a cookie is
2702 sent to the client, the date this cookie was emitted is sent too.
2703 Upon further presentations of this cookie, if the date is older
2704 than the delay indicated by the parameter (in seconds), it will
2705 be ignored. Otherwise, it will be refreshed if needed when the
2706 response is sent to the client. This is particularly useful to
2707 prevent users who never close their browsers from remaining for
2708 too long on the same server (eg: after a farm size change). When
2709 this option is set and a cookie has no date, it is always
2710 accepted, but gets refreshed in the response. This maintains the
2711 ability for admins to access their sites. Cookies that have a
2712 date in the future further than 24 hours are ignored. Doing so
2713 lets admins fix timezone issues without risking kicking users off
2714 the site.
2715
2716 maxlife This option allows inserted cookies to be ignored after some life
2717 time, whether they're in use or not. It only works with insert
2718 mode cookies. When a cookie is first sent to the client, the date
2719 this cookie was emitted is sent too. Upon further presentations
2720 of this cookie, if the date is older than the delay indicated by
2721 the parameter (in seconds), it will be ignored. If the cookie in
2722 the request has no date, it is accepted and a date will be set.
2723 Cookies that have a date in the future further than 24 hours are
2724 ignored. Doing so lets admins fix timezone issues without risking
2725 kicking users off the site. Contrary to maxidle, this value is
2726 not refreshed, only the first visit date counts. Both maxidle and
2727 maxlife may be used at the time. This is particularly useful to
2728 prevent users who never close their browsers from remaining for
2729 too long on the same server (eg: after a farm size change). This
2730 is stronger than the maxidle method in that it forces a
2731 redispatch after some absolute delay.
2732
Willy Tarreau0ba27502007-12-24 16:55:16 +01002733 There can be only one persistence cookie per HTTP backend, and it can be
2734 declared in a defaults section. The value of the cookie will be the value
2735 indicated after the "cookie" keyword in a "server" statement. If no cookie
2736 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002737
Willy Tarreau0ba27502007-12-24 16:55:16 +01002738 Examples :
2739 cookie JSESSIONID prefix
2740 cookie SRV insert indirect nocache
2741 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02002742 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01002743
Willy Tarreau294d0f02015-08-10 19:40:12 +02002744 See also : "balance source", "capture cookie", "server" and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002745
Willy Tarreau983e01e2010-01-11 18:42:06 +01002746
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02002747declare capture [ request | response ] len <length>
2748 Declares a capture slot.
2749 May be used in sections : defaults | frontend | listen | backend
2750 no | yes | yes | no
2751 Arguments:
2752 <length> is the length allowed for the capture.
2753
2754 This declaration is only available in the frontend or listen section, but the
2755 reserved slot can be used in the backends. The "request" keyword allocates a
2756 capture slot for use in the request, and "response" allocates a capture slot
2757 for use in the response.
2758
2759 See also: "capture-req", "capture-res" (sample converters),
2760 "http-request capture" and "http-response capture".
2761
2762
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002763default-server [param*]
2764 Change default options for a server in a backend
2765 May be used in sections : defaults | frontend | listen | backend
2766 yes | no | yes | yes
2767 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01002768 <param*> is a list of parameters for this server. The "default-server"
2769 keyword accepts an important number of options and has a complete
2770 section dedicated to it. Please refer to section 5 for more
2771 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002772
Willy Tarreau983e01e2010-01-11 18:42:06 +01002773 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002774 default-server inter 1000 weight 13
2775
2776 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01002777
Willy Tarreau983e01e2010-01-11 18:42:06 +01002778
Willy Tarreau0ba27502007-12-24 16:55:16 +01002779default_backend <backend>
2780 Specify the backend to use when no "use_backend" rule has been matched.
2781 May be used in sections : defaults | frontend | listen | backend
2782 yes | yes | yes | no
2783 Arguments :
2784 <backend> is the name of the backend to use.
2785
2786 When doing content-switching between frontend and backends using the
2787 "use_backend" keyword, it is often useful to indicate which backend will be
2788 used when no rule has matched. It generally is the dynamic backend which
2789 will catch all undetermined requests.
2790
Willy Tarreau0ba27502007-12-24 16:55:16 +01002791 Example :
2792
2793 use_backend dynamic if url_dyn
2794 use_backend static if url_css url_img extension_img
2795 default_backend dynamic
2796
Willy Tarreau98d04852015-05-26 12:18:29 +02002797 See also : "use_backend"
Willy Tarreau2769aa02007-12-27 18:26:09 +01002798
Willy Tarreau0ba27502007-12-24 16:55:16 +01002799
Baptiste Assmann27f51342013-10-09 06:51:49 +02002800description <string>
2801 Describe a listen, frontend or backend.
2802 May be used in sections : defaults | frontend | listen | backend
2803 no | yes | yes | yes
2804 Arguments : string
2805
2806 Allows to add a sentence to describe the related object in the HAProxy HTML
2807 stats page. The description will be printed on the right of the object name
2808 it describes.
2809 No need to backslash spaces in the <string> arguments.
2810
2811
Willy Tarreau0ba27502007-12-24 16:55:16 +01002812disabled
2813 Disable a proxy, frontend or backend.
2814 May be used in sections : defaults | frontend | listen | backend
2815 yes | yes | yes | yes
2816 Arguments : none
2817
2818 The "disabled" keyword is used to disable an instance, mainly in order to
2819 liberate a listening port or to temporarily disable a service. The instance
2820 will still be created and its configuration will be checked, but it will be
2821 created in the "stopped" state and will appear as such in the statistics. It
2822 will not receive any traffic nor will it send any health-checks or logs. It
2823 is possible to disable many instances at once by adding the "disabled"
2824 keyword in a "defaults" section.
2825
2826 See also : "enabled"
2827
2828
Willy Tarreau5ce94572010-06-07 14:35:41 +02002829dispatch <address>:<port>
2830 Set a default server address
2831 May be used in sections : defaults | frontend | listen | backend
2832 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02002833 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02002834
2835 <address> is the IPv4 address of the default server. Alternatively, a
2836 resolvable hostname is supported, but this name will be resolved
2837 during start-up.
2838
2839 <ports> is a mandatory port specification. All connections will be sent
2840 to this port, and it is not permitted to use port offsets as is
2841 possible with normal servers.
2842
Willy Tarreau787aed52011-04-15 06:45:37 +02002843 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02002844 server can take the connection. In the past it was used to forward non
2845 persistent connections to an auxiliary load balancer. Due to its simple
2846 syntax, it has also been used for simple TCP relays. It is recommended not to
2847 use it for more clarity, and to use the "server" directive instead.
2848
2849 See also : "server"
2850
2851
Willy Tarreau0ba27502007-12-24 16:55:16 +01002852enabled
2853 Enable a proxy, frontend or backend.
2854 May be used in sections : defaults | frontend | listen | backend
2855 yes | yes | yes | yes
2856 Arguments : none
2857
2858 The "enabled" keyword is used to explicitly enable an instance, when the
2859 defaults has been set to "disabled". This is very rarely used.
2860
2861 See also : "disabled"
2862
2863
2864errorfile <code> <file>
2865 Return a file contents instead of errors generated by HAProxy
2866 May be used in sections : defaults | frontend | listen | backend
2867 yes | yes | yes | yes
2868 Arguments :
2869 <code> is the HTTP status code. Currently, HAProxy is capable of
CJ Ess108b1dd2015-04-07 12:03:37 -04002870 generating codes 200, 400, 403, 405, 408, 429, 500, 502, 503, and
2871 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002872
2873 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002874 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01002875 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01002876 error pages, and to use absolute paths, since files are read
2877 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002878
2879 It is important to understand that this keyword is not meant to rewrite
2880 errors returned by the server, but errors detected and returned by HAProxy.
2881 This is why the list of supported errors is limited to a small set.
2882
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002883 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2884
Willy Tarreau0ba27502007-12-24 16:55:16 +01002885 The files are returned verbatim on the TCP socket. This allows any trick such
2886 as redirections to another URL or site, as well as tricks to clean cookies,
2887 force enable or disable caching, etc... The package provides default error
2888 files returning the same contents as default errors.
2889
Willy Tarreau59140a22009-02-22 12:02:30 +01002890 The files should not exceed the configured buffer size (BUFSIZE), which
2891 generally is 8 or 16 kB, otherwise they will be truncated. It is also wise
2892 not to put any reference to local contents (eg: images) in order to avoid
2893 loops between the client and HAProxy when all servers are down, causing an
2894 error to be returned instead of an image. For better HTTP compliance, it is
2895 recommended that all header lines end with CR-LF and not LF alone.
2896
Willy Tarreau0ba27502007-12-24 16:55:16 +01002897 The files are read at the same time as the configuration and kept in memory.
2898 For this reason, the errors continue to be returned even when the process is
2899 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01002900 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01002901 403 status code and interrogating a blocked URL.
2902
2903 See also : "errorloc", "errorloc302", "errorloc303"
2904
Willy Tarreau59140a22009-02-22 12:02:30 +01002905 Example :
2906 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
Willy Tarreau2705a612014-05-23 17:38:34 +02002907 errorfile 408 /dev/null # workaround Chrome pre-connect bug
Willy Tarreau59140a22009-02-22 12:02:30 +01002908 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
2909 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
2910
Willy Tarreau2769aa02007-12-27 18:26:09 +01002911
2912errorloc <code> <url>
2913errorloc302 <code> <url>
2914 Return an HTTP redirection to a URL instead of errors generated by HAProxy
2915 May be used in sections : defaults | frontend | listen | backend
2916 yes | yes | yes | yes
2917 Arguments :
2918 <code> is the HTTP status code. Currently, HAProxy is capable of
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002919 generating codes 200, 400, 403, 408, 500, 502, 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002920
2921 <url> it is the exact contents of the "Location" header. It may contain
2922 either a relative URI to an error page hosted on the same site,
2923 or an absolute URI designating an error page on another site.
2924 Special care should be given to relative URIs to avoid redirect
2925 loops if the URI itself may generate the same error (eg: 500).
2926
2927 It is important to understand that this keyword is not meant to rewrite
2928 errors returned by the server, but errors detected and returned by HAProxy.
2929 This is why the list of supported errors is limited to a small set.
2930
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002931 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2932
Willy Tarreau2769aa02007-12-27 18:26:09 +01002933 Note that both keyword return the HTTP 302 status code, which tells the
2934 client to fetch the designated URL using the same HTTP method. This can be
2935 quite problematic in case of non-GET methods such as POST, because the URL
2936 sent to the client might not be allowed for something other than GET. To
2937 workaround this problem, please use "errorloc303" which send the HTTP 303
2938 status code, indicating to the client that the URL must be fetched with a GET
2939 request.
2940
2941 See also : "errorfile", "errorloc303"
2942
2943
2944errorloc303 <code> <url>
2945 Return an HTTP redirection to a URL instead of errors generated by HAProxy
2946 May be used in sections : defaults | frontend | listen | backend
2947 yes | yes | yes | yes
2948 Arguments :
2949 <code> is the HTTP status code. Currently, HAProxy is capable of
2950 generating codes 400, 403, 408, 500, 502, 503, and 504.
2951
2952 <url> it is the exact contents of the "Location" header. It may contain
2953 either a relative URI to an error page hosted on the same site,
2954 or an absolute URI designating an error page on another site.
2955 Special care should be given to relative URIs to avoid redirect
2956 loops if the URI itself may generate the same error (eg: 500).
2957
2958 It is important to understand that this keyword is not meant to rewrite
2959 errors returned by the server, but errors detected and returned by HAProxy.
2960 This is why the list of supported errors is limited to a small set.
2961
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002962 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2963
Willy Tarreau2769aa02007-12-27 18:26:09 +01002964 Note that both keyword return the HTTP 303 status code, which tells the
2965 client to fetch the designated URL using the same HTTP GET method. This
2966 solves the usual problems associated with "errorloc" and the 302 code. It is
2967 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002968 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002969
2970 See also : "errorfile", "errorloc", "errorloc302"
2971
2972
Simon Horman51a1cf62015-02-03 13:00:44 +09002973email-alert from <emailaddr>
2974 Declare the from email address to be used in both the envelope and header
2975 of email alerts. This is the address that email alerts are sent from.
2976 May be used in sections: defaults | frontend | listen | backend
2977 yes | yes | yes | yes
2978
2979 Arguments :
2980
2981 <emailaddr> is the from email address to use when sending email alerts
2982
2983 Also requires "email-alert mailers" and "email-alert to" to be set
2984 and if so sending email alerts is enabled for the proxy.
2985
Simon Horman64e34162015-02-06 11:11:57 +09002986 See also : "email-alert level", "email-alert mailers",
Cyril Bonté307ee1e2015-09-28 23:16:06 +02002987 "email-alert myhostname", "email-alert to", section 3.6 about
2988 mailers.
Simon Horman64e34162015-02-06 11:11:57 +09002989
2990
2991email-alert level <level>
2992 Declare the maximum log level of messages for which email alerts will be
2993 sent. This acts as a filter on the sending of email alerts.
2994 May be used in sections: defaults | frontend | listen | backend
2995 yes | yes | yes | yes
2996
2997 Arguments :
2998
2999 <level> One of the 8 syslog levels:
3000 emerg alert crit err warning notice info debug
3001 The above syslog levels are ordered from lowest to highest.
3002
3003 By default level is alert
3004
3005 Also requires "email-alert from", "email-alert mailers" and
3006 "email-alert to" to be set and if so sending email alerts is enabled
3007 for the proxy.
3008
Simon Horman1421e212015-04-30 13:10:35 +09003009 Alerts are sent when :
3010
3011 * An un-paused server is marked as down and <level> is alert or lower
3012 * A paused server is marked as down and <level> is notice or lower
3013 * A server is marked as up or enters the drain state and <level>
3014 is notice or lower
3015 * "option log-health-checks" is enabled, <level> is info or lower,
3016 and a health check status update occurs
3017
Simon Horman64e34162015-02-06 11:11:57 +09003018 See also : "email-alert from", "email-alert mailers",
3019 "email-alert myhostname", "email-alert to",
Simon Horman51a1cf62015-02-03 13:00:44 +09003020 section 3.6 about mailers.
3021
3022
3023email-alert mailers <mailersect>
3024 Declare the mailers to be used when sending email alerts
3025 May be used in sections: defaults | frontend | listen | backend
3026 yes | yes | yes | yes
3027
3028 Arguments :
3029
3030 <mailersect> is the name of the mailers section to send email alerts.
3031
3032 Also requires "email-alert from" and "email-alert to" to be set
3033 and if so sending email alerts is enabled for the proxy.
3034
Simon Horman64e34162015-02-06 11:11:57 +09003035 See also : "email-alert from", "email-alert level", "email-alert myhostname",
3036 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003037
3038
3039email-alert myhostname <hostname>
3040 Declare the to hostname address to be used when communicating with
3041 mailers.
3042 May be used in sections: defaults | frontend | listen | backend
3043 yes | yes | yes | yes
3044
3045 Arguments :
3046
3047 <emailaddr> is the to email address to use when sending email alerts
3048
3049 By default the systems hostname is used.
3050
3051 Also requires "email-alert from", "email-alert mailers" and
3052 "email-alert to" to be set and if so sending email alerts is enabled
3053 for the proxy.
3054
Simon Horman64e34162015-02-06 11:11:57 +09003055 See also : "email-alert from", "email-alert level", "email-alert mailers",
3056 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003057
3058
3059email-alert to <emailaddr>
3060 Declare both the recipent address in the envelope and to address in the
3061 header of email alerts. This is the address that email alerts are sent to.
3062 May be used in sections: defaults | frontend | listen | backend
3063 yes | yes | yes | yes
3064
3065 Arguments :
3066
3067 <emailaddr> is the to email address to use when sending email alerts
3068
3069 Also requires "email-alert mailers" and "email-alert to" to be set
3070 and if so sending email alerts is enabled for the proxy.
3071
Simon Horman64e34162015-02-06 11:11:57 +09003072 See also : "email-alert from", "email-alert level", "email-alert mailers",
Simon Horman51a1cf62015-02-03 13:00:44 +09003073 "email-alert myhostname", section 3.6 about mailers.
3074
3075
Willy Tarreau4de91492010-01-22 19:10:05 +01003076force-persist { if | unless } <condition>
3077 Declare a condition to force persistence on down servers
3078 May be used in sections: defaults | frontend | listen | backend
3079 no | yes | yes | yes
3080
3081 By default, requests are not dispatched to down servers. It is possible to
3082 force this using "option persist", but it is unconditional and redispatches
3083 to a valid server if "option redispatch" is set. That leaves with very little
3084 possibilities to force some requests to reach a server which is artificially
3085 marked down for maintenance operations.
3086
3087 The "force-persist" statement allows one to declare various ACL-based
3088 conditions which, when met, will cause a request to ignore the down status of
3089 a server and still try to connect to it. That makes it possible to start a
3090 server, still replying an error to the health checks, and run a specially
3091 configured browser to test the service. Among the handy methods, one could
3092 use a specific source IP address, or a specific cookie. The cookie also has
3093 the advantage that it can easily be added/removed on the browser from a test
3094 page. Once the service is validated, it is then possible to open the service
3095 to the world by returning a valid response to health checks.
3096
3097 The forced persistence is enabled when an "if" condition is met, or unless an
3098 "unless" condition is met. The final redispatch is always disabled when this
3099 is used.
3100
Cyril Bonté0d4bf012010-04-25 23:21:46 +02003101 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02003102 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01003103
3104
Willy Tarreau2769aa02007-12-27 18:26:09 +01003105fullconn <conns>
3106 Specify at what backend load the servers will reach their maxconn
3107 May be used in sections : defaults | frontend | listen | backend
3108 yes | no | yes | yes
3109 Arguments :
3110 <conns> is the number of connections on the backend which will make the
3111 servers use the maximal number of connections.
3112
Willy Tarreau198a7442008-01-17 12:05:32 +01003113 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01003114 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01003115 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01003116 load. The server will then always accept at least <minconn> connections,
3117 never more than <maxconn>, and the limit will be on the ramp between both
3118 values when the backend has less than <conns> concurrent connections. This
3119 makes it possible to limit the load on the servers during normal loads, but
3120 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003121 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003122
Willy Tarreaufbb78422011-06-05 15:38:35 +02003123 Since it's hard to get this value right, haproxy automatically sets it to
3124 10% of the sum of the maxconns of all frontends that may branch to this
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01003125 backend (based on "use_backend" and "default_backend" rules). That way it's
3126 safe to leave it unset. However, "use_backend" involving dynamic names are
3127 not counted since there is no way to know if they could match or not.
Willy Tarreaufbb78422011-06-05 15:38:35 +02003128
Willy Tarreau2769aa02007-12-27 18:26:09 +01003129 Example :
3130 # The servers will accept between 100 and 1000 concurrent connections each
3131 # and the maximum of 1000 will be reached when the backend reaches 10000
3132 # connections.
3133 backend dynamic
3134 fullconn 10000
3135 server srv1 dyn1:80 minconn 100 maxconn 1000
3136 server srv2 dyn2:80 minconn 100 maxconn 1000
3137
3138 See also : "maxconn", "server"
3139
3140
3141grace <time>
3142 Maintain a proxy operational for some time after a soft stop
3143 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01003144 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01003145 Arguments :
3146 <time> is the time (by default in milliseconds) for which the instance
3147 will remain operational with the frontend sockets still listening
3148 when a soft-stop is received via the SIGUSR1 signal.
3149
3150 This may be used to ensure that the services disappear in a certain order.
3151 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003152 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01003153 needed by the equipment to detect the failure.
3154
3155 Note that currently, there is very little benefit in using this parameter,
3156 and it may in fact complicate the soft-reconfiguration process more than
3157 simplify it.
3158
Willy Tarreau0ba27502007-12-24 16:55:16 +01003159
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003160hash-type <method> <function> <modifier>
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003161 Specify a method to use for mapping hashes to servers
3162 May be used in sections : defaults | frontend | listen | backend
3163 yes | no | yes | yes
3164 Arguments :
Bhaskar98634f02013-10-29 23:30:51 -04003165 <method> is the method used to select a server from the hash computed by
3166 the <function> :
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003167
Bhaskar98634f02013-10-29 23:30:51 -04003168 map-based the hash table is a static array containing all alive servers.
3169 The hashes will be very smooth, will consider weights, but
3170 will be static in that weight changes while a server is up
3171 will be ignored. This means that there will be no slow start.
3172 Also, since a server is selected by its position in the array,
3173 most mappings are changed when the server count changes. This
3174 means that when a server goes up or down, or when a server is
3175 added to a farm, most connections will be redistributed to
3176 different servers. This can be inconvenient with caches for
3177 instance.
Willy Tarreau798a39c2010-11-24 15:04:29 +01003178
Bhaskar98634f02013-10-29 23:30:51 -04003179 consistent the hash table is a tree filled with many occurrences of each
3180 server. The hash key is looked up in the tree and the closest
3181 server is chosen. This hash is dynamic, it supports changing
3182 weights while the servers are up, so it is compatible with the
3183 slow start feature. It has the advantage that when a server
3184 goes up or down, only its associations are moved. When a
3185 server is added to the farm, only a few part of the mappings
3186 are redistributed, making it an ideal method for caches.
3187 However, due to its principle, the distribution will never be
3188 very smooth and it may sometimes be necessary to adjust a
3189 server's weight or its ID to get a more balanced distribution.
3190 In order to get the same distribution on multiple load
3191 balancers, it is important that all servers have the exact
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003192 same IDs. Note: consistent hash uses sdbm and avalanche if no
3193 hash function is specified.
Bhaskar98634f02013-10-29 23:30:51 -04003194
3195 <function> is the hash function to be used :
3196
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003197 sdbm this function was created initially for sdbm (a public-domain
Bhaskar98634f02013-10-29 23:30:51 -04003198 reimplementation of ndbm) database library. It was found to do
3199 well in scrambling bits, causing better distribution of the keys
3200 and fewer splits. It also happens to be a good general hashing
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003201 function with good distribution, unless the total server weight
3202 is a multiple of 64, in which case applying the avalanche
3203 modifier may help.
Bhaskar98634f02013-10-29 23:30:51 -04003204
3205 djb2 this function was first proposed by Dan Bernstein many years ago
3206 on comp.lang.c. Studies have shown that for certain workload this
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003207 function provides a better distribution than sdbm. It generally
3208 works well with text-based inputs though it can perform extremely
3209 poorly with numeric-only input or when the total server weight is
3210 a multiple of 33, unless the avalanche modifier is also used.
3211
Willy Tarreaua0f42712013-11-14 14:30:35 +01003212 wt6 this function was designed for haproxy while testing other
3213 functions in the past. It is not as smooth as the other ones, but
3214 is much less sensible to the input data set or to the number of
3215 servers. It can make sense as an alternative to sdbm+avalanche or
3216 djb2+avalanche for consistent hashing or when hashing on numeric
3217 data such as a source IP address or a visitor identifier in a URL
3218 parameter.
3219
Willy Tarreau324f07f2015-01-20 19:44:50 +01003220 crc32 this is the most common CRC32 implementation as used in Ethernet,
3221 gzip, PNG, etc. It is slower than the other ones but may provide
3222 a better distribution or less predictable results especially when
3223 used on strings.
3224
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003225 <modifier> indicates an optional method applied after hashing the key :
3226
3227 avalanche This directive indicates that the result from the hash
3228 function above should not be used in its raw form but that
3229 a 4-byte full avalanche hash must be applied first. The
3230 purpose of this step is to mix the resulting bits from the
3231 previous hash in order to avoid any undesired effect when
3232 the input contains some limited values or when the number of
3233 servers is a multiple of one of the hash's components (64
3234 for SDBM, 33 for DJB2). Enabling avalanche tends to make the
3235 result less predictable, but it's also not as smooth as when
3236 using the original function. Some testing might be needed
3237 with some workloads. This hash is one of the many proposed
3238 by Bob Jenkins.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003239
Bhaskar98634f02013-10-29 23:30:51 -04003240 The default hash type is "map-based" and is recommended for most usages. The
3241 default function is "sdbm", the selection of a function should be based on
3242 the range of the values being hashed.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003243
3244 See also : "balance", "server"
3245
3246
Willy Tarreau0ba27502007-12-24 16:55:16 +01003247http-check disable-on-404
3248 Enable a maintenance mode upon HTTP/404 response to health-checks
3249 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01003250 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01003251 Arguments : none
3252
3253 When this option is set, a server which returns an HTTP code 404 will be
3254 excluded from further load-balancing, but will still receive persistent
3255 connections. This provides a very convenient method for Web administrators
3256 to perform a graceful shutdown of their servers. It is also important to note
3257 that a server which is detected as failed while it was in this mode will not
3258 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
3259 will immediately be reinserted into the farm. The status on the stats page
3260 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01003261 option only works in conjunction with the "httpchk" option. If this option
3262 is used with "http-check expect", then it has precedence over it so that 404
3263 responses will still be considered as soft-stop.
3264
3265 See also : "option httpchk", "http-check expect"
3266
3267
3268http-check expect [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003269 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01003270 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02003271 yes | no | yes | yes
Willy Tarreaubd741542010-03-16 18:46:54 +01003272 Arguments :
3273 <match> is a keyword indicating how to look for a specific pattern in the
3274 response. The keyword may be one of "status", "rstatus",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003275 "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01003276 exclamation mark ("!") to negate the match. Spaces are allowed
3277 between the exclamation mark and the keyword. See below for more
3278 details on the supported keywords.
3279
3280 <pattern> is the pattern to look for. It may be a string or a regular
3281 expression. If the pattern contains spaces, they must be escaped
3282 with the usual backslash ('\').
3283
3284 By default, "option httpchk" considers that response statuses 2xx and 3xx
3285 are valid, and that others are invalid. When "http-check expect" is used,
3286 it defines what is considered valid or invalid. Only one "http-check"
3287 statement is supported in a backend. If a server fails to respond or times
3288 out, the check obviously fails. The available matches are :
3289
3290 status <string> : test the exact string match for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003291 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003292 response's status code is exactly this string. If the
3293 "status" keyword is prefixed with "!", then the response
3294 will be considered invalid if the status code matches.
3295
3296 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003297 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003298 response's status code matches the expression. If the
3299 "rstatus" keyword is prefixed with "!", then the response
3300 will be considered invalid if the status code matches.
3301 This is mostly used to check for multiple codes.
3302
3303 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003304 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003305 response's body contains this exact string. If the
3306 "string" keyword is prefixed with "!", then the response
3307 will be considered invalid if the body contains this
3308 string. This can be used to look for a mandatory word at
3309 the end of a dynamic page, or to detect a failure when a
3310 specific error appears on the check page (eg: a stack
3311 trace).
3312
3313 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003314 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003315 response's body matches this expression. If the "rstring"
3316 keyword is prefixed with "!", then the response will be
3317 considered invalid if the body matches the expression.
3318 This can be used to look for a mandatory word at the end
3319 of a dynamic page, or to detect a failure when a specific
3320 error appears on the check page (eg: a stack trace).
3321
3322 It is important to note that the responses will be limited to a certain size
3323 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
3324 Thus, too large responses may not contain the mandatory pattern when using
3325 "string" or "rstring". If a large response is absolutely required, it is
3326 possible to change the default max size by setting the global variable.
3327 However, it is worth keeping in mind that parsing very large responses can
3328 waste some CPU cycles, especially when regular expressions are used, and that
3329 it is always better to focus the checks on smaller resources.
3330
Cyril Bonté32602d22015-01-30 00:07:07 +01003331 Also "http-check expect" doesn't support HTTP keep-alive. Keep in mind that it
3332 will automatically append a "Connection: close" header, meaning that this
3333 header should not be present in the request provided by "option httpchk".
3334
Willy Tarreaubd741542010-03-16 18:46:54 +01003335 Last, if "http-check expect" is combined with "http-check disable-on-404",
3336 then this last one has precedence when the server responds with 404.
3337
3338 Examples :
3339 # only accept status 200 as valid
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003340 http-check expect status 200
Willy Tarreaubd741542010-03-16 18:46:54 +01003341
3342 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003343 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01003344
3345 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003346 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01003347
3348 # check that we have a correct hexadecimal tag before /html
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003349 http-check expect rstring <!--tag:[0-9a-f]*</html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01003350
Willy Tarreaubd741542010-03-16 18:46:54 +01003351 See also : "option httpchk", "http-check disable-on-404"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003352
3353
Willy Tarreauef781042010-01-27 11:53:01 +01003354http-check send-state
3355 Enable emission of a state header with HTTP health checks
3356 May be used in sections : defaults | frontend | listen | backend
3357 yes | no | yes | yes
3358 Arguments : none
3359
3360 When this option is set, haproxy will systematically send a special header
3361 "X-Haproxy-Server-State" with a list of parameters indicating to each server
3362 how they are seen by haproxy. This can be used for instance when a server is
3363 manipulated without access to haproxy and the operator needs to know whether
3364 haproxy still sees it up or not, or if the server is the last one in a farm.
3365
3366 The header is composed of fields delimited by semi-colons, the first of which
3367 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
3368 checks on the total number before transition, just as appears in the stats
3369 interface. Next headers are in the form "<variable>=<value>", indicating in
3370 no specific order some values available in the stats interface :
Joseph Lynch514061c2015-01-15 17:52:59 -08003371 - a variable "address", containing the address of the backend server.
3372 This corresponds to the <address> field in the server declaration. For
3373 unix domain sockets, it will read "unix".
3374
3375 - a variable "port", containing the port of the backend server. This
3376 corresponds to the <port> field in the server declaration. For unix
3377 domain sockets, it will read "unix".
3378
Willy Tarreauef781042010-01-27 11:53:01 +01003379 - a variable "name", containing the name of the backend followed by a slash
3380 ("/") then the name of the server. This can be used when a server is
3381 checked in multiple backends.
3382
3383 - a variable "node" containing the name of the haproxy node, as set in the
3384 global "node" variable, otherwise the system's hostname if unspecified.
3385
3386 - a variable "weight" indicating the weight of the server, a slash ("/")
3387 and the total weight of the farm (just counting usable servers). This
3388 helps to know if other servers are available to handle the load when this
3389 one fails.
3390
3391 - a variable "scur" indicating the current number of concurrent connections
3392 on the server, followed by a slash ("/") then the total number of
3393 connections on all servers of the same backend.
3394
3395 - a variable "qcur" indicating the current number of requests in the
3396 server's queue.
3397
3398 Example of a header received by the application server :
3399 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
3400 scur=13/22; qcur=0
3401
3402 See also : "option httpchk", "http-check disable-on-404"
3403
Willy Tarreauccbcc372012-12-27 12:37:57 +01003404http-request { allow | deny | tarpit | auth [realm <realm>] | redirect <rule> |
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003405 add-header <name> <fmt> | set-header <name> <fmt> |
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003406 capture <sample> [ len <length> | id <id> ] |
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003407 del-header <name> | set-nice <nice> | set-log-level <level> |
Sasha Pachev218f0642014-06-16 12:05:59 -06003408 replace-header <name> <match-regex> <replace-fmt> |
3409 replace-value <name> <match-regex> <replace-fmt> |
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01003410 set-method <fmt> | set-path <fmt> | set-query <fmt> |
3411 set-uri <fmt> | set-tos <tos> | set-mark <mark> |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003412 add-acl(<file name>) <key fmt> |
3413 del-acl(<file name>) <key fmt> |
3414 del-map(<file name>) <key fmt> |
Baptiste Assmannbb7e86a2014-09-03 18:29:47 +02003415 set-map(<file name>) <key fmt> <value fmt> |
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003416 set-var(<var name>) <expr> |
Thierry FOURNIER90da1912015-03-05 11:17:06 +01003417 { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] |
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02003418 sc-inc-gpc0(<sc-id>) |
Thierry FOURNIER236657b2015-08-19 08:25:14 +02003419 sc-set-gpt0(<sc-id>) <int> |
Willy Tarreau2d392c22015-08-24 01:43:45 +02003420 silent-drop |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003421 }
Cyril Bontéf0c60612010-02-06 14:44:47 +01003422 [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003423 Access control for Layer 7 requests
3424
3425 May be used in sections: defaults | frontend | listen | backend
3426 no | yes | yes | yes
3427
Willy Tarreau20b0de52012-12-24 15:45:22 +01003428 The http-request statement defines a set of rules which apply to layer 7
3429 processing. The rules are evaluated in their declaration order when they are
3430 met in a frontend, listen or backend section. Any rule may optionally be
3431 followed by an ACL-based condition, in which case it will only be evaluated
3432 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003433
Willy Tarreau20b0de52012-12-24 15:45:22 +01003434 The first keyword is the rule's action. Currently supported actions include :
3435 - "allow" : this stops the evaluation of the rules and lets the request
3436 pass the check. No further "http-request" rules are evaluated.
3437
3438 - "deny" : this stops the evaluation of the rules and immediately rejects
3439 the request and emits an HTTP 403 error. No further "http-request" rules
3440 are evaluated.
3441
Willy Tarreauccbcc372012-12-27 12:37:57 +01003442 - "tarpit" : this stops the evaluation of the rules and immediately blocks
3443 the request without responding for a delay specified by "timeout tarpit"
3444 or "timeout connect" if the former is not set. After that delay, if the
3445 client is still connected, an HTTP error 500 is returned so that the
3446 client does not suspect it has been tarpitted. Logs will report the flags
3447 "PT". The goal of the tarpit rule is to slow down robots during an attack
3448 when they're limited on the number of concurrent requests. It can be very
3449 efficient against very dumb robots, and will significantly reduce the
3450 load on firewalls compared to a "deny" rule. But when facing "correctly"
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003451 developed robots, it can make things worse by forcing haproxy and the
Willy Tarreau2d392c22015-08-24 01:43:45 +02003452 front firewall to support insane number of concurrent connections. See
3453 also the "silent-drop" action below.
Willy Tarreauccbcc372012-12-27 12:37:57 +01003454
Willy Tarreau20b0de52012-12-24 15:45:22 +01003455 - "auth" : this stops the evaluation of the rules and immediately responds
3456 with an HTTP 401 or 407 error code to invite the user to present a valid
3457 user name and password. No further "http-request" rules are evaluated. An
3458 optional "realm" parameter is supported, it sets the authentication realm
3459 that is returned with the response (typically the application's name).
3460
Willy Tarreau81499eb2012-12-27 12:19:02 +01003461 - "redirect" : this performs an HTTP redirection based on a redirect rule.
3462 This is exactly the same as the "redirect" statement except that it
3463 inserts a redirect rule which can be processed in the middle of other
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01003464 "http-request" rules and that these rules use the "log-format" strings.
3465 See the "redirect" keyword for the rule's syntax.
Willy Tarreau81499eb2012-12-27 12:19:02 +01003466
Willy Tarreau20b0de52012-12-24 15:45:22 +01003467 - "add-header" appends an HTTP header field whose name is specified in
3468 <name> and whose value is defined by <fmt> which follows the log-format
3469 rules (see Custom Log Format in section 8.2.4). This is particularly
3470 useful to pass connection-specific information to the server (eg: the
3471 client's SSL certificate), or to combine several headers into one. This
3472 rule is not final, so it is possible to add other similar rules. Note
3473 that header addition is performed immediately, so one rule might reuse
3474 the resulting header from a previous rule.
3475
3476 - "set-header" does the same as "add-header" except that the header name
3477 is first removed if it existed. This is useful when passing security
3478 information to the server, where the header must not be manipulated by
Willy Tarreau85603282015-01-21 20:39:27 +01003479 external users. Note that the new value is computed before the removal so
3480 it is possible to concatenate a value to an existing header.
Willy Tarreau20b0de52012-12-24 15:45:22 +01003481
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003482 - "del-header" removes all HTTP header fields whose name is specified in
3483 <name>.
3484
Sasha Pachev218f0642014-06-16 12:05:59 -06003485 - "replace-header" matches the regular expression in all occurrences of
3486 header field <name> according to <match-regex>, and replaces them with
3487 the <replace-fmt> argument. Format characters are allowed in replace-fmt
3488 and work like in <fmt> arguments in "add-header". The match is only
3489 case-sensitive. It is important to understand that this action only
3490 considers whole header lines, regardless of the number of values they
3491 may contain. This usage is suited to headers naturally containing commas
3492 in their value, such as If-Modified-Since and so on.
3493
3494 Example:
3495
3496 http-request replace-header Cookie foo=([^;]*);(.*) foo=\1;ip=%bi;\2
3497
3498 applied to:
3499
3500 Cookie: foo=foobar; expires=Tue, 14-Jun-2016 01:40:45 GMT;
3501
3502 outputs:
3503
3504 Cookie: foo=foobar;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT;
3505
3506 assuming the backend IP is 192.168.1.20
3507
3508 - "replace-value" works like "replace-header" except that it matches the
3509 regex against every comma-delimited value of the header field <name>
3510 instead of the entire header. This is suited for all headers which are
3511 allowed to carry more than one value. An example could be the Accept
3512 header.
3513
3514 Example:
3515
3516 http-request replace-value X-Forwarded-For ^192\.168\.(.*)$ 172.16.\1
3517
3518 applied to:
3519
3520 X-Forwarded-For: 192.168.10.1, 192.168.13.24, 10.0.0.37
3521
3522 outputs:
3523
3524 X-Forwarded-For: 172.16.10.1, 172.16.13.24, 10.0.0.37
3525
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01003526 - "set-method" rewrites the request method with the result of the
3527 evaluation of format string <fmt>. There should be very few valid reasons
3528 for having to do so as this is more likely to break something than to fix
3529 it.
3530
3531 - "set-path" rewrites the request path with the result of the evaluation of
3532 format string <fmt>. The query string, if any, is left intact. If a
3533 scheme and authority is found before the path, they are left intact as
3534 well. If the request doesn't have a path ("*"), this one is replaced with
3535 the format. This can be used to prepend a directory component in front of
3536 a path for example. See also "set-query" and "set-uri".
3537
3538 Example :
3539 # prepend the host name before the path
3540 http-request set-path /%[hdr(host)]%[path]
3541
3542 - "set-query" rewrites the request's query string which appears after the
3543 first question mark ("?") with the result of the evaluation of format
3544 string <fmt>. The part prior to the question mark is left intact. If the
3545 request doesn't contain a question mark and the new value is not empty,
3546 then one is added at the end of the URI, followed by the new value. If
3547 a question mark was present, it will never be removed even if the value
3548 is empty. This can be used to add or remove parameters from the query
3549 string. See also "set-query" and "set-uri".
3550
3551 Example :
3552 # replace "%3D" with "=" in the query string
3553 http-request set-query %[query,regsub(%3D,=,g)]
3554
3555 - "set-uri" rewrites the request URI with the result of the evaluation of
3556 format string <fmt>. The scheme, authority, path and query string are all
3557 replaced at once. This can be used to rewrite hosts in front of proxies,
3558 or to perform complex modifications to the URI such as moving parts
3559 between the path and the query string. See also "set-path" and
3560 "set-query".
3561
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003562 - "set-nice" sets the "nice" factor of the current request being processed.
3563 It only has effect against the other requests being processed at the same
3564 time. The default value is 0, unless altered by the "nice" setting on the
3565 "bind" line. The accepted range is -1024..1024. The higher the value, the
3566 nicest the request will be. Lower values will make the request more
3567 important than other ones. This can be useful to improve the speed of
3568 some requests, or lower the priority of non-important requests. Using
3569 this setting without prior experimentation can cause some major slowdown.
3570
Willy Tarreau9a355ec2013-06-11 17:45:46 +02003571 - "set-log-level" is used to change the log level of the current request
3572 when a certain condition is met. Valid levels are the 8 syslog levels
3573 (see the "log" keyword) plus the special level "silent" which disables
3574 logging for this request. This rule is not final so the last matching
3575 rule wins. This rule can be useful to disable health checks coming from
3576 another equipment.
3577
Willy Tarreau42cf39e2013-06-11 18:51:32 +02003578 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
3579 the client to the value passed in <tos> on platforms which support this.
3580 This value represents the whole 8 bits of the IP TOS field, and can be
3581 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
3582 that only the 6 higher bits are used in DSCP or TOS, and the two lower
3583 bits are always 0. This can be used to adjust some routing behaviour on
3584 border routers based on some information from the request. See RFC 2474,
3585 2597, 3260 and 4594 for more information.
3586
Willy Tarreau51347ed2013-06-11 19:34:13 +02003587 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
3588 client to the value passed in <mark> on platforms which support it. This
3589 value is an unsigned 32 bit value which can be matched by netfilter and
3590 by the routing table. It can be expressed both in decimal or hexadecimal
3591 format (prefixed by "0x"). This can be useful to force certain packets to
3592 take a different route (for example a cheaper network path for bulk
3593 downloads). This works on Linux kernels 2.6.32 and above and requires
3594 admin privileges.
3595
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003596 - "add-acl" is used to add a new entry into an ACL. The ACL must be loaded
3597 from a file (even a dummy empty file). The file name of the ACL to be
3598 updated is passed between parentheses. It takes one argument: <key fmt>,
3599 which follows log-format rules, to collect content of the new entry. It
3600 performs a lookup in the ACL before insertion, to avoid duplicated (or
3601 more) values. This lookup is done by a linear search and can be expensive
3602 with large lists! It is the equivalent of the "add acl" command from the
3603 stats socket, but can be triggered by an HTTP request.
3604
3605 - "del-acl" is used to delete an entry from an ACL. The ACL must be loaded
3606 from a file (even a dummy empty file). The file name of the ACL to be
3607 updated is passed between parentheses. It takes one argument: <key fmt>,
3608 which follows log-format rules, to collect content of the entry to delete.
3609 It is the equivalent of the "del acl" command from the stats socket, but
3610 can be triggered by an HTTP request.
3611
3612 - "del-map" is used to delete an entry from a MAP. The MAP must be loaded
3613 from a file (even a dummy empty file). The file name of the MAP to be
3614 updated is passed between parentheses. It takes one argument: <key fmt>,
3615 which follows log-format rules, to collect content of the entry to delete.
3616 It takes one argument: "file name" It is the equivalent of the "del map"
3617 command from the stats socket, but can be triggered by an HTTP request.
3618
3619 - "set-map" is used to add a new entry into a MAP. The MAP must be loaded
3620 from a file (even a dummy empty file). The file name of the MAP to be
3621 updated is passed between parentheses. It takes 2 arguments: <key fmt>,
3622 which follows log-format rules, used to collect MAP key, and <value fmt>,
3623 which follows log-format rules, used to collect content for the new entry.
3624 It performs a lookup in the MAP before insertion, to avoid duplicated (or
3625 more) values. This lookup is done by a linear search and can be expensive
3626 with large lists! It is the equivalent of the "set map" command from the
3627 stats socket, but can be triggered by an HTTP request.
3628
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003629 - capture <sample> [ len <length> | id <id> ] :
Willy Tarreaua9083d02015-05-08 15:27:59 +02003630 captures sample expression <sample> from the request buffer, and converts
3631 it to a string of at most <len> characters. The resulting string is
3632 stored into the next request "capture" slot, so it will possibly appear
3633 next to some captured HTTP headers. It will then automatically appear in
3634 the logs, and it will be possible to extract it using sample fetch rules
3635 to feed it into headers or anything. The length should be limited given
3636 that this size will be allocated for each capture during the whole
3637 session life. Please check section 7.3 (Fetching samples) and "capture
3638 request header" for more information.
3639
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003640 If the keyword "id" is used instead of "len", the action tries to store
3641 the captured string in a previously declared capture slot. This is useful
3642 to run captures in backends. The slot id can be declared by a previous
3643 directive "http-request capture" or with the "declare capture" keyword.
3644
Willy Tarreau09448f72014-06-25 18:12:15 +02003645 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
3646 enables tracking of sticky counters from current request. These rules
3647 do not stop evaluation and do not change default action. Three sets of
3648 counters may be simultaneously tracked by the same connection. The first
3649 "track-sc0" rule executed enables tracking of the counters of the
3650 specified table as the first set. The first "track-sc1" rule executed
3651 enables tracking of the counters of the specified table as the second
3652 set. The first "track-sc2" rule executed enables tracking of the
3653 counters of the specified table as the third set. It is a recommended
3654 practice to use the first set of counters for the per-frontend counters
3655 and the second set for the per-backend ones. But this is just a
3656 guideline, all may be used everywhere.
3657
3658 These actions take one or two arguments :
3659 <key> is mandatory, and is a sample expression rule as described
3660 in section 7.3. It describes what elements of the incoming
3661 request or connection will be analysed, extracted, combined,
3662 and used to select which table entry to update the counters.
3663
3664 <table> is an optional table to be used instead of the default one,
3665 which is the stick-table declared in the current proxy. All
3666 the counters for the matches and updates for the key will
3667 then be performed in that table until the session ends.
3668
3669 Once a "track-sc*" rule is executed, the key is looked up in the table
3670 and if it is not found, an entry is allocated for it. Then a pointer to
3671 that entry is kept during all the session's life, and this entry's
3672 counters are updated as often as possible, every time the session's
3673 counters are updated, and also systematically when the session ends.
3674 Counters are only updated for events that happen after the tracking has
3675 been started. As an exception, connection counters and request counters
3676 are systematically updated so that they reflect useful information.
3677
3678 If the entry tracks concurrent connection counters, one connection is
3679 counted for as long as the entry is tracked, and the entry will not
3680 expire during that time. Tracking counters also provides a performance
3681 advantage over just checking the keys, because only one table lookup is
3682 performed for all ACL checks that make use of it.
3683
Thierry FOURNIER236657b2015-08-19 08:25:14 +02003684 - sc-set-gpt0(<sc-id>) <int> :
3685 This action sets the GPT0 tag according to the sticky counter designated
3686 by <sc-id> and the value of <int>. The expected result is a boolean. If
3687 an error occurs, this action silently fails and the actions evaluation
3688 continues.
3689
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02003690 - sc-inc-gpc0(<sc-id>):
3691 This action increments the GPC0 counter according with the sticky counter
3692 designated by <sc-id>. If an error occurs, this action silently fails and
3693 the actions evaluation continues.
3694
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003695 - set-var(<var-name>) <expr> :
3696 Is used to set the contents of a variable. The variable is declared
3697 inline.
3698
3699 <var-name> The name of the variable starts by an indication about its
3700 scope. The allowed scopes are:
3701 "sess" : the variable is shared with all the session,
3702 "txn" : the variable is shared with all the transaction
3703 (request and response)
3704 "req" : the variable is shared only during the request
3705 processing
3706 "res" : the variable is shared only during the response
3707 processing.
3708 This prefix is followed by a name. The separator is a '.'.
3709 The name may only contain characters 'a-z', 'A-Z', '0-9',
3710 and '_'.
3711
3712 <expr> Is a standard HAProxy expression formed by a sample-fetch
3713 followed by some converters.
3714
3715 Example:
3716
3717 http-request set-var(req.my_var) req.fhdr(user-agent),lower
3718
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02003719 - set-src <expr> :
3720 Is used to set the source IP address to the value of specified
3721 expression. Useful when a proxy in front of HAProxy rewrites source IP,
3722 but provides the correct IP in a HTTP header; or you want to mask
3723 source IP for privacy.
3724
3725 <expr> Is a standard HAProxy expression formed by a sample-fetch
3726 followed by some converters.
3727
3728 Example:
3729
3730 http-request set-src hdr(x-forwarded-for)
3731 http-request set-src src,ipmask(24)
3732
3733 When set-src is successful, the source port is set to 0.
3734
Willy Tarreau2d392c22015-08-24 01:43:45 +02003735 - "silent-drop" : this stops the evaluation of the rules and makes the
3736 client-facing connection suddenly disappear using a system-dependant way
3737 that tries to prevent the client from being notified. The effect it then
3738 that the client still sees an established connection while there's none
3739 on HAProxy. The purpose is to achieve a comparable effect to "tarpit"
3740 except that it doesn't use any local resource at all on the machine
3741 running HAProxy. It can resist much higher loads than "tarpit", and slow
3742 down stronger attackers. It is important to undestand the impact of using
3743 this mechanism. All stateful equipments placed between the client and
3744 HAProxy (firewalls, proxies, load balancers) will also keep the
3745 established connection for a long time and may suffer from this action.
3746 On modern Linux systems running with enough privileges, the TCP_REPAIR
3747 socket option is used to block the emission of a TCP reset. On other
3748 systems, the socket's TTL is reduced to 1 so that the TCP reset doesn't
3749 pass the first router, though it's still delivered to local networks. Do
3750 not use it unless you fully understand how it works.
3751
Willy Tarreau20b0de52012-12-24 15:45:22 +01003752 There is no limit to the number of http-request statements per instance.
3753
3754 It is important to know that http-request rules are processed very early in
3755 the HTTP processing, just after "block" rules and before "reqdel" or "reqrep"
3756 rules. That way, headers added by "add-header"/"set-header" are visible by
3757 almost all further ACL rules.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003758
3759 Example:
Cyril Bonté78caf842010-03-10 22:41:43 +01003760 acl nagios src 192.168.129.3
3761 acl local_net src 192.168.0.0/16
3762 acl auth_ok http_auth(L1)
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003763
Cyril Bonté78caf842010-03-10 22:41:43 +01003764 http-request allow if nagios
3765 http-request allow if local_net auth_ok
3766 http-request auth realm Gimme if local_net auth_ok
3767 http-request deny
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003768
Cyril Bonté78caf842010-03-10 22:41:43 +01003769 Example:
3770 acl auth_ok http_auth_group(L1) G1
Cyril Bonté78caf842010-03-10 22:41:43 +01003771 http-request auth unless auth_ok
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003772
Willy Tarreau20b0de52012-12-24 15:45:22 +01003773 Example:
3774 http-request set-header X-Haproxy-Current-Date %T
3775 http-request set-header X-SSL %[ssl_fc]
Willy Tarreaufca42612015-08-27 17:15:05 +02003776 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id,hex]
Willy Tarreau20b0de52012-12-24 15:45:22 +01003777 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
3778 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
3779 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
3780 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
3781 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
3782 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
3783
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003784 Example:
3785 acl key req.hdr(X-Add-Acl-Key) -m found
3786 acl add path /addacl
3787 acl del path /delacl
3788
3789 acl myhost hdr(Host) -f myhost.lst
3790
3791 http-request add-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key add
3792 http-request del-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key del
3793
3794 Example:
3795 acl value req.hdr(X-Value) -m found
3796 acl setmap path /setmap
3797 acl delmap path /delmap
3798
3799 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
3800
3801 http-request set-map(map.lst) %[src] %[req.hdr(X-Value)] if setmap value
3802 http-request del-map(map.lst) %[src] if delmap
3803
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02003804 See also : "stats http-request", section 3.4 about userlists and section 7
3805 about ACL usage.
Willy Tarreauef781042010-01-27 11:53:01 +01003806
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003807http-response { allow | deny | add-header <name> <fmt> | set-nice <nice> |
Willy Tarreau51d861a2015-05-22 17:30:48 +02003808 capture <sample> id <id> | redirect <rule> |
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003809 set-header <name> <fmt> | del-header <name> |
Sasha Pachev218f0642014-06-16 12:05:59 -06003810 replace-header <name> <regex-match> <replace-fmt> |
3811 replace-value <name> <regex-match> <replace-fmt> |
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02003812 set-status <status> |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003813 set-log-level <level> | set-mark <mark> | set-tos <tos> |
3814 add-acl(<file name>) <key fmt> |
3815 del-acl(<file name>) <key fmt> |
3816 del-map(<file name>) <key fmt> |
Thierry FOURNIER90da1912015-03-05 11:17:06 +01003817 set-map(<file name>) <key fmt> <value fmt> |
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003818 set-var(<var-name>) <expr> |
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02003819 sc-inc-gpc0(<sc-id>) |
Thierry FOURNIER236657b2015-08-19 08:25:14 +02003820 sc-set-gpt0(<sc-id>) <int> |
Willy Tarreau2d392c22015-08-24 01:43:45 +02003821 silent-drop |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003822 }
Lukas Tribus2dd1d1a2013-06-19 23:34:41 +02003823 [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02003824 Access control for Layer 7 responses
3825
3826 May be used in sections: defaults | frontend | listen | backend
3827 no | yes | yes | yes
3828
3829 The http-response statement defines a set of rules which apply to layer 7
3830 processing. The rules are evaluated in their declaration order when they are
3831 met in a frontend, listen or backend section. Any rule may optionally be
3832 followed by an ACL-based condition, in which case it will only be evaluated
3833 if the condition is true. Since these rules apply on responses, the backend
3834 rules are applied first, followed by the frontend's rules.
3835
3836 The first keyword is the rule's action. Currently supported actions include :
3837 - "allow" : this stops the evaluation of the rules and lets the response
3838 pass the check. No further "http-response" rules are evaluated for the
3839 current section.
3840
3841 - "deny" : this stops the evaluation of the rules and immediately rejects
3842 the response and emits an HTTP 502 error. No further "http-response"
3843 rules are evaluated.
3844
3845 - "add-header" appends an HTTP header field whose name is specified in
3846 <name> and whose value is defined by <fmt> which follows the log-format
3847 rules (see Custom Log Format in section 8.2.4). This may be used to send
3848 a cookie to a client for example, or to pass some internal information.
3849 This rule is not final, so it is possible to add other similar rules.
3850 Note that header addition is performed immediately, so one rule might
3851 reuse the resulting header from a previous rule.
3852
3853 - "set-header" does the same as "add-header" except that the header name
3854 is first removed if it existed. This is useful when passing security
3855 information to the server, where the header must not be manipulated by
3856 external users.
3857
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003858 - "del-header" removes all HTTP header fields whose name is specified in
3859 <name>.
3860
Sasha Pachev218f0642014-06-16 12:05:59 -06003861 - "replace-header" matches the regular expression in all occurrences of
3862 header field <name> according to <match-regex>, and replaces them with
3863 the <replace-fmt> argument. Format characters are allowed in replace-fmt
3864 and work like in <fmt> arguments in "add-header". The match is only
3865 case-sensitive. It is important to understand that this action only
3866 considers whole header lines, regardless of the number of values they
3867 may contain. This usage is suited to headers naturally containing commas
3868 in their value, such as Set-Cookie, Expires and so on.
3869
3870 Example:
3871
3872 http-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
3873
3874 applied to:
3875
3876 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
3877
3878 outputs:
3879
3880 Set-Cookie: C=1;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT
3881
3882 assuming the backend IP is 192.168.1.20.
3883
3884 - "replace-value" works like "replace-header" except that it matches the
3885 regex against every comma-delimited value of the header field <name>
3886 instead of the entire header. This is suited for all headers which are
3887 allowed to carry more than one value. An example could be the Accept
3888 header.
3889
3890 Example:
3891
3892 http-response replace-value Cache-control ^public$ private
3893
3894 applied to:
3895
3896 Cache-Control: max-age=3600, public
3897
3898 outputs:
3899
3900 Cache-Control: max-age=3600, private
3901
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02003902 - "set-status" replaces the response status code with <status> which must
3903 be an integer between 100 and 999. Note that the reason is automatically
3904 adapted to the new code.
3905
3906 Example:
3907
3908 # return "431 Request Header Fields Too Large"
3909 http-response set-status 431
3910
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003911 - "set-nice" sets the "nice" factor of the current request being processed.
3912 It only has effect against the other requests being processed at the same
3913 time. The default value is 0, unless altered by the "nice" setting on the
3914 "bind" line. The accepted range is -1024..1024. The higher the value, the
3915 nicest the request will be. Lower values will make the request more
3916 important than other ones. This can be useful to improve the speed of
3917 some requests, or lower the priority of non-important requests. Using
3918 this setting without prior experimentation can cause some major slowdown.
3919
Willy Tarreau9a355ec2013-06-11 17:45:46 +02003920 - "set-log-level" is used to change the log level of the current request
3921 when a certain condition is met. Valid levels are the 8 syslog levels
3922 (see the "log" keyword) plus the special level "silent" which disables
3923 logging for this request. This rule is not final so the last matching
3924 rule wins. This rule can be useful to disable health checks coming from
3925 another equipment.
3926
Willy Tarreau42cf39e2013-06-11 18:51:32 +02003927 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
3928 the client to the value passed in <tos> on platforms which support this.
3929 This value represents the whole 8 bits of the IP TOS field, and can be
3930 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
3931 that only the 6 higher bits are used in DSCP or TOS, and the two lower
3932 bits are always 0. This can be used to adjust some routing behaviour on
3933 border routers based on some information from the request. See RFC 2474,
3934 2597, 3260 and 4594 for more information.
3935
Willy Tarreau51347ed2013-06-11 19:34:13 +02003936 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
3937 client to the value passed in <mark> on platforms which support it. This
3938 value is an unsigned 32 bit value which can be matched by netfilter and
3939 by the routing table. It can be expressed both in decimal or hexadecimal
3940 format (prefixed by "0x"). This can be useful to force certain packets to
3941 take a different route (for example a cheaper network path for bulk
3942 downloads). This works on Linux kernels 2.6.32 and above and requires
3943 admin privileges.
3944
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003945 - "add-acl" is used to add a new entry into an ACL. The ACL must be loaded
3946 from a file (even a dummy empty file). The file name of the ACL to be
3947 updated is passed between parentheses. It takes one argument: <key fmt>,
3948 which follows log-format rules, to collect content of the new entry. It
3949 performs a lookup in the ACL before insertion, to avoid duplicated (or
3950 more) values. This lookup is done by a linear search and can be expensive
3951 with large lists! It is the equivalent of the "add acl" command from the
3952 stats socket, but can be triggered by an HTTP response.
3953
3954 - "del-acl" is used to delete an entry from an ACL. The ACL must be loaded
3955 from a file (even a dummy empty file). The file name of the ACL to be
3956 updated is passed between parentheses. It takes one argument: <key fmt>,
3957 which follows log-format rules, to collect content of the entry to delete.
3958 It is the equivalent of the "del acl" command from the stats socket, but
3959 can be triggered by an HTTP response.
3960
3961 - "del-map" is used to delete an entry from a MAP. The MAP must be loaded
3962 from a file (even a dummy empty file). The file name of the MAP to be
3963 updated is passed between parentheses. It takes one argument: <key fmt>,
3964 which follows log-format rules, to collect content of the entry to delete.
3965 It takes one argument: "file name" It is the equivalent of the "del map"
3966 command from the stats socket, but can be triggered by an HTTP response.
3967
3968 - "set-map" is used to add a new entry into a MAP. The MAP must be loaded
3969 from a file (even a dummy empty file). The file name of the MAP to be
3970 updated is passed between parentheses. It takes 2 arguments: <key fmt>,
3971 which follows log-format rules, used to collect MAP key, and <value fmt>,
3972 which follows log-format rules, used to collect content for the new entry.
3973 It performs a lookup in the MAP before insertion, to avoid duplicated (or
3974 more) values. This lookup is done by a linear search and can be expensive
3975 with large lists! It is the equivalent of the "set map" command from the
3976 stats socket, but can be triggered by an HTTP response.
3977
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02003978 - capture <sample> id <id> :
3979 captures sample expression <sample> from the response buffer, and converts
3980 it to a string. The resulting string is stored into the next request
3981 "capture" slot, so it will possibly appear next to some captured HTTP
3982 headers. It will then automatically appear in the logs, and it will be
3983 possible to extract it using sample fetch rules to feed it into headers or
3984 anything. Please check section 7.3 (Fetching samples) and "capture
3985 response header" for more information.
3986
3987 The keyword "id" is the id of the capture slot which is used for storing
3988 the string. The capture slot must be defined in an associated frontend.
3989 This is useful to run captures in backends. The slot id can be declared by
3990 a previous directive "http-response capture" or with the "declare capture"
3991 keyword.
3992
Willy Tarreau51d861a2015-05-22 17:30:48 +02003993 - "redirect" : this performs an HTTP redirection based on a redirect rule.
3994 This supports a format string similarly to "http-request redirect" rules,
3995 with the exception that only the "location" type of redirect is possible
3996 on the response. See the "redirect" keyword for the rule's syntax. When
3997 a redirect rule is applied during a response, connections to the server
3998 are closed so that no data can be forwarded from the server to the client.
3999
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004000 - set-var(<var-name>) expr:
4001 Is used to set the contents of a variable. The variable is declared
4002 inline.
4003
4004 <var-name> The name of the variable starts by an indication about its
4005 scope. The allowed scopes are:
4006 "sess" : the variable is shared with all the session,
4007 "txn" : the variable is shared with all the transaction
4008 (request and response)
4009 "req" : the variable is shared only during the request
4010 processing
4011 "res" : the variable is shared only during the response
4012 processing.
4013 This prefix is followed by a name. The separator is a '.'.
4014 The name may only contain characters 'a-z', 'A-Z', '0-9',
4015 and '_'.
4016
4017 <expr> Is a standard HAProxy expression formed by a sample-fetch
4018 followed by some converters.
4019
4020 Example:
4021
4022 http-response set-var(sess.last_redir) res.hdr(location)
4023
Thierry FOURNIER236657b2015-08-19 08:25:14 +02004024 - sc-set-gpt0(<sc-id>) <int> :
4025 This action sets the GPT0 tag according to the sticky counter designated
4026 by <sc-id> and the value of <int>. The expected result is a boolean. If
4027 an error occurs, this action silently fails and the actions evaluation
4028 continues.
4029
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02004030 - sc-inc-gpc0(<sc-id>):
4031 This action increments the GPC0 counter according with the sticky counter
4032 designated by <sc-id>. If an error occurs, this action silently fails and
4033 the actions evaluation continues.
4034
Willy Tarreau2d392c22015-08-24 01:43:45 +02004035 - "silent-drop" : this stops the evaluation of the rules and makes the
4036 client-facing connection suddenly disappear using a system-dependant way
4037 that tries to prevent the client from being notified. The effect it then
4038 that the client still sees an established connection while there's none
4039 on HAProxy. The purpose is to achieve a comparable effect to "tarpit"
4040 except that it doesn't use any local resource at all on the machine
4041 running HAProxy. It can resist much higher loads than "tarpit", and slow
4042 down stronger attackers. It is important to undestand the impact of using
4043 this mechanism. All stateful equipments placed between the client and
4044 HAProxy (firewalls, proxies, load balancers) will also keep the
4045 established connection for a long time and may suffer from this action.
4046 On modern Linux systems running with enough privileges, the TCP_REPAIR
4047 socket option is used to block the emission of a TCP reset. On other
4048 systems, the socket's TTL is reduced to 1 so that the TCP reset doesn't
4049 pass the first router, though it's still delivered to local networks. Do
4050 not use it unless you fully understand how it works.
4051
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004052 There is no limit to the number of http-response statements per instance.
4053
Godbach09250262013-07-02 01:19:15 +08004054 It is important to know that http-response rules are processed very early in
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004055 the HTTP processing, before "reqdel" or "reqrep" rules. That way, headers
4056 added by "add-header"/"set-header" are visible by almost all further ACL
4057 rules.
4058
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004059 Example:
4060 acl key_acl res.hdr(X-Acl-Key) -m found
4061
4062 acl myhost hdr(Host) -f myhost.lst
4063
4064 http-response add-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
4065 http-response del-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
4066
4067 Example:
4068 acl value res.hdr(X-Value) -m found
4069
4070 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
4071
4072 http-response set-map(map.lst) %[src] %[res.hdr(X-Value)] if value
4073 http-response del-map(map.lst) %[src] if ! value
4074
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004075 See also : "http-request", section 3.4 about userlists and section 7 about
4076 ACL usage.
4077
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02004078
Willy Tarreau30631952015-08-06 15:05:24 +02004079http-reuse { never | safe | aggressive | always }
4080 Declare how idle HTTP connections may be shared between requests
4081
4082 May be used in sections: defaults | frontend | listen | backend
4083 yes | no | yes | yes
4084
4085 By default, a connection established between haproxy and the backend server
4086 belongs to the session that initiated it. The downside is that between the
4087 response and the next request, the connection remains idle and is not used.
4088 In many cases for performance reasons it is desirable to make it possible to
4089 reuse these idle connections to serve other requests from different sessions.
4090 This directive allows to tune this behaviour.
4091
4092 The argument indicates the desired connection reuse strategy :
4093
4094 - "never" : idle connections are never shared between sessions. This is
4095 the default choice. It may be enforced to cancel a different
4096 strategy inherited from a defaults section or for
4097 troubleshooting. For example, if an old bogus application
4098 considers that multiple requests over the same connection come
4099 from the same client and it is not possible to fix the
4100 application, it may be desirable to disable connection sharing
4101 in a single backend. An example of such an application could
4102 be an old haproxy using cookie insertion in tunnel mode and
4103 not checking any request past the first one.
4104
4105 - "safe" : this is the recommended strategy. The first request of a
4106 session is always sent over its own connection, and only
4107 subsequent requests may be dispatched over other existing
4108 connections. This ensures that in case the server closes the
4109 connection when the request is being sent, the browser can
4110 decide to silently retry it. Since it is exactly equivalent to
4111 regular keep-alive, there should be no side effects.
4112
4113 - "aggressive" : this mode may be useful in webservices environments where
4114 all servers are not necessarily known and where it would be
4115 appreciable to deliver most first requests over existing
4116 connections. In this case, first requests are only delivered
4117 over existing connections that have been reused at least once,
4118 proving that the server correctly supports connection reuse.
4119 It should only be used when it's sure that the client can
4120 retry a failed request once in a while and where the benefit
4121 of aggressive connection reuse significantly outweights the
4122 downsides of rare connection failures.
4123
4124 - "always" : this mode is only recommended when the path to the server is
4125 known for never breaking existing connections quickly after
4126 releasing them. It allows the first request of a session to be
4127 sent to an existing connection. This can provide a significant
4128 performance increase over the "safe" strategy when the backend
4129 is a cache farm, since such components tend to show a
4130 consistent behaviour and will benefit from the connection
4131 sharing. It is recommended that the "http-keep-alive" timeout
4132 remains low in this mode so that no dead connections remain
4133 usable. In most cases, this will lead to the same performance
4134 gains as "aggressive" but with more risks. It should only be
4135 used when it improves the situation over "aggressive".
4136
4137 When http connection sharing is enabled, a great care is taken to respect the
4138 connection properties and compatiblities. Specifically :
4139 - connections made with "usesrc" followed by a client-dependant value
4140 ("client", "clientip", "hdr_ip") are marked private and never shared ;
4141
4142 - connections sent to a server with a TLS SNI extension are marked private
4143 and are never shared ;
4144
4145 - connections receiving a status code 401 or 407 expect some authentication
4146 to be sent in return. Due to certain bogus authentication schemes (such
4147 as NTLM) relying on the connection, these connections are marked private
4148 and are never shared ;
4149
4150 No connection pool is involved, once a session dies, the last idle connection
4151 it was attached to is deleted at the same time. This ensures that connections
4152 may not last after all sessions are closed.
4153
4154 Note: connection reuse improves the accuracy of the "server maxconn" setting,
4155 because almost no new connection will be established while idle connections
4156 remain available. This is particularly true with the "always" strategy.
4157
4158 See also : "option http-keep-alive", "server maxconn"
4159
4160
Mark Lamourinec2247f02012-01-04 13:02:01 -05004161http-send-name-header [<header>]
4162 Add the server name to a request. Use the header string given by <header>
4163
4164 May be used in sections: defaults | frontend | listen | backend
4165 yes | no | yes | yes
4166
4167 Arguments :
4168
4169 <header> The header string to use to send the server name
4170
4171 The "http-send-name-header" statement causes the name of the target
4172 server to be added to the headers of an HTTP request. The name
4173 is added with the header string proved.
4174
4175 See also : "server"
4176
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01004177id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02004178 Set a persistent ID to a proxy.
4179 May be used in sections : defaults | frontend | listen | backend
4180 no | yes | yes | yes
4181 Arguments : none
4182
4183 Set a persistent ID for the proxy. This ID must be unique and positive.
4184 An unused ID will automatically be assigned if unset. The first assigned
4185 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01004186
4187
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004188ignore-persist { if | unless } <condition>
4189 Declare a condition to ignore persistence
4190 May be used in sections: defaults | frontend | listen | backend
4191 no | yes | yes | yes
4192
4193 By default, when cookie persistence is enabled, every requests containing
4194 the cookie are unconditionally persistent (assuming the target server is up
4195 and running).
4196
4197 The "ignore-persist" statement allows one to declare various ACL-based
4198 conditions which, when met, will cause a request to ignore persistence.
4199 This is sometimes useful to load balance requests for static files, which
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03004200 often don't require persistence. This can also be used to fully disable
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004201 persistence for a specific User-Agent (for example, some web crawler bots).
4202
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004203 The persistence is ignored when an "if" condition is met, or unless an
4204 "unless" condition is met.
4205
4206 See also : "force-persist", "cookie", and section 7 about ACL usage.
4207
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004208load-server-state-from-file { global | local | none }
4209 Allow seamless reload of HAProxy
4210 May be used in sections: defaults | frontend | listen | backend
4211 yes | no | yes | yes
4212
4213 This directive points HAProxy to a file where server state from previous
4214 running process has been saved. That way, when starting up, before handling
4215 traffic, the new process can apply old states to servers exactly has if no
4216 reload occured. The purpose of the "load-server-state-from-file" directive is
4217 to tell haproxy which file to use. For now, only 2 arguments to either prevent
4218 loading state or load states from a file containing all backends and servers.
4219 The state file can be generated by running the command "show servers state"
4220 over the stats socket and redirect output.
4221
4222 The format of the file is versionned and is very specific. To understand it,
4223 please read the documentation of the "show servers state" command (chapter
4224 9.2).
4225
4226 Arguments:
4227 global load the content of the file pointed by the global directive
4228 named "server-state-file".
4229
4230 local load the content of the file pointed by the directive
4231 "server-state-file-name" if set. If not set, then the backend
4232 name is used as a file name.
4233
4234 none don't load any stat for this backend
4235
4236 Notes:
4237 - server's IP address is not updated unless DNS resolution is enabled on
4238 the server. It means that if a server IP address has been changed using
4239 the stat socket, this information won't be re-applied after reloading.
4240
4241 - server's weight is applied from previous running process unless it has
4242 has changed between previous and new configuration files.
4243
4244 Example 1:
4245
4246 Minimal configuration:
4247
4248 global
4249 stats socket /tmp/socket
4250 server-state-file /tmp/server_state
4251
4252 defaults
4253 load-server-state-from-file global
4254
4255 backend bk
4256 server s1 127.0.0.1:22 check weight 11
4257 server s2 127.0.0.1:22 check weight 12
4258
4259 Then one can run :
4260
4261 socat /tmp/socket - <<< "show servers state" > /tmp/server_state
4262
4263 Content of the file /tmp/server_state would be like this:
4264
4265 1
4266 # <field names skipped for the doc example>
4267 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
4268 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
4269
4270 Example 2:
4271
4272 Minimal configuration:
4273
4274 global
4275 stats socket /tmp/socket
4276 server-state-base /etc/haproxy/states
4277
4278 defaults
4279 load-server-state-from-file local
4280
4281 backend bk
4282 server s1 127.0.0.1:22 check weight 11
4283 server s2 127.0.0.1:22 check weight 12
4284
4285 Then one can run :
4286
4287 socat /tmp/socket - <<< "show servers state bk" > /etc/haproxy/states/bk
4288
4289 Content of the file /etc/haproxy/states/bk would be like this:
4290
4291 1
4292 # <field names skipped for the doc example>
4293 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
4294 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
4295
4296 See also: "server-state-file", "server-state-file-name", and
4297 "show servers state"
4298
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004299
Willy Tarreau2769aa02007-12-27 18:26:09 +01004300log global
Willy Tarreau18324f52014-06-27 18:10:07 +02004301log <address> [len <length>] <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02004302no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01004303 Enable per-instance logging of events and traffic.
4304 May be used in sections : defaults | frontend | listen | backend
4305 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02004306
4307 Prefix :
4308 no should be used when the logger list must be flushed. For example,
4309 if you don't want to inherit from the default logger list. This
4310 prefix does not allow arguments.
4311
Willy Tarreau2769aa02007-12-27 18:26:09 +01004312 Arguments :
4313 global should be used when the instance's logging parameters are the
4314 same as the global ones. This is the most common usage. "global"
4315 replaces <address>, <facility> and <level> with those of the log
4316 entries found in the "global" section. Only one "log global"
4317 statement may be used per instance, and this form takes no other
4318 parameter.
4319
4320 <address> indicates where to send the logs. It takes the same format as
4321 for the "global" section's logs, and can be one of :
4322
4323 - An IPv4 address optionally followed by a colon (':') and a UDP
4324 port. If no port is specified, 514 is used by default (the
4325 standard syslog port).
4326
David du Colombier24bb5f52011-03-17 10:40:23 +01004327 - An IPv6 address followed by a colon (':') and optionally a UDP
4328 port. If no port is specified, 514 is used by default (the
4329 standard syslog port).
4330
Willy Tarreau2769aa02007-12-27 18:26:09 +01004331 - A filesystem path to a UNIX domain socket, keeping in mind
4332 considerations for chroot (be sure the path is accessible
4333 inside the chroot) and uid/gid (be sure the path is
4334 appropriately writeable).
4335
William Lallemandb2f07452015-05-12 14:27:13 +02004336 You may want to reference some environment variables in the
4337 address parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01004338
Willy Tarreau18324f52014-06-27 18:10:07 +02004339 <length> is an optional maximum line length. Log lines larger than this
4340 value will be truncated before being sent. The reason is that
4341 syslog servers act differently on log line length. All servers
4342 support the default value of 1024, but some servers simply drop
4343 larger lines while others do log them. If a server supports long
4344 lines, it may make sense to set this value here in order to avoid
4345 truncating long lines. Similarly, if a server drops long lines,
4346 it is preferable to truncate them before sending them. Accepted
4347 values are 80 to 65535 inclusive. The default value of 1024 is
4348 generally fine for all standard usages. Some specific cases of
4349 long captures or JSON-formated logs may require larger values.
4350
Willy Tarreau2769aa02007-12-27 18:26:09 +01004351 <facility> must be one of the 24 standard syslog facilities :
4352
4353 kern user mail daemon auth syslog lpr news
4354 uucp cron auth2 ftp ntp audit alert cron2
4355 local0 local1 local2 local3 local4 local5 local6 local7
4356
4357 <level> is optional and can be specified to filter outgoing messages. By
4358 default, all messages are sent. If a level is specified, only
4359 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02004360 will be sent. An optional minimum level can be specified. If it
4361 is set, logs emitted with a more severe level than this one will
4362 be capped to this level. This is used to avoid sending "emerg"
4363 messages on all terminals on some default syslog configurations.
4364 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01004365
4366 emerg alert crit err warning notice info debug
4367
William Lallemand0f99e342011-10-12 17:50:54 +02004368 It is important to keep in mind that it is the frontend which decides what to
4369 log from a connection, and that in case of content switching, the log entries
4370 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01004371
4372 However, backend log declaration define how and where servers status changes
4373 will be logged. Level "notice" will be used to indicate a server going up,
4374 "warning" will be used for termination signals and definitive service
4375 termination, and "alert" will be used for when a server goes down.
4376
4377 Note : According to RFC3164, messages are truncated to 1024 bytes before
4378 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004379
4380 Example :
4381 log global
Willy Tarreauf7edefa2009-05-10 17:20:05 +02004382 log 127.0.0.1:514 local0 notice # only send important events
4383 log 127.0.0.1:514 local0 notice notice # same but limit output level
William Lallemandb2f07452015-05-12 14:27:13 +02004384 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
Willy Tarreaudad36a32013-03-11 01:20:04 +01004385
Willy Tarreau2769aa02007-12-27 18:26:09 +01004386
William Lallemand48940402012-01-30 16:47:22 +01004387log-format <string>
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01004388 Specifies the log format string to use for traffic logs
4389 May be used in sections: defaults | frontend | listen | backend
4390 yes | yes | yes | no
William Lallemand48940402012-01-30 16:47:22 +01004391
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01004392 This directive specifies the log format string that will be used for all logs
4393 resulting from traffic passing through the frontend using this line. If the
4394 directive is used in a defaults section, all subsequent frontends will use
4395 the same log format. Please see section 8.2.4 which covers the log format
4396 string in depth.
William Lallemand48940402012-01-30 16:47:22 +01004397
Dragan Dosen7ad31542015-09-28 17:16:47 +02004398log-format-sd <string>
4399 Specifies the RFC5424 structured-data log format string
4400 May be used in sections: defaults | frontend | listen | backend
4401 yes | yes | yes | no
4402
4403 This directive specifies the RFC5424 structured-data log format string that
4404 will be used for all logs resulting from traffic passing through the frontend
4405 using this line. If the directive is used in a defaults section, all
4406 subsequent frontends will use the same log format. Please see section 8.2.4
4407 which covers the log format string in depth.
4408
4409 See https://tools.ietf.org/html/rfc5424#section-6.3 for more information
4410 about the RFC5424 structured-data part.
4411
4412 Note : This log format string will be used only for loggers that have set
4413 log format to "rfc5424".
4414
4415 Example :
4416 log-format-sd [exampleSDID@1234\ bytes=\"%B\"\ status=\"%ST\"]
4417
4418
Willy Tarreau094af4e2015-01-07 15:03:42 +01004419log-tag <string>
4420 Specifies the log tag to use for all outgoing logs
4421 May be used in sections: defaults | frontend | listen | backend
4422 yes | yes | yes | yes
4423
4424 Sets the tag field in the syslog header to this string. It defaults to the
4425 log-tag set in the global section, otherwise the program name as launched
4426 from the command line, which usually is "haproxy". Sometimes it can be useful
4427 to differentiate between multiple processes running on the same host, or to
4428 differentiate customer instances running in the same process. In the backend,
4429 logs about servers up/down will use this tag. As a hint, it can be convenient
4430 to set a log-tag related to a hosted customer in a defaults section then put
4431 all the frontends and backends for that customer, then start another customer
4432 in a new defaults section. See also the global "log-tag" directive.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004433
Willy Tarreauc35362a2014-04-25 13:58:37 +02004434max-keep-alive-queue <value>
4435 Set the maximum server queue size for maintaining keep-alive connections
4436 May be used in sections: defaults | frontend | listen | backend
4437 yes | no | yes | yes
4438
4439 HTTP keep-alive tries to reuse the same server connection whenever possible,
4440 but sometimes it can be counter-productive, for example if a server has a lot
4441 of connections while other ones are idle. This is especially true for static
4442 servers.
4443
4444 The purpose of this setting is to set a threshold on the number of queued
4445 connections at which haproxy stops trying to reuse the same server and prefers
4446 to find another one. The default value, -1, means there is no limit. A value
4447 of zero means that keep-alive requests will never be queued. For very close
4448 servers which can be reached with a low latency and which are not sensible to
4449 breaking keep-alive, a low value is recommended (eg: local static server can
4450 use a value of 10 or less). For remote servers suffering from a high latency,
4451 higher values might be needed to cover for the latency and/or the cost of
4452 picking a different server.
4453
4454 Note that this has no impact on responses which are maintained to the same
4455 server consecutively to a 401 response. They will still go to the same server
4456 even if they have to be queued.
4457
4458 See also : "option http-server-close", "option prefer-last-server", server
4459 "maxconn" and cookie persistence.
4460
4461
Willy Tarreau2769aa02007-12-27 18:26:09 +01004462maxconn <conns>
4463 Fix the maximum number of concurrent connections on a frontend
4464 May be used in sections : defaults | frontend | listen | backend
4465 yes | yes | yes | no
4466 Arguments :
4467 <conns> is the maximum number of concurrent connections the frontend will
4468 accept to serve. Excess connections will be queued by the system
4469 in the socket's listen queue and will be served once a connection
4470 closes.
4471
4472 If the system supports it, it can be useful on big sites to raise this limit
4473 very high so that haproxy manages connection queues, instead of leaving the
4474 clients with unanswered connection attempts. This value should not exceed the
4475 global maxconn. Also, keep in mind that a connection contains two buffers
4476 of 8kB each, as well as some other data resulting in about 17 kB of RAM being
4477 consumed per established connection. That means that a medium system equipped
4478 with 1GB of RAM can withstand around 40000-50000 concurrent connections if
4479 properly tuned.
4480
4481 Also, when <conns> is set to large values, it is possible that the servers
4482 are not sized to accept such loads, and for this reason it is generally wise
4483 to assign them some reasonable connection limits.
4484
Vincent Bernat6341be52012-06-27 17:18:30 +02004485 By default, this value is set to 2000.
4486
Willy Tarreau2769aa02007-12-27 18:26:09 +01004487 See also : "server", global section's "maxconn", "fullconn"
4488
4489
4490mode { tcp|http|health }
4491 Set the running mode or protocol of the instance
4492 May be used in sections : defaults | frontend | listen | backend
4493 yes | yes | yes | yes
4494 Arguments :
4495 tcp The instance will work in pure TCP mode. A full-duplex connection
4496 will be established between clients and servers, and no layer 7
4497 examination will be performed. This is the default mode. It
4498 should be used for SSL, SSH, SMTP, ...
4499
4500 http The instance will work in HTTP mode. The client request will be
4501 analyzed in depth before connecting to any server. Any request
4502 which is not RFC-compliant will be rejected. Layer 7 filtering,
4503 processing and switching will be possible. This is the mode which
4504 brings HAProxy most of its value.
4505
4506 health The instance will work in "health" mode. It will just reply "OK"
Willy Tarreau82569f92012-09-27 23:48:56 +02004507 to incoming connections and close the connection. Alternatively,
4508 If the "httpchk" option is set, "HTTP/1.0 200 OK" will be sent
4509 instead. Nothing will be logged in either case. This mode is used
4510 to reply to external components health checks. This mode is
4511 deprecated and should not be used anymore as it is possible to do
4512 the same and even better by combining TCP or HTTP modes with the
4513 "monitor" keyword.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004514
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004515 When doing content switching, it is mandatory that the frontend and the
4516 backend are in the same mode (generally HTTP), otherwise the configuration
4517 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004518
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004519 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01004520 defaults http_instances
4521 mode http
4522
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004523 See also : "monitor", "monitor-net"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004524
Willy Tarreau0ba27502007-12-24 16:55:16 +01004525
Cyril Bontéf0c60612010-02-06 14:44:47 +01004526monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01004527 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004528 May be used in sections : defaults | frontend | listen | backend
4529 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01004530 Arguments :
4531 if <cond> the monitor request will fail if the condition is satisfied,
4532 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004533 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01004534 are met, for instance a low number of servers both in a
4535 backend and its backup.
4536
4537 unless <cond> the monitor request will succeed only if the condition is
4538 satisfied, and will fail otherwise. Such a condition may be
4539 based on a test on the presence of a minimum number of active
4540 servers in a list of backends.
4541
4542 This statement adds a condition which can force the response to a monitor
4543 request to report a failure. By default, when an external component queries
4544 the URI dedicated to monitoring, a 200 response is returned. When one of the
4545 conditions above is met, haproxy will return 503 instead of 200. This is
4546 very useful to report a site failure to an external component which may base
4547 routing advertisements between multiple sites on the availability reported by
4548 haproxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02004549 criterion. Note that "monitor fail" only works in HTTP mode. Both status
4550 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004551
4552 Example:
4553 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01004554 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01004555 acl site_dead nbsrv(dynamic) lt 2
4556 acl site_dead nbsrv(static) lt 2
4557 monitor-uri /site_alive
4558 monitor fail if site_dead
4559
Willy Tarreauae94d4d2011-05-11 16:28:49 +02004560 See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004561
4562
4563monitor-net <source>
4564 Declare a source network which is limited to monitor requests
4565 May be used in sections : defaults | frontend | listen | backend
4566 yes | yes | yes | no
4567 Arguments :
4568 <source> is the source IPv4 address or network which will only be able to
4569 get monitor responses to any request. It can be either an IPv4
4570 address, a host name, or an address followed by a slash ('/')
4571 followed by a mask.
4572
4573 In TCP mode, any connection coming from a source matching <source> will cause
4574 the connection to be immediately closed without any log. This allows another
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004575 equipment to probe the port and verify that it is still listening, without
Willy Tarreau2769aa02007-12-27 18:26:09 +01004576 forwarding the connection to a remote server.
4577
4578 In HTTP mode, a connection coming from a source matching <source> will be
4579 accepted, the following response will be sent without waiting for a request,
4580 then the connection will be closed : "HTTP/1.0 200 OK". This is normally
4581 enough for any front-end HTTP probe to detect that the service is UP and
Willy Tarreau82569f92012-09-27 23:48:56 +02004582 running without forwarding the request to a backend server. Note that this
4583 response is sent in raw format, without any transformation. This is important
4584 as it means that it will not be SSL-encrypted on SSL listeners.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004585
Willy Tarreau82569f92012-09-27 23:48:56 +02004586 Monitor requests are processed very early, just after tcp-request connection
4587 ACLs which are the only ones able to block them. These connections are short
4588 lived and never wait for any data from the client. They cannot be logged, and
4589 it is the intended purpose. They are only used to report HAProxy's health to
4590 an upper component, nothing more. Please note that "monitor fail" rules do
4591 not apply to connections intercepted by "monitor-net".
Willy Tarreau2769aa02007-12-27 18:26:09 +01004592
Willy Tarreau95cd2832010-03-04 23:36:33 +01004593 Last, please note that only one "monitor-net" statement can be specified in
4594 a frontend. If more than one is found, only the last one will be considered.
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004595
Willy Tarreau2769aa02007-12-27 18:26:09 +01004596 Example :
4597 # addresses .252 and .253 are just probing us.
4598 frontend www
4599 monitor-net 192.168.0.252/31
4600
4601 See also : "monitor fail", "monitor-uri"
4602
4603
4604monitor-uri <uri>
4605 Intercept a URI used by external components' monitor requests
4606 May be used in sections : defaults | frontend | listen | backend
4607 yes | yes | yes | no
4608 Arguments :
4609 <uri> is the exact URI which we want to intercept to return HAProxy's
4610 health status instead of forwarding the request.
4611
4612 When an HTTP request referencing <uri> will be received on a frontend,
4613 HAProxy will not forward it nor log it, but instead will return either
4614 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
4615 conditions defined with "monitor fail". This is normally enough for any
4616 front-end HTTP probe to detect that the service is UP and running without
4617 forwarding the request to a backend server. Note that the HTTP method, the
4618 version and all headers are ignored, but the request must at least be valid
4619 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
4620
4621 Monitor requests are processed very early. It is not possible to block nor
4622 divert them using ACLs. They cannot be logged either, and it is the intended
4623 purpose. They are only used to report HAProxy's health to an upper component,
4624 nothing more. However, it is possible to add any number of conditions using
4625 "monitor fail" and ACLs so that the result can be adjusted to whatever check
4626 can be imagined (most often the number of available servers in a backend).
4627
4628 Example :
4629 # Use /haproxy_test to report haproxy's status
4630 frontend www
4631 mode http
4632 monitor-uri /haproxy_test
4633
4634 See also : "monitor fail", "monitor-net"
4635
Willy Tarreau0ba27502007-12-24 16:55:16 +01004636
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004637option abortonclose
4638no option abortonclose
4639 Enable or disable early dropping of aborted requests pending in queues.
4640 May be used in sections : defaults | frontend | listen | backend
4641 yes | no | yes | yes
4642 Arguments : none
4643
4644 In presence of very high loads, the servers will take some time to respond.
4645 The per-instance connection queue will inflate, and the response time will
4646 increase respective to the size of the queue times the average per-session
4647 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01004648 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004649 the queue, and slowing down other users, and the servers as well, because the
4650 request will eventually be served, then aborted at the first error
4651 encountered while delivering the response.
4652
4653 As there is no way to distinguish between a full STOP and a simple output
4654 close on the client side, HTTP agents should be conservative and consider
4655 that the client might only have closed its output channel while waiting for
4656 the response. However, this introduces risks of congestion when lots of users
4657 do the same, and is completely useless nowadays because probably no client at
4658 all will close the session while waiting for the response. Some HTTP agents
Willy Tarreaud72758d2010-01-12 10:42:19 +01004659 support this behaviour (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004660 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01004661 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004662 of being the single component to break rare but valid traffic is extremely
4663 low, which adds to the temptation to be able to abort a session early while
4664 still not served and not pollute the servers.
4665
4666 In HAProxy, the user can choose the desired behaviour using the option
4667 "abortonclose". By default (without the option) the behaviour is HTTP
4668 compliant and aborted requests will be served. But when the option is
4669 specified, a session with an incoming channel closed will be aborted while
4670 it is still possible, either pending in the queue for a connection slot, or
4671 during the connection establishment if the server has not yet acknowledged
4672 the connection request. This considerably reduces the queue size and the load
4673 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01004674 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004675
4676 If this option has been enabled in a "defaults" section, it can be disabled
4677 in a specific instance by prepending the "no" keyword before it.
4678
4679 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
4680
4681
Willy Tarreau4076a152009-04-02 15:18:36 +02004682option accept-invalid-http-request
4683no option accept-invalid-http-request
4684 Enable or disable relaxing of HTTP request parsing
4685 May be used in sections : defaults | frontend | listen | backend
4686 yes | yes | yes | no
4687 Arguments : none
4688
Willy Tarreau91852eb2015-05-01 13:26:00 +02004689 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02004690 means that invalid characters in header names are not permitted and cause an
4691 error to be returned to the client. This is the desired behaviour as such
4692 forbidden characters are essentially used to build attacks exploiting server
4693 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
4694 server will emit invalid header names for whatever reason (configuration,
4695 implementation) and the issue will not be immediately fixed. In such a case,
4696 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01004697 even if that does not make sense, by specifying this option. Similarly, the
4698 list of characters allowed to appear in a URI is well defined by RFC3986, and
4699 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
4700 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
4701 not allowed at all. Haproxy always blocks a number of them (0..32, 127). The
Willy Tarreau91852eb2015-05-01 13:26:00 +02004702 remaining ones are blocked by default unless this option is enabled. This
Willy Tarreau13317662015-05-01 13:47:08 +02004703 option also relaxes the test on the HTTP version, it allows HTTP/0.9 requests
4704 to pass through (no version specified) and multiple digits for both the major
4705 and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02004706
4707 This option should never be enabled by default as it hides application bugs
4708 and open security breaches. It should only be deployed after a problem has
4709 been confirmed.
4710
4711 When this option is enabled, erroneous header names will still be accepted in
4712 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01004713 analysis using the "show errors" request on the UNIX stats socket. Similarly,
4714 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02004715 also helps confirming that the issue has been solved.
4716
4717 If this option has been enabled in a "defaults" section, it can be disabled
4718 in a specific instance by prepending the "no" keyword before it.
4719
4720 See also : "option accept-invalid-http-response" and "show errors" on the
4721 stats socket.
4722
4723
4724option accept-invalid-http-response
4725no option accept-invalid-http-response
4726 Enable or disable relaxing of HTTP response parsing
4727 May be used in sections : defaults | frontend | listen | backend
4728 yes | no | yes | yes
4729 Arguments : none
4730
Willy Tarreau91852eb2015-05-01 13:26:00 +02004731 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02004732 means that invalid characters in header names are not permitted and cause an
4733 error to be returned to the client. This is the desired behaviour as such
4734 forbidden characters are essentially used to build attacks exploiting server
4735 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
4736 server will emit invalid header names for whatever reason (configuration,
4737 implementation) and the issue will not be immediately fixed. In such a case,
4738 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau91852eb2015-05-01 13:26:00 +02004739 even if that does not make sense, by specifying this option. This option also
4740 relaxes the test on the HTTP version format, it allows multiple digits for
4741 both the major and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02004742
4743 This option should never be enabled by default as it hides application bugs
4744 and open security breaches. It should only be deployed after a problem has
4745 been confirmed.
4746
4747 When this option is enabled, erroneous header names will still be accepted in
4748 responses, but the complete response will be captured in order to permit
4749 later analysis using the "show errors" request on the UNIX stats socket.
4750 Doing this also helps confirming that the issue has been solved.
4751
4752 If this option has been enabled in a "defaults" section, it can be disabled
4753 in a specific instance by prepending the "no" keyword before it.
4754
4755 See also : "option accept-invalid-http-request" and "show errors" on the
4756 stats socket.
4757
4758
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004759option allbackups
4760no option allbackups
4761 Use either all backup servers at a time or only the first one
4762 May be used in sections : defaults | frontend | listen | backend
4763 yes | no | yes | yes
4764 Arguments : none
4765
4766 By default, the first operational backup server gets all traffic when normal
4767 servers are all down. Sometimes, it may be preferred to use multiple backups
4768 at once, because one will not be enough. When "option allbackups" is enabled,
4769 the load balancing will be performed among all backup servers when all normal
4770 ones are unavailable. The same load balancing algorithm will be used and the
4771 servers' weights will be respected. Thus, there will not be any priority
4772 order between the backup servers anymore.
4773
4774 This option is mostly used with static server farms dedicated to return a
4775 "sorry" page when an application is completely offline.
4776
4777 If this option has been enabled in a "defaults" section, it can be disabled
4778 in a specific instance by prepending the "no" keyword before it.
4779
4780
4781option checkcache
4782no option checkcache
Godbach7056a352013-12-11 20:01:07 +08004783 Analyze all server responses and block responses with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004784 May be used in sections : defaults | frontend | listen | backend
4785 yes | no | yes | yes
4786 Arguments : none
4787
4788 Some high-level frameworks set application cookies everywhere and do not
4789 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004790 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004791 high risk of session crossing or stealing between users traversing the same
4792 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02004793 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004794
4795 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004796 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01004797 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004798 response to check if there's a risk of caching a cookie on a client-side
4799 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01004800 to the client are :
4801 - all those without "Set-Cookie" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004802 - all those with a return code other than 200, 203, 206, 300, 301, 410,
Willy Tarreau198a7442008-01-17 12:05:32 +01004803 provided that the server has not set a "Cache-control: public" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004804 - all those that come from a POST request, provided that the server has not
4805 set a 'Cache-Control: public' header ;
4806 - those with a 'Pragma: no-cache' header
4807 - those with a 'Cache-control: private' header
4808 - those with a 'Cache-control: no-store' header
4809 - those with a 'Cache-control: max-age=0' header
4810 - those with a 'Cache-control: s-maxage=0' header
4811 - those with a 'Cache-control: no-cache' header
4812 - those with a 'Cache-control: no-cache="set-cookie"' header
4813 - those with a 'Cache-control: no-cache="set-cookie,' header
4814 (allowing other fields after set-cookie)
4815
4816 If a response doesn't respect these requirements, then it will be blocked
Willy Tarreau198a7442008-01-17 12:05:32 +01004817 just as if it was from an "rspdeny" filter, with an "HTTP 502 bad gateway".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004818 The session state shows "PH--" meaning that the proxy blocked the response
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004819 during headers processing. Additionally, an alert will be sent in the logs so
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004820 that admins are informed that there's something to be fixed.
4821
4822 Due to the high impact on the application, the application should be tested
4823 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01004824 good practice to always activate it during tests, even if it is not used in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004825 production, as it will report potentially dangerous application behaviours.
4826
4827 If this option has been enabled in a "defaults" section, it can be disabled
4828 in a specific instance by prepending the "no" keyword before it.
4829
4830
4831option clitcpka
4832no option clitcpka
4833 Enable or disable the sending of TCP keepalive packets on the client side
4834 May be used in sections : defaults | frontend | listen | backend
4835 yes | yes | yes | no
4836 Arguments : none
4837
4838 When there is a firewall or any session-aware component between a client and
4839 a server, and when the protocol involves very long sessions with long idle
4840 periods (eg: remote desktops), there is a risk that one of the intermediate
4841 components decides to expire a session which has remained idle for too long.
4842
4843 Enabling socket-level TCP keep-alives makes the system regularly send packets
4844 to the other end of the connection, leaving it active. The delay between
4845 keep-alive probes is controlled by the system only and depends both on the
4846 operating system and its tuning parameters.
4847
4848 It is important to understand that keep-alive packets are neither emitted nor
4849 received at the application level. It is only the network stacks which sees
4850 them. For this reason, even if one side of the proxy already uses keep-alives
4851 to maintain its connection alive, those keep-alive packets will not be
4852 forwarded to the other side of the proxy.
4853
4854 Please note that this has nothing to do with HTTP keep-alive.
4855
4856 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
4857 client side of a connection, which should help when session expirations are
4858 noticed between HAProxy and a client.
4859
4860 If this option has been enabled in a "defaults" section, it can be disabled
4861 in a specific instance by prepending the "no" keyword before it.
4862
4863 See also : "option srvtcpka", "option tcpka"
4864
4865
Willy Tarreau0ba27502007-12-24 16:55:16 +01004866option contstats
4867 Enable continuous traffic statistics updates
4868 May be used in sections : defaults | frontend | listen | backend
4869 yes | yes | yes | no
4870 Arguments : none
4871
4872 By default, counters used for statistics calculation are incremented
4873 only when a session finishes. It works quite well when serving small
4874 objects, but with big ones (for example large images or archives) or
4875 with A/V streaming, a graph generated from haproxy counters looks like
4876 a hedgehog. With this option enabled counters get incremented continuously,
4877 during a whole session. Recounting touches a hotpath directly so
4878 it is not enabled by default, as it has small performance impact (~0.5%).
4879
4880
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02004881option dontlog-normal
4882no option dontlog-normal
4883 Enable or disable logging of normal, successful connections
4884 May be used in sections : defaults | frontend | listen | backend
4885 yes | yes | yes | no
4886 Arguments : none
4887
4888 There are large sites dealing with several thousand connections per second
4889 and for which logging is a major pain. Some of them are even forced to turn
4890 logs off and cannot debug production issues. Setting this option ensures that
4891 normal connections, those which experience no error, no timeout, no retry nor
4892 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
4893 mode, the response status code is checked and return codes 5xx will still be
4894 logged.
4895
4896 It is strongly discouraged to use this option as most of the time, the key to
4897 complex issues is in the normal logs which will not be logged here. If you
4898 need to separate logs, see the "log-separate-errors" option instead.
4899
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004900 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02004901 logging.
4902
4903
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004904option dontlognull
4905no option dontlognull
4906 Enable or disable logging of null connections
4907 May be used in sections : defaults | frontend | listen | backend
4908 yes | yes | yes | no
4909 Arguments : none
4910
4911 In certain environments, there are components which will regularly connect to
4912 various systems to ensure that they are still alive. It can be the case from
4913 another load balancer as well as from monitoring systems. By default, even a
4914 simple port probe or scan will produce a log. If those connections pollute
4915 the logs too much, it is possible to enable option "dontlognull" to indicate
4916 that a connection on which no data has been transferred will not be logged,
Willy Tarreau0f228a02015-05-01 15:37:53 +02004917 which typically corresponds to those probes. Note that errors will still be
4918 returned to the client and accounted for in the stats. If this is not what is
4919 desired, option http-ignore-probes can be used instead.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004920
4921 It is generally recommended not to use this option in uncontrolled
4922 environments (eg: internet), otherwise scans and other malicious activities
4923 would not be logged.
4924
4925 If this option has been enabled in a "defaults" section, it can be disabled
4926 in a specific instance by prepending the "no" keyword before it.
4927
Willy Tarreau0f228a02015-05-01 15:37:53 +02004928 See also : "log", "http-ignore-probes", "monitor-net", "monitor-uri", and
4929 section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004930
4931
4932option forceclose
4933no option forceclose
4934 Enable or disable active connection closing after response is transferred.
4935 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaua31e5df2009-12-30 01:10:35 +01004936 yes | yes | yes | yes
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004937 Arguments : none
4938
4939 Some HTTP servers do not necessarily close the connections when they receive
4940 the "Connection: close" set by "option httpclose", and if the client does not
4941 close either, then the connection remains open till the timeout expires. This
4942 causes high number of simultaneous connections on the servers and shows high
4943 global session times in the logs.
4944
4945 When this happens, it is possible to use "option forceclose". It will
Willy Tarreau82eeaf22009-12-29 12:09:05 +01004946 actively close the outgoing server channel as soon as the server has finished
Cyril Bonté653dcd62014-02-20 00:13:15 +01004947 to respond and release some resources earlier than with "option httpclose".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004948
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004949 This option may also be combined with "option http-pretend-keepalive", which
4950 will disable sending of the "Connection: close" header, but will still cause
4951 the connection to be closed once the whole response is received.
4952
Cyril Bonté653dcd62014-02-20 00:13:15 +01004953 This option disables and replaces any previous "option httpclose", "option
4954 http-server-close", "option http-keep-alive", or "option http-tunnel".
Willy Tarreau02bce8b2014-01-30 00:15:28 +01004955
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004956 If this option has been enabled in a "defaults" section, it can be disabled
4957 in a specific instance by prepending the "no" keyword before it.
4958
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004959 See also : "option httpclose" and "option http-pretend-keepalive"
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004960
4961
Willy Tarreau87cf5142011-08-19 22:57:24 +02004962option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01004963 Enable insertion of the X-Forwarded-For header to requests sent to servers
4964 May be used in sections : defaults | frontend | listen | backend
4965 yes | yes | yes | yes
4966 Arguments :
4967 <network> is an optional argument used to disable this option for sources
4968 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02004969 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01004970 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004971
4972 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
4973 their client address. This is sometimes annoying when the client's IP address
4974 is expected in server logs. To solve this problem, the well-known HTTP header
4975 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
4976 This header contains a value representing the client's IP address. Since this
4977 header is always appended at the end of the existing header list, the server
4978 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02004979 the server's manual to find how to enable use of this standard header. Note
4980 that only the last occurrence of the header must be used, since it is really
4981 possible that the client has already brought one.
4982
Willy Tarreaud72758d2010-01-12 10:42:19 +01004983 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02004984 the default "X-Forwarded-For". This can be useful where you might already
Willy Tarreaud72758d2010-01-12 10:42:19 +01004985 have a "X-Forwarded-For" header from a different application (eg: stunnel),
4986 and you need preserve it. Also if your backend server doesn't use the
Ross Westaf72a1d2008-08-03 10:51:45 +02004987 "X-Forwarded-For" header and requires different one (eg: Zeus Web Servers
4988 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01004989
4990 Sometimes, a same HAProxy instance may be shared between a direct client
4991 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
4992 used to decrypt HTTPS traffic). It is possible to disable the addition of the
4993 header for a known source address or network by adding the "except" keyword
4994 followed by the network address. In this case, any source IP matching the
4995 network will not cause an addition of this header. Most common uses are with
4996 private networks or 127.0.0.1.
4997
Willy Tarreau87cf5142011-08-19 22:57:24 +02004998 Alternatively, the keyword "if-none" states that the header will only be
4999 added if it is not present. This should only be used in perfectly trusted
5000 environment, as this might cause a security issue if headers reaching haproxy
5001 are under the control of the end-user.
5002
Willy Tarreauc27debf2008-01-06 08:57:02 +01005003 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02005004 least one of them uses it, the header will be added. Note that the backend's
5005 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02005006 both are defined. In the case of the "if-none" argument, if at least one of
5007 the frontend or the backend does not specify it, it wants the addition to be
5008 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005009
Ross Westaf72a1d2008-08-03 10:51:45 +02005010 Examples :
Willy Tarreauc27debf2008-01-06 08:57:02 +01005011 # Public HTTP address also used by stunnel on the same machine
5012 frontend www
5013 mode http
5014 option forwardfor except 127.0.0.1 # stunnel already adds the header
5015
Ross Westaf72a1d2008-08-03 10:51:45 +02005016 # Those servers want the IP Address in X-Client
5017 backend www
5018 mode http
5019 option forwardfor header X-Client
5020
Willy Tarreau87cf5142011-08-19 22:57:24 +02005021 See also : "option httpclose", "option http-server-close",
Willy Tarreau70dffda2014-01-30 03:07:23 +01005022 "option forceclose", "option http-keep-alive"
Willy Tarreauc27debf2008-01-06 08:57:02 +01005023
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005024
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005025option http-buffer-request
5026no option http-buffer-request
5027 Enable or disable waiting for whole HTTP request body before proceeding
5028 May be used in sections : defaults | frontend | listen | backend
5029 yes | yes | yes | yes
5030 Arguments : none
5031
5032 It is sometimes desirable to wait for the body of an HTTP request before
5033 taking a decision. This is what is being done by "balance url_param" for
5034 example. The first use case is to buffer requests from slow clients before
5035 connecting to the server. Another use case consists in taking the routing
5036 decision based on the request body's contents. This option placed in a
5037 frontend or backend forces the HTTP processing to wait until either the whole
5038 body is received, or the request buffer is full, or the first chunk is
5039 complete in case of chunked encoding. It can have undesired side effects with
5040 some applications abusing HTTP by expecting unbufferred transmissions between
5041 the frontend and the backend, so this should definitely not be used by
5042 default.
5043
5044 See also : "option http-no-delay"
5045
5046
Willy Tarreau0f228a02015-05-01 15:37:53 +02005047option http-ignore-probes
5048no option http-ignore-probes
5049 Enable or disable logging of null connections and request timeouts
5050 May be used in sections : defaults | frontend | listen | backend
5051 yes | yes | yes | no
5052 Arguments : none
5053
5054 Recently some browsers started to implement a "pre-connect" feature
5055 consisting in speculatively connecting to some recently visited web sites
5056 just in case the user would like to visit them. This results in many
5057 connections being established to web sites, which end up in 408 Request
5058 Timeout if the timeout strikes first, or 400 Bad Request when the browser
5059 decides to close them first. These ones pollute the log and feed the error
5060 counters. There was already "option dontlognull" but it's insufficient in
5061 this case. Instead, this option does the following things :
5062 - prevent any 400/408 message from being sent to the client if nothing
5063 was received over a connection before it was closed ;
5064 - prevent any log from being emitted in this situation ;
5065 - prevent any error counter from being incremented
5066
5067 That way the empty connection is silently ignored. Note that it is better
5068 not to use this unless it is clear that it is needed, because it will hide
5069 real problems. The most common reason for not receiving a request and seeing
5070 a 408 is due to an MTU inconsistency between the client and an intermediary
5071 element such as a VPN, which blocks too large packets. These issues are
5072 generally seen with POST requests as well as GET with large cookies. The logs
5073 are often the only way to detect them.
5074
5075 If this option has been enabled in a "defaults" section, it can be disabled
5076 in a specific instance by prepending the "no" keyword before it.
5077
5078 See also : "log", "dontlognull", "errorfile", and section 8 about logging.
5079
5080
Willy Tarreau16bfb022010-01-16 19:48:41 +01005081option http-keep-alive
5082no option http-keep-alive
5083 Enable or disable HTTP keep-alive from client to server
5084 May be used in sections : defaults | frontend | listen | backend
5085 yes | yes | yes | yes
5086 Arguments : none
5087
Willy Tarreau70dffda2014-01-30 03:07:23 +01005088 By default HAProxy operates in keep-alive mode with regards to persistent
5089 connections: for each connection it processes each request and response, and
5090 leaves the connection idle on both sides between the end of a response and the
5091 start of a new request. This mode may be changed by several options such as
5092 "option http-server-close", "option forceclose", "option httpclose" or
5093 "option http-tunnel". This option allows to set back the keep-alive mode,
5094 which can be useful when another mode was used in a defaults section.
5095
5096 Setting "option http-keep-alive" enables HTTP keep-alive mode on the client-
5097 and server- sides. This provides the lowest latency on the client side (slow
Willy Tarreau16bfb022010-01-16 19:48:41 +01005098 network) and the fastest session reuse on the server side at the expense
5099 of maintaining idle connections to the servers. In general, it is possible
5100 with this option to achieve approximately twice the request rate that the
5101 "http-server-close" option achieves on small objects. There are mainly two
5102 situations where this option may be useful :
5103
5104 - when the server is non-HTTP compliant and authenticates the connection
5105 instead of requests (eg: NTLM authentication)
5106
5107 - when the cost of establishing the connection to the server is significant
5108 compared to the cost of retrieving the associated object from the server.
5109
5110 This last case can happen when the server is a fast static server of cache.
5111 In this case, the server will need to be properly tuned to support high enough
5112 connection counts because connections will last until the client sends another
5113 request.
5114
5115 If the client request has to go to another backend or another server due to
5116 content switching or the load balancing algorithm, the idle connection will
Willy Tarreau9420b122013-12-15 18:58:25 +01005117 immediately be closed and a new one re-opened. Option "prefer-last-server" is
5118 available to try optimize server selection so that if the server currently
5119 attached to an idle connection is usable, it will be used.
Willy Tarreau16bfb022010-01-16 19:48:41 +01005120
5121 In general it is preferred to use "option http-server-close" with application
5122 servers, and some static servers might benefit from "option http-keep-alive".
5123
5124 At the moment, logs will not indicate whether requests came from the same
5125 session or not. The accept date reported in the logs corresponds to the end
5126 of the previous request, and the request time corresponds to the time spent
5127 waiting for a new request. The keep-alive request time is still bound to the
5128 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
5129 not set.
5130
Cyril Bonté653dcd62014-02-20 00:13:15 +01005131 This option disables and replaces any previous "option httpclose", "option
5132 http-server-close", "option forceclose" or "option http-tunnel". When backend
Willy Tarreau70dffda2014-01-30 03:07:23 +01005133 and frontend options differ, all of these 4 options have precedence over
Cyril Bonté653dcd62014-02-20 00:13:15 +01005134 "option http-keep-alive".
Willy Tarreau16bfb022010-01-16 19:48:41 +01005135
5136 See also : "option forceclose", "option http-server-close",
Willy Tarreau9420b122013-12-15 18:58:25 +01005137 "option prefer-last-server", "option http-pretend-keepalive",
5138 "option httpclose", and "1.1. The HTTP transaction model".
Willy Tarreau16bfb022010-01-16 19:48:41 +01005139
5140
Willy Tarreau96e31212011-05-30 18:10:30 +02005141option http-no-delay
5142no option http-no-delay
5143 Instruct the system to favor low interactive delays over performance in HTTP
5144 May be used in sections : defaults | frontend | listen | backend
5145 yes | yes | yes | yes
5146 Arguments : none
5147
5148 In HTTP, each payload is unidirectional and has no notion of interactivity.
5149 Any agent is expected to queue data somewhat for a reasonably low delay.
5150 There are some very rare server-to-server applications that abuse the HTTP
5151 protocol and expect the payload phase to be highly interactive, with many
5152 interleaved data chunks in both directions within a single request. This is
5153 absolutely not supported by the HTTP specification and will not work across
5154 most proxies or servers. When such applications attempt to do this through
5155 haproxy, it works but they will experience high delays due to the network
5156 optimizations which favor performance by instructing the system to wait for
5157 enough data to be available in order to only send full packets. Typical
5158 delays are around 200 ms per round trip. Note that this only happens with
5159 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
5160 affected.
5161
5162 When "option http-no-delay" is present in either the frontend or the backend
5163 used by a connection, all such optimizations will be disabled in order to
5164 make the exchanges as fast as possible. Of course this offers no guarantee on
5165 the functionality, as it may break at any other place. But if it works via
5166 HAProxy, it will work as fast as possible. This option should never be used
5167 by default, and should never be used at all unless such a buggy application
5168 is discovered. The impact of using this option is an increase of bandwidth
5169 usage and CPU usage, which may significantly lower performance in high
5170 latency environments.
5171
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005172 See also : "option http-buffer-request"
5173
Willy Tarreau96e31212011-05-30 18:10:30 +02005174
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005175option http-pretend-keepalive
5176no option http-pretend-keepalive
5177 Define whether haproxy will announce keepalive to the server or not
5178 May be used in sections : defaults | frontend | listen | backend
5179 yes | yes | yes | yes
5180 Arguments : none
5181
5182 When running with "option http-server-close" or "option forceclose", haproxy
5183 adds a "Connection: close" header to the request forwarded to the server.
5184 Unfortunately, when some servers see this header, they automatically refrain
5185 from using the chunked encoding for responses of unknown length, while this
5186 is totally unrelated. The immediate effect is that this prevents haproxy from
5187 maintaining the client connection alive. A second effect is that a client or
5188 a cache could receive an incomplete response without being aware of it, and
5189 consider the response complete.
5190
5191 By setting "option http-pretend-keepalive", haproxy will make the server
5192 believe it will keep the connection alive. The server will then not fall back
5193 to the abnormal undesired above. When haproxy gets the whole response, it
5194 will close the connection with the server just as it would do with the
5195 "forceclose" option. That way the client gets a normal response and the
5196 connection is correctly closed on the server side.
5197
5198 It is recommended not to enable this option by default, because most servers
5199 will more efficiently close the connection themselves after the last packet,
5200 and release its buffers slightly earlier. Also, the added packet on the
5201 network could slightly reduce the overall peak performance. However it is
5202 worth noting that when this option is enabled, haproxy will have slightly
5203 less work to do. So if haproxy is the bottleneck on the whole architecture,
5204 enabling this option might save a few CPU cycles.
5205
5206 This option may be set both in a frontend and in a backend. It is enabled if
5207 at least one of the frontend or backend holding a connection has it enabled.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005208 This option may be combined with "option httpclose", which will cause
Willy Tarreau22a95342010-09-29 14:31:41 +02005209 keepalive to be announced to the server and close to be announced to the
5210 client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005211
5212 If this option has been enabled in a "defaults" section, it can be disabled
5213 in a specific instance by prepending the "no" keyword before it.
5214
Willy Tarreau16bfb022010-01-16 19:48:41 +01005215 See also : "option forceclose", "option http-server-close", and
5216 "option http-keep-alive"
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005217
Willy Tarreauc27debf2008-01-06 08:57:02 +01005218
Willy Tarreaub608feb2010-01-02 22:47:18 +01005219option http-server-close
5220no option http-server-close
5221 Enable or disable HTTP connection closing on the server side
5222 May be used in sections : defaults | frontend | listen | backend
5223 yes | yes | yes | yes
5224 Arguments : none
5225
Willy Tarreau70dffda2014-01-30 03:07:23 +01005226 By default HAProxy operates in keep-alive mode with regards to persistent
5227 connections: for each connection it processes each request and response, and
5228 leaves the connection idle on both sides between the end of a response and
5229 the start of a new request. This mode may be changed by several options such
5230 as "option http-server-close", "option forceclose", "option httpclose" or
5231 "option http-tunnel". Setting "option http-server-close" enables HTTP
5232 connection-close mode on the server side while keeping the ability to support
5233 HTTP keep-alive and pipelining on the client side. This provides the lowest
5234 latency on the client side (slow network) and the fastest session reuse on
5235 the server side to save server resources, similarly to "option forceclose".
5236 It also permits non-keepalive capable servers to be served in keep-alive mode
5237 to the clients if they conform to the requirements of RFC2616. Please note
5238 that some servers do not always conform to those requirements when they see
5239 "Connection: close" in the request. The effect will be that keep-alive will
5240 never be used. A workaround consists in enabling "option
5241 http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01005242
5243 At the moment, logs will not indicate whether requests came from the same
5244 session or not. The accept date reported in the logs corresponds to the end
5245 of the previous request, and the request time corresponds to the time spent
5246 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01005247 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
5248 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01005249
5250 This option may be set both in a frontend and in a backend. It is enabled if
5251 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01005252 It disables and replaces any previous "option httpclose", "option forceclose",
5253 "option http-tunnel" or "option http-keep-alive". Please check section 4
Willy Tarreau70dffda2014-01-30 03:07:23 +01005254 ("Proxies") to see how this option combines with others when frontend and
5255 backend options differ.
Willy Tarreaub608feb2010-01-02 22:47:18 +01005256
5257 If this option has been enabled in a "defaults" section, it can be disabled
5258 in a specific instance by prepending the "no" keyword before it.
5259
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02005260 See also : "option forceclose", "option http-pretend-keepalive",
Willy Tarreau16bfb022010-01-16 19:48:41 +01005261 "option httpclose", "option http-keep-alive", and
5262 "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01005263
5264
Willy Tarreau02bce8b2014-01-30 00:15:28 +01005265option http-tunnel
5266no option http-tunnel
5267 Disable or enable HTTP connection processing after first transaction
5268 May be used in sections : defaults | frontend | listen | backend
5269 yes | yes | yes | yes
5270 Arguments : none
5271
Willy Tarreau70dffda2014-01-30 03:07:23 +01005272 By default HAProxy operates in keep-alive mode with regards to persistent
5273 connections: for each connection it processes each request and response, and
5274 leaves the connection idle on both sides between the end of a response and
5275 the start of a new request. This mode may be changed by several options such
5276 as "option http-server-close", "option forceclose", "option httpclose" or
5277 "option http-tunnel".
5278
5279 Option "http-tunnel" disables any HTTP processing past the first request and
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005280 the first response. This is the mode which was used by default in versions
Willy Tarreau70dffda2014-01-30 03:07:23 +01005281 1.0 to 1.5-dev21. It is the mode with the lowest processing overhead, which
5282 is normally not needed anymore unless in very specific cases such as when
5283 using an in-house protocol that looks like HTTP but is not compatible, or
5284 just to log one request per client in order to reduce log size. Note that
5285 everything which works at the HTTP level, including header parsing/addition,
5286 cookie processing or content switching will only work for the first request
5287 and will be ignored after the first response.
Willy Tarreau02bce8b2014-01-30 00:15:28 +01005288
5289 If this option has been enabled in a "defaults" section, it can be disabled
5290 in a specific instance by prepending the "no" keyword before it.
5291
5292 See also : "option forceclose", "option http-server-close",
5293 "option httpclose", "option http-keep-alive", and
5294 "1.1. The HTTP transaction model".
5295
5296
Willy Tarreau88d349d2010-01-25 12:15:43 +01005297option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01005298no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01005299 Make use of non-standard Proxy-Connection header instead of Connection
5300 May be used in sections : defaults | frontend | listen | backend
5301 yes | yes | yes | no
5302 Arguments : none
5303
5304 While RFC2616 explicitly states that HTTP/1.1 agents must use the
5305 Connection header to indicate their wish of persistent or non-persistent
5306 connections, both browsers and proxies ignore this header for proxied
5307 connections and make use of the undocumented, non-standard Proxy-Connection
5308 header instead. The issue begins when trying to put a load balancer between
5309 browsers and such proxies, because there will be a difference between what
5310 haproxy understands and what the client and the proxy agree on.
5311
5312 By setting this option in a frontend, haproxy can automatically switch to use
5313 that non-standard header if it sees proxied requests. A proxied request is
5314 defined here as one where the URI begins with neither a '/' nor a '*'. The
5315 choice of header only affects requests passing through proxies making use of
5316 one of the "httpclose", "forceclose" and "http-server-close" options. Note
5317 that this option can only be specified in a frontend and will affect the
5318 request along its whole life.
5319
Willy Tarreau844a7e72010-01-31 21:46:18 +01005320 Also, when this option is set, a request which requires authentication will
5321 automatically switch to use proxy authentication headers if it is itself a
5322 proxied request. That makes it possible to check or enforce authentication in
5323 front of an existing proxy.
5324
Willy Tarreau88d349d2010-01-25 12:15:43 +01005325 This option should normally never be used, except in front of a proxy.
5326
5327 See also : "option httpclose", "option forceclose" and "option
5328 http-server-close".
5329
5330
Willy Tarreaud63335a2010-02-26 12:56:52 +01005331option httpchk
5332option httpchk <uri>
5333option httpchk <method> <uri>
5334option httpchk <method> <uri> <version>
5335 Enable HTTP protocol to check on the servers health
5336 May be used in sections : defaults | frontend | listen | backend
5337 yes | no | yes | yes
5338 Arguments :
5339 <method> is the optional HTTP method used with the requests. When not set,
5340 the "OPTIONS" method is used, as it generally requires low server
5341 processing and is easy to filter out from the logs. Any method
5342 may be used, though it is not recommended to invent non-standard
5343 ones.
5344
5345 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
5346 which is accessible by default on almost any server, but may be
5347 changed to any other URI. Query strings are permitted.
5348
5349 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
5350 but some servers might behave incorrectly in HTTP 1.0, so turning
5351 it to HTTP/1.1 may sometimes help. Note that the Host field is
5352 mandatory in HTTP/1.1, and as a trick, it is possible to pass it
5353 after "\r\n" following the version string.
5354
5355 By default, server health checks only consist in trying to establish a TCP
5356 connection. When "option httpchk" is specified, a complete HTTP request is
5357 sent once the TCP connection is established, and responses 2xx and 3xx are
5358 considered valid, while all other ones indicate a server failure, including
5359 the lack of any response.
5360
5361 The port and interval are specified in the server configuration.
5362
5363 This option does not necessarily require an HTTP backend, it also works with
5364 plain TCP backends. This is particularly useful to check simple scripts bound
5365 to some dedicated ports using the inetd daemon.
5366
5367 Examples :
5368 # Relay HTTPS traffic to Apache instance and check service availability
5369 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
5370 backend https_relay
5371 mode tcp
5372 option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www
5373 server apache1 192.168.1.1:443 check port 80
5374
Simon Hormanafc47ee2013-11-25 10:46:35 +09005375 See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",
5376 "option pgsql-check", "http-check" and the "check", "port" and
5377 "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01005378
5379
Willy Tarreauc27debf2008-01-06 08:57:02 +01005380option httpclose
5381no option httpclose
5382 Enable or disable passive HTTP connection closing
5383 May be used in sections : defaults | frontend | listen | backend
5384 yes | yes | yes | yes
5385 Arguments : none
5386
Willy Tarreau70dffda2014-01-30 03:07:23 +01005387 By default HAProxy operates in keep-alive mode with regards to persistent
5388 connections: for each connection it processes each request and response, and
5389 leaves the connection idle on both sides between the end of a response and
5390 the start of a new request. This mode may be changed by several options such
Cyril Bonté653dcd62014-02-20 00:13:15 +01005391 as "option http-server-close", "option forceclose", "option httpclose" or
Willy Tarreau70dffda2014-01-30 03:07:23 +01005392 "option http-tunnel".
5393
5394 If "option httpclose" is set, HAProxy will work in HTTP tunnel mode and check
5395 if a "Connection: close" header is already set in each direction, and will
5396 add one if missing. Each end should react to this by actively closing the TCP
5397 connection after each transfer, thus resulting in a switch to the HTTP close
5398 mode. Any "Connection" header different from "close" will also be removed.
5399 Note that this option is deprecated since what it does is very cheap but not
5400 reliable. Using "option http-server-close" or "option forceclose" is strongly
5401 recommended instead.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005402
5403 It seldom happens that some servers incorrectly ignore this header and do not
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005404 close the connection even though they reply "Connection: close". For this
Willy Tarreau0dfdf192010-01-05 11:33:11 +01005405 reason, they are not compatible with older HTTP 1.0 browsers. If this happens
5406 it is possible to use the "option forceclose" which actively closes the
5407 request connection once the server responds. Option "forceclose" also
5408 releases the server connection earlier because it does not have to wait for
5409 the client to acknowledge it.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005410
5411 This option may be set both in a frontend and in a backend. It is enabled if
5412 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01005413 It disables and replaces any previous "option http-server-close",
5414 "option forceclose", "option http-keep-alive" or "option http-tunnel". Please
Willy Tarreau70dffda2014-01-30 03:07:23 +01005415 check section 4 ("Proxies") to see how this option combines with others when
5416 frontend and backend options differ.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005417
5418 If this option has been enabled in a "defaults" section, it can be disabled
5419 in a specific instance by prepending the "no" keyword before it.
5420
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02005421 See also : "option forceclose", "option http-server-close" and
5422 "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01005423
5424
Emeric Brun3a058f32009-06-30 18:26:00 +02005425option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01005426 Enable logging of HTTP request, session state and timers
5427 May be used in sections : defaults | frontend | listen | backend
5428 yes | yes | yes | yes
Emeric Brun3a058f32009-06-30 18:26:00 +02005429 Arguments :
5430 clf if the "clf" argument is added, then the output format will be
5431 the CLF format instead of HAProxy's default HTTP format. You can
5432 use this when you need to feed HAProxy's logs through a specific
5433 log analyser which only support the CLF format and which is not
5434 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005435
5436 By default, the log output format is very poor, as it only contains the
5437 source and destination addresses, and the instance name. By specifying
5438 "option httplog", each log line turns into a much richer format including,
5439 but not limited to, the HTTP request, the connection timers, the session
5440 status, the connections numbers, the captured headers and cookies, the
5441 frontend, backend and server name, and of course the source address and
5442 ports.
5443
5444 This option may be set either in the frontend or the backend.
5445
PiBa-NLbd556bf2014-12-11 21:31:54 +01005446 Specifying only "option httplog" will automatically clear the 'clf' mode
5447 if it was set by default.
Emeric Brun3a058f32009-06-30 18:26:00 +02005448
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005449 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005450
Willy Tarreau55165fe2009-05-10 12:02:55 +02005451
5452option http_proxy
5453no option http_proxy
5454 Enable or disable plain HTTP proxy mode
5455 May be used in sections : defaults | frontend | listen | backend
5456 yes | yes | yes | yes
5457 Arguments : none
5458
5459 It sometimes happens that people need a pure HTTP proxy which understands
5460 basic proxy requests without caching nor any fancy feature. In this case,
5461 it may be worth setting up an HAProxy instance with the "option http_proxy"
5462 set. In this mode, no server is declared, and the connection is forwarded to
5463 the IP address and port found in the URL after the "http://" scheme.
5464
5465 No host address resolution is performed, so this only works when pure IP
5466 addresses are passed. Since this option's usage perimeter is rather limited,
5467 it will probably be used only by experts who know they need exactly it. Last,
5468 if the clients are susceptible of sending keep-alive requests, it will be
Cyril Bonté2409e682010-12-14 22:47:51 +01005469 needed to add "option httpclose" to ensure that all requests will correctly
Willy Tarreau55165fe2009-05-10 12:02:55 +02005470 be analyzed.
5471
5472 If this option has been enabled in a "defaults" section, it can be disabled
5473 in a specific instance by prepending the "no" keyword before it.
5474
5475 Example :
5476 # this backend understands HTTP proxy requests and forwards them directly.
5477 backend direct_forward
5478 option httpclose
5479 option http_proxy
5480
5481 See also : "option httpclose"
5482
Willy Tarreau211ad242009-10-03 21:45:07 +02005483
Jamie Gloudon801a0a32012-08-25 00:18:33 -04005484option independent-streams
5485no option independent-streams
5486 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02005487 May be used in sections : defaults | frontend | listen | backend
5488 yes | yes | yes | yes
5489 Arguments : none
5490
5491 By default, when data is sent over a socket, both the write timeout and the
5492 read timeout for that socket are refreshed, because we consider that there is
5493 activity on that socket, and we have no other means of guessing if we should
5494 receive data or not.
5495
5496 While this default behaviour is desirable for almost all applications, there
5497 exists a situation where it is desirable to disable it, and only refresh the
5498 read timeout if there are incoming data. This happens on sessions with large
5499 timeouts and low amounts of exchanged data such as telnet session. If the
5500 server suddenly disappears, the output data accumulates in the system's
5501 socket buffers, both timeouts are correctly refreshed, and there is no way
5502 to know the server does not receive them, so we don't timeout. However, when
5503 the underlying protocol always echoes sent data, it would be enough by itself
5504 to detect the issue using the read timeout. Note that this problem does not
5505 happen with more verbose protocols because data won't accumulate long in the
5506 socket buffers.
5507
5508 When this option is set on the frontend, it will disable read timeout updates
5509 on data sent to the client. There probably is little use of this case. When
5510 the option is set on the backend, it will disable read timeout updates on
5511 data sent to the server. Doing so will typically break large HTTP posts from
5512 slow lines, so use it with caution.
5513
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005514 Note: older versions used to call this setting "option independent-streams"
Jamie Gloudon801a0a32012-08-25 00:18:33 -04005515 with a spelling mistake. This spelling is still supported but
5516 deprecated.
5517
Willy Tarreauce887fd2012-05-12 12:50:00 +02005518 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02005519
5520
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02005521option ldap-check
5522 Use LDAPv3 health checks for server testing
5523 May be used in sections : defaults | frontend | listen | backend
5524 yes | no | yes | yes
5525 Arguments : none
5526
5527 It is possible to test that the server correctly talks LDAPv3 instead of just
5528 testing that it accepts the TCP connection. When this option is set, an
5529 LDAPv3 anonymous simple bind message is sent to the server, and the response
5530 is analyzed to find an LDAPv3 bind response message.
5531
5532 The server is considered valid only when the LDAP response contains success
5533 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
5534
5535 Logging of bind requests is server dependent see your documentation how to
5536 configure it.
5537
5538 Example :
5539 option ldap-check
5540
5541 See also : "option httpchk"
5542
5543
Simon Horman98637e52014-06-20 12:30:16 +09005544option external-check
5545 Use external processes for server health checks
5546 May be used in sections : defaults | frontend | listen | backend
5547 yes | no | yes | yes
5548
5549 It is possible to test the health of a server using an external command.
5550 This is achieved by running the executable set using "external-check
5551 command".
5552
5553 Requires the "external-check" global to be set.
5554
5555 See also : "external-check", "external-check command", "external-check path"
5556
5557
Willy Tarreau211ad242009-10-03 21:45:07 +02005558option log-health-checks
5559no option log-health-checks
Willy Tarreaubef1b322014-05-13 21:01:39 +02005560 Enable or disable logging of health checks status updates
Willy Tarreau211ad242009-10-03 21:45:07 +02005561 May be used in sections : defaults | frontend | listen | backend
5562 yes | no | yes | yes
5563 Arguments : none
5564
Willy Tarreaubef1b322014-05-13 21:01:39 +02005565 By default, failed health check are logged if server is UP and successful
5566 health checks are logged if server is DOWN, so the amount of additional
5567 information is limited.
Willy Tarreau211ad242009-10-03 21:45:07 +02005568
Willy Tarreaubef1b322014-05-13 21:01:39 +02005569 When this option is enabled, any change of the health check status or to
5570 the server's health will be logged, so that it becomes possible to know
5571 that a server was failing occasional checks before crashing, or exactly when
5572 it failed to respond a valid HTTP status, then when the port started to
5573 reject connections, then when the server stopped responding at all.
5574
5575 Note that status changes not caused by health checks (eg: enable/disable on
5576 the CLI) are intentionally not logged by this option.
Willy Tarreau211ad242009-10-03 21:45:07 +02005577
Willy Tarreaubef1b322014-05-13 21:01:39 +02005578 See also: "option httpchk", "option ldap-check", "option mysql-check",
5579 "option pgsql-check", "option redis-check", "option smtpchk",
5580 "option tcp-check", "log" and section 8 about logging.
Willy Tarreau211ad242009-10-03 21:45:07 +02005581
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005582
5583option log-separate-errors
5584no option log-separate-errors
5585 Change log level for non-completely successful connections
5586 May be used in sections : defaults | frontend | listen | backend
5587 yes | yes | yes | no
5588 Arguments : none
5589
5590 Sometimes looking for errors in logs is not easy. This option makes haproxy
5591 raise the level of logs containing potentially interesting information such
5592 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
5593 level changes from "info" to "err". This makes it possible to log them
5594 separately to a different file with most syslog daemons. Be careful not to
5595 remove them from the original file, otherwise you would lose ordering which
5596 provides very important information.
5597
5598 Using this option, large sites dealing with several thousand connections per
5599 second may log normal traffic to a rotating buffer and only archive smaller
5600 error logs.
5601
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005602 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005603 logging.
5604
Willy Tarreauc27debf2008-01-06 08:57:02 +01005605
5606option logasap
5607no option logasap
5608 Enable or disable early logging of HTTP requests
5609 May be used in sections : defaults | frontend | listen | backend
5610 yes | yes | yes | no
5611 Arguments : none
5612
5613 By default, HTTP requests are logged upon termination so that the total
5614 transfer time and the number of bytes appear in the logs. When large objects
5615 are being transferred, it may take a while before the request appears in the
5616 logs. Using "option logasap", the request gets logged as soon as the server
5617 sends the complete headers. The only missing information in the logs will be
5618 the total number of bytes which will indicate everything except the amount
5619 of data transferred, and the total time which will not take the transfer
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01005620 time into account. In such a situation, it's a good practice to capture the
Willy Tarreauc27debf2008-01-06 08:57:02 +01005621 "Content-Length" response header so that the logs at least indicate how many
5622 bytes are expected to be transferred.
5623
Willy Tarreaucc6c8912009-02-22 10:53:55 +01005624 Examples :
5625 listen http_proxy 0.0.0.0:80
5626 mode http
5627 option httplog
5628 option logasap
5629 log 192.168.2.200 local3
5630
5631 >>> Feb 6 12:14:14 localhost \
5632 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
5633 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
5634 "GET /image.iso HTTP/1.0"
5635
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005636 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01005637 logging.
5638
5639
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02005640option mysql-check [ user <username> [ post-41 ] ]
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005641 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005642 May be used in sections : defaults | frontend | listen | backend
5643 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005644 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005645 <username> This is the username which will be used when connecting to MySQL
5646 server.
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02005647 post-41 Send post v4.1 client compatible checks
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005648
5649 If you specify a username, the check consists of sending two MySQL packet,
5650 one Client Authentication packet, and one QUIT packet, to correctly close
5651 MySQL session. We then parse the MySQL Handshake Initialisation packet and/or
5652 Error packet. It is a basic but useful test which does not produce error nor
5653 aborted connect on the server. However, it requires adding an authorization
5654 in the MySQL table, like this :
5655
5656 USE mysql;
5657 INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');
5658 FLUSH PRIVILEGES;
5659
5660 If you don't specify a username (it is deprecated and not recommended), the
5661 check only consists in parsing the Mysql Handshake Initialisation packet or
5662 Error packet, we don't send anything in this mode. It was reported that it
5663 can generate lockout if check is too frequent and/or if there is not enough
5664 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
5665 value as if a connection is established successfully within fewer than MySQL
5666 "max_connect_errors" attempts after a previous connection was interrupted,
5667 the error count for the host is cleared to zero. If HAProxy's server get
5668 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
5669
5670 Remember that this does not check database presence nor database consistency.
5671 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005672
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02005673 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005674
5675 Most often, an incoming MySQL server needs to see the client's IP address for
5676 various purposes, including IP privilege matching and connection logging.
5677 When possible, it is often wise to masquerade the client's IP address when
5678 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02005679 which requires the transparent proxy feature to be compiled in, and the MySQL
5680 server to route the client via the machine hosting haproxy.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005681
5682 See also: "option httpchk"
5683
5684
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005685option nolinger
5686no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005687 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005688 May be used in sections: defaults | frontend | listen | backend
5689 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005690 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005691
5692 When clients or servers abort connections in a dirty way (eg: they are
5693 physically disconnected), the session timeouts triggers and the session is
5694 closed. But it will remain in FIN_WAIT1 state for some time in the system,
5695 using some resources and possibly limiting the ability to establish newer
5696 connections.
5697
5698 When this happens, it is possible to activate "option nolinger" which forces
5699 the system to immediately remove any socket's pending data on close. Thus,
5700 the session is instantly purged from the system's tables. This usually has
5701 side effects such as increased number of TCP resets due to old retransmits
5702 getting immediately rejected. Some firewalls may sometimes complain about
5703 this too.
5704
5705 For this reason, it is not recommended to use this option when not absolutely
5706 needed. You know that you need it when you have thousands of FIN_WAIT1
5707 sessions on your system (TIME_WAIT ones do not count).
5708
5709 This option may be used both on frontends and backends, depending on the side
5710 where it is required. Use it on the frontend for clients, and on the backend
5711 for servers.
5712
5713 If this option has been enabled in a "defaults" section, it can be disabled
5714 in a specific instance by prepending the "no" keyword before it.
5715
5716
Willy Tarreau55165fe2009-05-10 12:02:55 +02005717option originalto [ except <network> ] [ header <name> ]
5718 Enable insertion of the X-Original-To header to requests sent to servers
5719 May be used in sections : defaults | frontend | listen | backend
5720 yes | yes | yes | yes
5721 Arguments :
5722 <network> is an optional argument used to disable this option for sources
5723 matching <network>
5724 <name> an optional argument to specify a different "X-Original-To"
5725 header name.
5726
5727 Since HAProxy can work in transparent mode, every request from a client can
5728 be redirected to the proxy and HAProxy itself can proxy every request to a
5729 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
5730 be lost. This is annoying when you want access rules based on destination ip
5731 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
5732 added by HAProxy to all requests sent to the server. This header contains a
5733 value representing the original destination IP address. Since this must be
5734 configured to always use the last occurrence of this header only. Note that
5735 only the last occurrence of the header must be used, since it is really
5736 possible that the client has already brought one.
5737
5738 The keyword "header" may be used to supply a different header name to replace
5739 the default "X-Original-To". This can be useful where you might already
5740 have a "X-Original-To" header from a different application, and you need
5741 preserve it. Also if your backend server doesn't use the "X-Original-To"
5742 header and requires different one.
5743
5744 Sometimes, a same HAProxy instance may be shared between a direct client
5745 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
5746 used to decrypt HTTPS traffic). It is possible to disable the addition of the
5747 header for a known source address or network by adding the "except" keyword
5748 followed by the network address. In this case, any source IP matching the
5749 network will not cause an addition of this header. Most common uses are with
5750 private networks or 127.0.0.1.
5751
5752 This option may be specified either in the frontend or in the backend. If at
5753 least one of them uses it, the header will be added. Note that the backend's
5754 setting of the header subargument takes precedence over the frontend's if
5755 both are defined.
5756
Willy Tarreau55165fe2009-05-10 12:02:55 +02005757 Examples :
5758 # Original Destination address
5759 frontend www
5760 mode http
5761 option originalto except 127.0.0.1
5762
5763 # Those servers want the IP Address in X-Client-Dst
5764 backend www
5765 mode http
5766 option originalto header X-Client-Dst
5767
Willy Tarreau87cf5142011-08-19 22:57:24 +02005768 See also : "option httpclose", "option http-server-close",
5769 "option forceclose"
Willy Tarreau55165fe2009-05-10 12:02:55 +02005770
5771
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005772option persist
5773no option persist
5774 Enable or disable forced persistence on down servers
5775 May be used in sections: defaults | frontend | listen | backend
5776 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005777 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005778
5779 When an HTTP request reaches a backend with a cookie which references a dead
5780 server, by default it is redispatched to another server. It is possible to
5781 force the request to be sent to the dead server first using "option persist"
5782 if absolutely needed. A common use case is when servers are under extreme
5783 load and spend their time flapping. In this case, the users would still be
5784 directed to the server they opened the session on, in the hope they would be
5785 correctly served. It is recommended to use "option redispatch" in conjunction
5786 with this option so that in the event it would not be possible to connect to
5787 the server at all (server definitely dead), the client would finally be
5788 redirected to another valid server.
5789
5790 If this option has been enabled in a "defaults" section, it can be disabled
5791 in a specific instance by prepending the "no" keyword before it.
5792
Willy Tarreau4de91492010-01-22 19:10:05 +01005793 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005794
5795
Willy Tarreau0c122822013-12-15 18:49:01 +01005796option pgsql-check [ user <username> ]
5797 Use PostgreSQL health checks for server testing
5798 May be used in sections : defaults | frontend | listen | backend
5799 yes | no | yes | yes
5800 Arguments :
5801 <username> This is the username which will be used when connecting to
5802 PostgreSQL server.
5803
5804 The check sends a PostgreSQL StartupMessage and waits for either
5805 Authentication request or ErrorResponse message. It is a basic but useful
5806 test which does not produce error nor aborted connect on the server.
5807 This check is identical with the "mysql-check".
5808
5809 See also: "option httpchk"
5810
5811
Willy Tarreau9420b122013-12-15 18:58:25 +01005812option prefer-last-server
5813no option prefer-last-server
5814 Allow multiple load balanced requests to remain on the same server
5815 May be used in sections: defaults | frontend | listen | backend
5816 yes | no | yes | yes
5817 Arguments : none
5818
5819 When the load balancing algorithm in use is not deterministic, and a previous
5820 request was sent to a server to which haproxy still holds a connection, it is
5821 sometimes desirable that subsequent requests on a same session go to the same
5822 server as much as possible. Note that this is different from persistence, as
5823 we only indicate a preference which haproxy tries to apply without any form
5824 of warranty. The real use is for keep-alive connections sent to servers. When
5825 this option is used, haproxy will try to reuse the same connection that is
5826 attached to the server instead of rebalancing to another server, causing a
5827 close of the connection. This can make sense for static file servers. It does
Willy Tarreau068621e2013-12-23 15:11:25 +01005828 not make much sense to use this in combination with hashing algorithms. Note,
5829 haproxy already automatically tries to stick to a server which sends a 401 or
5830 to a proxy which sends a 407 (authentication required). This is mandatory for
5831 use with the broken NTLM authentication challenge, and significantly helps in
5832 troubleshooting some faulty applications. Option prefer-last-server might be
5833 desirable in these environments as well, to avoid redistributing the traffic
5834 after every other response.
Willy Tarreau9420b122013-12-15 18:58:25 +01005835
5836 If this option has been enabled in a "defaults" section, it can be disabled
5837 in a specific instance by prepending the "no" keyword before it.
5838
5839 See also: "option http-keep-alive"
5840
5841
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005842option redispatch
Joseph Lynch726ab712015-05-11 23:25:34 -07005843option redispatch <interval>
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005844no option redispatch
5845 Enable or disable session redistribution in case of connection failure
5846 May be used in sections: defaults | frontend | listen | backend
5847 yes | no | yes | yes
Joseph Lynch726ab712015-05-11 23:25:34 -07005848 Arguments :
5849 <interval> The optional integer value that controls how often redispatches
5850 occur when retrying connections. Positive value P indicates a
5851 redispatch is desired on every Pth retry, and negative value
5852 N indicate a redispath is desired on the Nth retry prior to the
5853 last retry. For example, the default of -1 preserves the
5854 historical behaviour of redispatching on the last retry, a
5855 positive value of 1 would indicate a redispatch on every retry,
5856 and a positive value of 3 would indicate a redispatch on every
5857 third retry. You can disable redispatches with a value of 0.
5858
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005859
5860 In HTTP mode, if a server designated by a cookie is down, clients may
5861 definitely stick to it because they cannot flush the cookie, so they will not
5862 be able to access the service anymore.
5863
5864 Specifying "option redispatch" will allow the proxy to break their
5865 persistence and redistribute them to a working server.
5866
Joseph Lynch726ab712015-05-11 23:25:34 -07005867 It also allows to retry connections to another server in case of multiple
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005868 connection failures. Of course, it requires having "retries" set to a nonzero
5869 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01005870
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005871 This form is the preferred form, which replaces both the "redispatch" and
5872 "redisp" keywords.
5873
5874 If this option has been enabled in a "defaults" section, it can be disabled
5875 in a specific instance by prepending the "no" keyword before it.
5876
Willy Tarreau4de91492010-01-22 19:10:05 +01005877 See also : "redispatch", "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005878
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005879
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02005880option redis-check
5881 Use redis health checks for server testing
5882 May be used in sections : defaults | frontend | listen | backend
5883 yes | no | yes | yes
5884 Arguments : none
5885
5886 It is possible to test that the server correctly talks REDIS protocol instead
5887 of just testing that it accepts the TCP connection. When this option is set,
5888 a PING redis command is sent to the server, and the response is analyzed to
5889 find the "+PONG" response message.
5890
5891 Example :
5892 option redis-check
5893
5894 See also : "option httpchk"
5895
5896
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005897option smtpchk
5898option smtpchk <hello> <domain>
5899 Use SMTP health checks for server testing
5900 May be used in sections : defaults | frontend | listen | backend
5901 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01005902 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005903 <hello> is an optional argument. It is the "hello" command to use. It can
5904 be either "HELO" (for SMTP) or "EHLO" (for ESTMP). All other
5905 values will be turned into the default command ("HELO").
5906
5907 <domain> is the domain name to present to the server. It may only be
5908 specified (and is mandatory) if the hello command has been
5909 specified. By default, "localhost" is used.
5910
5911 When "option smtpchk" is set, the health checks will consist in TCP
5912 connections followed by an SMTP command. By default, this command is
5913 "HELO localhost". The server's return code is analyzed and only return codes
5914 starting with a "2" will be considered as valid. All other responses,
5915 including a lack of response will constitute an error and will indicate a
5916 dead server.
5917
5918 This test is meant to be used with SMTP servers or relays. Depending on the
5919 request, it is possible that some servers do not log each connection attempt,
5920 so you may want to experiment to improve the behaviour. Using telnet on port
5921 25 is often easier than adjusting the configuration.
5922
5923 Most often, an incoming SMTP server needs to see the client's IP address for
5924 various purposes, including spam filtering, anti-spoofing and logging. When
5925 possible, it is often wise to masquerade the client's IP address when
5926 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02005927 which requires the transparent proxy feature to be compiled in.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005928
5929 Example :
5930 option smtpchk HELO mydomain.org
5931
5932 See also : "option httpchk", "source"
5933
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005934
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02005935option socket-stats
5936no option socket-stats
5937
5938 Enable or disable collecting & providing separate statistics for each socket.
5939 May be used in sections : defaults | frontend | listen | backend
5940 yes | yes | yes | no
5941
5942 Arguments : none
5943
5944
Willy Tarreauff4f82d2009-02-06 11:28:13 +01005945option splice-auto
5946no option splice-auto
5947 Enable or disable automatic kernel acceleration on sockets in both directions
5948 May be used in sections : defaults | frontend | listen | backend
5949 yes | yes | yes | yes
5950 Arguments : none
5951
5952 When this option is enabled either on a frontend or on a backend, haproxy
5953 will automatically evaluate the opportunity to use kernel tcp splicing to
5954 forward data between the client and the server, in either direction. Haproxy
5955 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005956 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01005957 are not much aggressive in order to limit excessive use of splicing. This
5958 option requires splicing to be enabled at compile time, and may be globally
5959 disabled with the global option "nosplice". Since splice uses pipes, using it
5960 requires that there are enough spare pipes.
5961
5962 Important note: kernel-based TCP splicing is a Linux-specific feature which
5963 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
5964 transfer data between sockets without copying these data to user-space, thus
5965 providing noticeable performance gains and CPU cycles savings. Since many
5966 early implementations are buggy, corrupt data and/or are inefficient, this
5967 feature is not enabled by default, and it should be used with extreme care.
5968 While it is not possible to detect the correctness of an implementation,
5969 2.6.29 is the first version offering a properly working implementation. In
5970 case of doubt, splicing may be globally disabled using the global "nosplice"
5971 keyword.
5972
5973 Example :
5974 option splice-auto
5975
5976 If this option has been enabled in a "defaults" section, it can be disabled
5977 in a specific instance by prepending the "no" keyword before it.
5978
5979 See also : "option splice-request", "option splice-response", and global
5980 options "nosplice" and "maxpipes"
5981
5982
5983option splice-request
5984no option splice-request
5985 Enable or disable automatic kernel acceleration on sockets for requests
5986 May be used in sections : defaults | frontend | listen | backend
5987 yes | yes | yes | yes
5988 Arguments : none
5989
5990 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005991 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01005992 the client to the server. It might still use the recv/send scheme if there
5993 are no spare pipes left. This option requires splicing to be enabled at
5994 compile time, and may be globally disabled with the global option "nosplice".
5995 Since splice uses pipes, using it requires that there are enough spare pipes.
5996
5997 Important note: see "option splice-auto" for usage limitations.
5998
5999 Example :
6000 option splice-request
6001
6002 If this option has been enabled in a "defaults" section, it can be disabled
6003 in a specific instance by prepending the "no" keyword before it.
6004
6005 See also : "option splice-auto", "option splice-response", and global options
6006 "nosplice" and "maxpipes"
6007
6008
6009option splice-response
6010no option splice-response
6011 Enable or disable automatic kernel acceleration on sockets for responses
6012 May be used in sections : defaults | frontend | listen | backend
6013 yes | yes | yes | yes
6014 Arguments : none
6015
6016 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04006017 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006018 the server to the client. It might still use the recv/send scheme if there
6019 are no spare pipes left. This option requires splicing to be enabled at
6020 compile time, and may be globally disabled with the global option "nosplice".
6021 Since splice uses pipes, using it requires that there are enough spare pipes.
6022
6023 Important note: see "option splice-auto" for usage limitations.
6024
6025 Example :
6026 option splice-response
6027
6028 If this option has been enabled in a "defaults" section, it can be disabled
6029 in a specific instance by prepending the "no" keyword before it.
6030
6031 See also : "option splice-auto", "option splice-request", and global options
6032 "nosplice" and "maxpipes"
6033
6034
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006035option srvtcpka
6036no option srvtcpka
6037 Enable or disable the sending of TCP keepalive packets on the server side
6038 May be used in sections : defaults | frontend | listen | backend
6039 yes | no | yes | yes
6040 Arguments : none
6041
6042 When there is a firewall or any session-aware component between a client and
6043 a server, and when the protocol involves very long sessions with long idle
6044 periods (eg: remote desktops), there is a risk that one of the intermediate
6045 components decides to expire a session which has remained idle for too long.
6046
6047 Enabling socket-level TCP keep-alives makes the system regularly send packets
6048 to the other end of the connection, leaving it active. The delay between
6049 keep-alive probes is controlled by the system only and depends both on the
6050 operating system and its tuning parameters.
6051
6052 It is important to understand that keep-alive packets are neither emitted nor
6053 received at the application level. It is only the network stacks which sees
6054 them. For this reason, even if one side of the proxy already uses keep-alives
6055 to maintain its connection alive, those keep-alive packets will not be
6056 forwarded to the other side of the proxy.
6057
6058 Please note that this has nothing to do with HTTP keep-alive.
6059
6060 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
6061 server side of a connection, which should help when session expirations are
6062 noticed between HAProxy and a server.
6063
6064 If this option has been enabled in a "defaults" section, it can be disabled
6065 in a specific instance by prepending the "no" keyword before it.
6066
6067 See also : "option clitcpka", "option tcpka"
6068
6069
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006070option ssl-hello-chk
6071 Use SSLv3 client hello health checks for server testing
6072 May be used in sections : defaults | frontend | listen | backend
6073 yes | no | yes | yes
6074 Arguments : none
6075
6076 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
6077 possible to test that the server correctly talks SSL instead of just testing
6078 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
6079 SSLv3 client hello messages are sent once the connection is established to
6080 the server, and the response is analyzed to find an SSL server hello message.
6081 The server is considered valid only when the response contains this server
6082 hello message.
6083
6084 All servers tested till there correctly reply to SSLv3 client hello messages,
6085 and most servers tested do not even log the requests containing only hello
6086 messages, which is appreciable.
6087
Willy Tarreau763a95b2012-10-04 23:15:39 +02006088 Note that this check works even when SSL support was not built into haproxy
6089 because it forges the SSL message. When SSL support is available, it is best
6090 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006091
Willy Tarreau763a95b2012-10-04 23:15:39 +02006092 See also: "option httpchk", "check-ssl"
6093
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006094
Willy Tarreaued179852013-12-16 01:07:00 +01006095option tcp-check
6096 Perform health checks using tcp-check send/expect sequences
6097 May be used in sections: defaults | frontend | listen | backend
6098 yes | no | yes | yes
6099
6100 This health check method is intended to be combined with "tcp-check" command
6101 lists in order to support send/expect types of health check sequences.
6102
6103 TCP checks currently support 4 modes of operations :
6104 - no "tcp-check" directive : the health check only consists in a connection
6105 attempt, which remains the default mode.
6106
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006107 - "tcp-check send" or "tcp-check send-binary" only is mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01006108 used to send a string along with a connection opening. With some
6109 protocols, it helps sending a "QUIT" message for example that prevents
6110 the server from logging a connection error for each health check. The
6111 check result will still be based on the ability to open the connection
6112 only.
6113
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006114 - "tcp-check expect" only is mentioned : this is used to test a banner.
Willy Tarreaued179852013-12-16 01:07:00 +01006115 The connection is opened and haproxy waits for the server to present some
6116 contents which must validate some rules. The check result will be based
6117 on the matching between the contents and the rules. This is suited for
6118 POP, IMAP, SMTP, FTP, SSH, TELNET.
6119
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006120 - both "tcp-check send" and "tcp-check expect" are mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01006121 used to test a hello-type protocol. Haproxy sends a message, the server
6122 responds and its response is analysed. the check result will be based on
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006123 the matching between the response contents and the rules. This is often
Willy Tarreaued179852013-12-16 01:07:00 +01006124 suited for protocols which require a binding or a request/response model.
6125 LDAP, MySQL, Redis and SSL are example of such protocols, though they
6126 already all have their dedicated checks with a deeper understanding of
6127 the respective protocols.
6128 In this mode, many questions may be sent and many answers may be
6129 analysed.
6130
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006131 A fifth mode can be used to insert comments in different steps of the
6132 script.
6133
6134 For each tcp-check rule you create, you can add a "comment" directive,
6135 followed by a string. This string will be reported in the log and stderr
6136 in debug mode. It is useful to make user-friendly error reporting.
6137 The "comment" is of course optional.
6138
6139
Willy Tarreaued179852013-12-16 01:07:00 +01006140 Examples :
6141 # perform a POP check (analyse only server's banner)
6142 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006143 tcp-check expect string +OK\ POP3\ ready comment POP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01006144
6145 # perform an IMAP check (analyse only server's banner)
6146 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006147 tcp-check expect string *\ OK\ IMAP4\ ready comment IMAP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01006148
6149 # look for the redis master server after ensuring it speaks well
6150 # redis protocol, then it exits properly.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006151 # (send a command then analyse the response 3 times)
Willy Tarreaued179852013-12-16 01:07:00 +01006152 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006153 tcp-check comment PING\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01006154 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02006155 tcp-check expect string +PONG
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006156 tcp-check comment role\ check
Willy Tarreaued179852013-12-16 01:07:00 +01006157 tcp-check send info\ replication\r\n
6158 tcp-check expect string role:master
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006159 tcp-check comment QUIT\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01006160 tcp-check send QUIT\r\n
6161 tcp-check expect string +OK
6162
6163 forge a HTTP request, then analyse the response
6164 (send many headers before analyzing)
6165 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006166 tcp-check comment forge\ and\ send\ HTTP\ request
Willy Tarreaued179852013-12-16 01:07:00 +01006167 tcp-check send HEAD\ /\ HTTP/1.1\r\n
6168 tcp-check send Host:\ www.mydomain.com\r\n
6169 tcp-check send User-Agent:\ HAProxy\ tcpcheck\r\n
6170 tcp-check send \r\n
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006171 tcp-check expect rstring HTTP/1\..\ (2..|3..) comment check\ HTTP\ response
Willy Tarreaued179852013-12-16 01:07:00 +01006172
6173
6174 See also : "tcp-check expect", "tcp-check send"
6175
6176
Willy Tarreau9ea05a72009-06-14 12:07:01 +02006177option tcp-smart-accept
6178no option tcp-smart-accept
6179 Enable or disable the saving of one ACK packet during the accept sequence
6180 May be used in sections : defaults | frontend | listen | backend
6181 yes | yes | yes | no
6182 Arguments : none
6183
6184 When an HTTP connection request comes in, the system acknowledges it on
6185 behalf of HAProxy, then the client immediately sends its request, and the
6186 system acknowledges it too while it is notifying HAProxy about the new
6187 connection. HAProxy then reads the request and responds. This means that we
6188 have one TCP ACK sent by the system for nothing, because the request could
6189 very well be acknowledged by HAProxy when it sends its response.
6190
6191 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
6192 sending this useless ACK on platforms which support it (currently at least
6193 Linux). It must not cause any problem, because the system will send it anyway
6194 after 40 ms if the response takes more time than expected to come.
6195
6196 During complex network debugging sessions, it may be desirable to disable
6197 this optimization because delayed ACKs can make troubleshooting more complex
6198 when trying to identify where packets are delayed. It is then possible to
6199 fall back to normal behaviour by specifying "no option tcp-smart-accept".
6200
6201 It is also possible to force it for non-HTTP proxies by simply specifying
6202 "option tcp-smart-accept". For instance, it can make sense with some services
6203 such as SMTP where the server speaks first.
6204
6205 It is recommended to avoid forcing this option in a defaults section. In case
6206 of doubt, consider setting it back to automatic values by prepending the
6207 "default" keyword before it, or disabling it using the "no" keyword.
6208
Willy Tarreaud88edf22009-06-14 15:48:17 +02006209 See also : "option tcp-smart-connect"
6210
6211
6212option tcp-smart-connect
6213no option tcp-smart-connect
6214 Enable or disable the saving of one ACK packet during the connect sequence
6215 May be used in sections : defaults | frontend | listen | backend
6216 yes | no | yes | yes
6217 Arguments : none
6218
6219 On certain systems (at least Linux), HAProxy can ask the kernel not to
6220 immediately send an empty ACK upon a connection request, but to directly
6221 send the buffer request instead. This saves one packet on the network and
6222 thus boosts performance. It can also be useful for some servers, because they
6223 immediately get the request along with the incoming connection.
6224
6225 This feature is enabled when "option tcp-smart-connect" is set in a backend.
6226 It is not enabled by default because it makes network troubleshooting more
6227 complex.
6228
6229 It only makes sense to enable it with protocols where the client speaks first
6230 such as HTTP. In other situations, if there is no data to send in place of
6231 the ACK, a normal ACK is sent.
6232
6233 If this option has been enabled in a "defaults" section, it can be disabled
6234 in a specific instance by prepending the "no" keyword before it.
6235
6236 See also : "option tcp-smart-accept"
6237
Willy Tarreau9ea05a72009-06-14 12:07:01 +02006238
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006239option tcpka
6240 Enable or disable the sending of TCP keepalive packets on both sides
6241 May be used in sections : defaults | frontend | listen | backend
6242 yes | yes | yes | yes
6243 Arguments : none
6244
6245 When there is a firewall or any session-aware component between a client and
6246 a server, and when the protocol involves very long sessions with long idle
6247 periods (eg: remote desktops), there is a risk that one of the intermediate
6248 components decides to expire a session which has remained idle for too long.
6249
6250 Enabling socket-level TCP keep-alives makes the system regularly send packets
6251 to the other end of the connection, leaving it active. The delay between
6252 keep-alive probes is controlled by the system only and depends both on the
6253 operating system and its tuning parameters.
6254
6255 It is important to understand that keep-alive packets are neither emitted nor
6256 received at the application level. It is only the network stacks which sees
6257 them. For this reason, even if one side of the proxy already uses keep-alives
6258 to maintain its connection alive, those keep-alive packets will not be
6259 forwarded to the other side of the proxy.
6260
6261 Please note that this has nothing to do with HTTP keep-alive.
6262
6263 Using option "tcpka" enables the emission of TCP keep-alive probes on both
6264 the client and server sides of a connection. Note that this is meaningful
6265 only in "defaults" or "listen" sections. If this option is used in a
6266 frontend, only the client side will get keep-alives, and if this option is
6267 used in a backend, only the server side will get keep-alives. For this
6268 reason, it is strongly recommended to explicitly use "option clitcpka" and
6269 "option srvtcpka" when the configuration is split between frontends and
6270 backends.
6271
6272 See also : "option clitcpka", "option srvtcpka"
6273
Willy Tarreau844e3c52008-01-11 16:28:18 +01006274
6275option tcplog
6276 Enable advanced logging of TCP connections with session state and timers
6277 May be used in sections : defaults | frontend | listen | backend
6278 yes | yes | yes | yes
6279 Arguments : none
6280
6281 By default, the log output format is very poor, as it only contains the
6282 source and destination addresses, and the instance name. By specifying
6283 "option tcplog", each log line turns into a much richer format including, but
6284 not limited to, the connection timers, the session status, the connections
6285 numbers, the frontend, backend and server name, and of course the source
6286 address and ports. This option is useful for pure TCP proxies in order to
6287 find which of the client or server disconnects or times out. For normal HTTP
6288 proxies, it's better to use "option httplog" which is even more complete.
6289
6290 This option may be set either in the frontend or the backend.
6291
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006292 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01006293
6294
Willy Tarreau844e3c52008-01-11 16:28:18 +01006295option transparent
6296no option transparent
6297 Enable client-side transparent proxying
6298 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01006299 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01006300 Arguments : none
6301
6302 This option was introduced in order to provide layer 7 persistence to layer 3
6303 load balancers. The idea is to use the OS's ability to redirect an incoming
6304 connection for a remote address to a local process (here HAProxy), and let
6305 this process know what address was initially requested. When this option is
6306 used, sessions without cookies will be forwarded to the original destination
6307 IP address of the incoming request (which should match that of another
6308 equipment), while requests with cookies will still be forwarded to the
6309 appropriate server.
6310
6311 Note that contrary to a common belief, this option does NOT make HAProxy
6312 present the client's IP to the server when establishing the connection.
6313
Willy Tarreaua1146052011-03-01 09:51:54 +01006314 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006315 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01006316
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006317
Simon Horman98637e52014-06-20 12:30:16 +09006318external-check command <command>
6319 Executable to run when performing an external-check
6320 May be used in sections : defaults | frontend | listen | backend
6321 yes | no | yes | yes
6322
6323 Arguments :
6324 <command> is the external command to run
6325
Simon Horman98637e52014-06-20 12:30:16 +09006326 The arguments passed to the to the command are:
6327
Cyril Bonté777be862014-12-02 21:21:35 +01006328 <proxy_address> <proxy_port> <server_address> <server_port>
Simon Horman98637e52014-06-20 12:30:16 +09006329
Cyril Bonté777be862014-12-02 21:21:35 +01006330 The <proxy_address> and <proxy_port> are derived from the first listener
6331 that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
6332 listener the proxy_address will be the path of the socket and the
6333 <proxy_port> will be the string "NOT_USED". In a backend section, it's not
6334 possible to determine a listener, and both <proxy_address> and <proxy_port>
6335 will have the string value "NOT_USED".
Simon Horman98637e52014-06-20 12:30:16 +09006336
Cyril Bonté72cda2a2014-12-27 22:28:39 +01006337 Some values are also provided through environment variables.
6338
6339 Environment variables :
6340 HAPROXY_PROXY_ADDR The first bind address if available (or empty if not
6341 applicable, for example in a "backend" section).
6342
6343 HAPROXY_PROXY_ID The backend id.
6344
6345 HAPROXY_PROXY_NAME The backend name.
6346
6347 HAPROXY_PROXY_PORT The first bind port if available (or empty if not
6348 applicable, for example in a "backend" section or
6349 for a UNIX socket).
6350
6351 HAPROXY_SERVER_ADDR The server address.
6352
6353 HAPROXY_SERVER_CURCONN The current number of connections on the server.
6354
6355 HAPROXY_SERVER_ID The server id.
6356
6357 HAPROXY_SERVER_MAXCONN The server max connections.
6358
6359 HAPROXY_SERVER_NAME The server name.
6360
6361 HAPROXY_SERVER_PORT The server port if available (or empty for a UNIX
6362 socket).
6363
6364 PATH The PATH environment variable used when executing
6365 the command may be set using "external-check path".
6366
Simon Horman98637e52014-06-20 12:30:16 +09006367 If the command executed and exits with a zero status then the check is
6368 considered to have passed, otherwise the check is considered to have
6369 failed.
6370
6371 Example :
6372 external-check command /bin/true
6373
6374 See also : "external-check", "option external-check", "external-check path"
6375
6376
6377external-check path <path>
6378 The value of the PATH environment variable used when running an external-check
6379 May be used in sections : defaults | frontend | listen | backend
6380 yes | no | yes | yes
6381
6382 Arguments :
6383 <path> is the path used when executing external command to run
6384
6385 The default path is "".
6386
6387 Example :
6388 external-check path "/usr/bin:/bin"
6389
6390 See also : "external-check", "option external-check",
6391 "external-check command"
6392
6393
Emeric Brun647caf12009-06-30 17:57:00 +02006394persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02006395persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02006396 Enable RDP cookie-based persistence
6397 May be used in sections : defaults | frontend | listen | backend
6398 yes | no | yes | yes
6399 Arguments :
6400 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02006401 default cookie name "msts" will be used. There currently is no
6402 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02006403
6404 This statement enables persistence based on an RDP cookie. The RDP cookie
6405 contains all information required to find the server in the list of known
6406 servers. So when this option is set in the backend, the request is analysed
6407 and if an RDP cookie is found, it is decoded. If it matches a known server
6408 which is still UP (or if "option persist" is set), then the connection is
6409 forwarded to this server.
6410
6411 Note that this only makes sense in a TCP backend, but for this to work, the
6412 frontend must have waited long enough to ensure that an RDP cookie is present
6413 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006414 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02006415 a single "listen" section.
6416
Willy Tarreau61e28f22010-05-16 22:31:05 +02006417 Also, it is important to understand that the terminal server will emit this
6418 RDP cookie only if it is configured for "token redirection mode", which means
6419 that the "IP address redirection" option is disabled.
6420
Emeric Brun647caf12009-06-30 17:57:00 +02006421 Example :
6422 listen tse-farm
6423 bind :3389
6424 # wait up to 5s for an RDP cookie in the request
6425 tcp-request inspect-delay 5s
6426 tcp-request content accept if RDP_COOKIE
6427 # apply RDP cookie persistence
6428 persist rdp-cookie
6429 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02006430 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02006431 balance rdp-cookie
6432 server srv1 1.1.1.1:3389
6433 server srv2 1.1.1.2:3389
6434
Simon Hormanab814e02011-06-24 14:50:20 +09006435 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
6436 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02006437
6438
Willy Tarreau3a7d2072009-03-05 23:48:25 +01006439rate-limit sessions <rate>
6440 Set a limit on the number of new sessions accepted per second on a frontend
6441 May be used in sections : defaults | frontend | listen | backend
6442 yes | yes | yes | no
6443 Arguments :
6444 <rate> The <rate> parameter is an integer designating the maximum number
6445 of new sessions per second to accept on the frontend.
6446
6447 When the frontend reaches the specified number of new sessions per second, it
6448 stops accepting new connections until the rate drops below the limit again.
6449 During this time, the pending sessions will be kept in the socket's backlog
6450 (in system buffers) and haproxy will not even be aware that sessions are
6451 pending. When applying very low limit on a highly loaded service, it may make
6452 sense to increase the socket's backlog using the "backlog" keyword.
6453
6454 This feature is particularly efficient at blocking connection-based attacks
6455 or service abuse on fragile servers. Since the session rate is measured every
6456 millisecond, it is extremely accurate. Also, the limit applies immediately,
6457 no delay is needed at all to detect the threshold.
6458
6459 Example : limit the connection rate on SMTP to 10 per second max
6460 listen smtp
6461 mode tcp
6462 bind :25
6463 rate-limit sessions 10
6464 server 127.0.0.1:1025
6465
Willy Tarreaua17c2d92011-07-25 08:16:20 +02006466 Note : when the maximum rate is reached, the frontend's status is not changed
6467 but its sockets appear as "WAITING" in the statistics if the
6468 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +01006469
6470 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
6471
6472
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006473redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
6474redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
6475redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006476 Return an HTTP redirection if/unless a condition is matched
6477 May be used in sections : defaults | frontend | listen | backend
6478 no | yes | yes | yes
6479
6480 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +01006481 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006482
Willy Tarreau0140f252008-11-19 21:07:09 +01006483 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006484 <loc> With "redirect location", the exact value in <loc> is placed into
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006485 the HTTP "Location" header. When used in an "http-request" rule,
6486 <loc> value follows the log-format rules and can include some
6487 dynamic values (see Custom Log Format in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006488
6489 <pfx> With "redirect prefix", the "Location" header is built from the
6490 concatenation of <pfx> and the complete URI path, including the
6491 query string, unless the "drop-query" option is specified (see
6492 below). As a special case, if <pfx> equals exactly "/", then
6493 nothing is inserted before the original URI. It allows one to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006494 redirect to the same URL (for instance, to insert a cookie). When
6495 used in an "http-request" rule, <pfx> value follows the log-format
6496 rules and can include some dynamic values (see Custom Log Format
6497 in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006498
6499 <sch> With "redirect scheme", then the "Location" header is built by
6500 concatenating <sch> with "://" then the first occurrence of the
6501 "Host" header, and then the URI path, including the query string
6502 unless the "drop-query" option is specified (see below). If no
6503 path is found or if the path is "*", then "/" is used instead. If
6504 no "Host" header is found, then an empty host component will be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006505 returned, which most recent browsers interpret as redirecting to
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006506 the same host. This directive is mostly used to redirect HTTP to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006507 HTTPS. When used in an "http-request" rule, <sch> value follows
6508 the log-format rules and can include some dynamic values (see
6509 Custom Log Format in section 8.2.4).
Willy Tarreau0140f252008-11-19 21:07:09 +01006510
6511 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +01006512 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
6513 with 302 used by default if no code is specified. 301 means
6514 "Moved permanently", and a browser may cache the Location. 302
Baptiste Assmannea849c02015-08-03 11:42:50 +02006515 means "Moved temporarily" and means that the browser should not
Willy Tarreaub67fdc42013-03-29 19:28:11 +01006516 cache the redirection. 303 is equivalent to 302 except that the
6517 browser will fetch the location with a GET method. 307 is just
6518 like 302 but makes it clear that the same method must be reused.
6519 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +01006520
6521 <option> There are several options which can be specified to adjust the
6522 expected behaviour of a redirection :
6523
6524 - "drop-query"
6525 When this keyword is used in a prefix-based redirection, then the
6526 location will be set without any possible query-string, which is useful
6527 for directing users to a non-secure page for instance. It has no effect
6528 with a location-type redirect.
6529
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01006530 - "append-slash"
6531 This keyword may be used in conjunction with "drop-query" to redirect
6532 users who use a URL not ending with a '/' to the same one with the '/'.
6533 It can be useful to ensure that search engines will only see one URL.
6534 For this, a return code 301 is preferred.
6535
Willy Tarreau0140f252008-11-19 21:07:09 +01006536 - "set-cookie NAME[=value]"
6537 A "Set-Cookie" header will be added with NAME (and optionally "=value")
6538 to the response. This is sometimes used to indicate that a user has
6539 been seen, for instance to protect against some types of DoS. No other
6540 cookie option is added, so the cookie will be a session cookie. Note
6541 that for a browser, a sole cookie name without an equal sign is
6542 different from a cookie with an equal sign.
6543
6544 - "clear-cookie NAME[=]"
6545 A "Set-Cookie" header will be added with NAME (and optionally "="), but
6546 with the "Max-Age" attribute set to zero. This will tell the browser to
6547 delete this cookie. It is useful for instance on logout pages. It is
6548 important to note that clearing the cookie "NAME" will not remove a
6549 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
6550 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006551
6552 Example: move the login URL only to HTTPS.
6553 acl clear dst_port 80
6554 acl secure dst_port 8080
6555 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +01006556 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +01006557 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +01006558 acl cookie_set hdr_sub(cookie) SEEN=1
6559
6560 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +01006561 redirect prefix https://mysite.com if login_page !secure
6562 redirect prefix http://mysite.com drop-query if login_page !uid_given
6563 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +01006564 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006565
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01006566 Example: send redirects for request for articles without a '/'.
6567 acl missing_slash path_reg ^/article/[^/]*$
6568 redirect code 301 prefix / drop-query append-slash if missing_slash
6569
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006570 Example: redirect all HTTP traffic to HTTPS when SSL is handled by haproxy.
David BERARDe7153042012-11-03 00:11:31 +01006571 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006572
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006573 Example: append 'www.' prefix in front of all hosts not having it
6574 http-request redirect code 301 location www.%[hdr(host)]%[req.uri] \
6575 unless { hdr_beg(host) -i www }
6576
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006577 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006578
6579
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006580redisp (deprecated)
6581redispatch (deprecated)
6582 Enable or disable session redistribution in case of connection failure
6583 May be used in sections: defaults | frontend | listen | backend
6584 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006585 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006586
6587 In HTTP mode, if a server designated by a cookie is down, clients may
6588 definitely stick to it because they cannot flush the cookie, so they will not
6589 be able to access the service anymore.
6590
6591 Specifying "redispatch" will allow the proxy to break their persistence and
6592 redistribute them to a working server.
6593
6594 It also allows to retry last connection to another server in case of multiple
6595 connection failures. Of course, it requires having "retries" set to a nonzero
6596 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01006597
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006598 This form is deprecated, do not use it in any new configuration, use the new
6599 "option redispatch" instead.
6600
6601 See also : "option redispatch"
6602
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006603
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006604reqadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01006605 Add a header at the end of the HTTP request
6606 May be used in sections : defaults | frontend | listen | backend
6607 no | yes | yes | yes
6608 Arguments :
6609 <string> is the complete line to be added. Any space or known delimiter
6610 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006611 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006612
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006613 <cond> is an optional matching condition built from ACLs. It makes it
6614 possible to ignore this rule when other conditions are not met.
6615
Willy Tarreau303c0352008-01-17 19:01:39 +01006616 A new line consisting in <string> followed by a line feed will be added after
6617 the last header of an HTTP request.
6618
6619 Header transformations only apply to traffic which passes through HAProxy,
6620 and not to traffic generated by HAProxy, such as health-checks or error
6621 responses.
6622
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006623 Example : add "X-Proto: SSL" to requests coming via port 81
6624 acl is-ssl dst_port 81
6625 reqadd X-Proto:\ SSL if is-ssl
6626
6627 See also: "rspadd", section 6 about HTTP header manipulation, and section 7
6628 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006629
6630
Willy Tarreau5321c422010-01-28 20:35:13 +01006631reqallow <search> [{if | unless} <cond>]
6632reqiallow <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006633 Definitely allow an HTTP request if a line matches a regular expression
6634 May be used in sections : defaults | frontend | listen | backend
6635 no | yes | yes | yes
6636 Arguments :
6637 <search> is the regular expression applied to HTTP headers and to the
6638 request line. This is an extended regular expression. Parenthesis
6639 grouping is supported and no preliminary backslash is required.
6640 Any space or known delimiter must be escaped using a backslash
6641 ('\'). The pattern applies to a full line at a time. The
6642 "reqallow" keyword strictly matches case while "reqiallow"
6643 ignores case.
6644
Willy Tarreau5321c422010-01-28 20:35:13 +01006645 <cond> is an optional matching condition built from ACLs. It makes it
6646 possible to ignore this rule when other conditions are not met.
6647
Willy Tarreau303c0352008-01-17 19:01:39 +01006648 A request containing any line which matches extended regular expression
6649 <search> will mark the request as allowed, even if any later test would
6650 result in a deny. The test applies both to the request line and to request
6651 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01006652 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01006653
6654 It is easier, faster and more powerful to use ACLs to write access policies.
6655 Reqdeny, reqallow and reqpass should be avoided in new designs.
6656
6657 Example :
6658 # allow www.* but refuse *.local
6659 reqiallow ^Host:\ www\.
6660 reqideny ^Host:\ .*\.local
6661
Willy Tarreau5321c422010-01-28 20:35:13 +01006662 See also: "reqdeny", "block", section 6 about HTTP header manipulation, and
6663 section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006664
6665
Willy Tarreau5321c422010-01-28 20:35:13 +01006666reqdel <search> [{if | unless} <cond>]
6667reqidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006668 Delete all headers matching a regular expression in an HTTP request
6669 May be used in sections : defaults | frontend | listen | backend
6670 no | yes | yes | yes
6671 Arguments :
6672 <search> is the regular expression applied to HTTP headers and to the
6673 request line. This is an extended regular expression. Parenthesis
6674 grouping is supported and no preliminary backslash is required.
6675 Any space or known delimiter must be escaped using a backslash
6676 ('\'). The pattern applies to a full line at a time. The "reqdel"
6677 keyword strictly matches case while "reqidel" ignores case.
6678
Willy Tarreau5321c422010-01-28 20:35:13 +01006679 <cond> is an optional matching condition built from ACLs. It makes it
6680 possible to ignore this rule when other conditions are not met.
6681
Willy Tarreau303c0352008-01-17 19:01:39 +01006682 Any header line matching extended regular expression <search> in the request
6683 will be completely deleted. Most common use of this is to remove unwanted
6684 and/or dangerous headers or cookies from a request before passing it to the
6685 next servers.
6686
6687 Header transformations only apply to traffic which passes through HAProxy,
6688 and not to traffic generated by HAProxy, such as health-checks or error
6689 responses. Keep in mind that header names are not case-sensitive.
6690
6691 Example :
6692 # remove X-Forwarded-For header and SERVER cookie
6693 reqidel ^X-Forwarded-For:.*
6694 reqidel ^Cookie:.*SERVER=
6695
Willy Tarreau5321c422010-01-28 20:35:13 +01006696 See also: "reqadd", "reqrep", "rspdel", section 6 about HTTP header
6697 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006698
6699
Willy Tarreau5321c422010-01-28 20:35:13 +01006700reqdeny <search> [{if | unless} <cond>]
6701reqideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006702 Deny an HTTP request if a line matches a regular expression
6703 May be used in sections : defaults | frontend | listen | backend
6704 no | yes | yes | yes
6705 Arguments :
6706 <search> is the regular expression applied to HTTP headers and to the
6707 request line. This is an extended regular expression. Parenthesis
6708 grouping is supported and no preliminary backslash is required.
6709 Any space or known delimiter must be escaped using a backslash
6710 ('\'). The pattern applies to a full line at a time. The
6711 "reqdeny" keyword strictly matches case while "reqideny" ignores
6712 case.
6713
Willy Tarreau5321c422010-01-28 20:35:13 +01006714 <cond> is an optional matching condition built from ACLs. It makes it
6715 possible to ignore this rule when other conditions are not met.
6716
Willy Tarreau303c0352008-01-17 19:01:39 +01006717 A request containing any line which matches extended regular expression
6718 <search> will mark the request as denied, even if any later test would
6719 result in an allow. The test applies both to the request line and to request
6720 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01006721 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01006722
Willy Tarreauced27012008-01-17 20:35:34 +01006723 A denied request will generate an "HTTP 403 forbidden" response once the
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006724 complete request has been parsed. This is consistent with what is practiced
Willy Tarreaud72758d2010-01-12 10:42:19 +01006725 using ACLs.
Willy Tarreauced27012008-01-17 20:35:34 +01006726
Willy Tarreau303c0352008-01-17 19:01:39 +01006727 It is easier, faster and more powerful to use ACLs to write access policies.
6728 Reqdeny, reqallow and reqpass should be avoided in new designs.
6729
6730 Example :
6731 # refuse *.local, then allow www.*
6732 reqideny ^Host:\ .*\.local
6733 reqiallow ^Host:\ www\.
6734
Willy Tarreau5321c422010-01-28 20:35:13 +01006735 See also: "reqallow", "rspdeny", "block", section 6 about HTTP header
6736 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006737
6738
Willy Tarreau5321c422010-01-28 20:35:13 +01006739reqpass <search> [{if | unless} <cond>]
6740reqipass <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006741 Ignore any HTTP request line matching a regular expression in next rules
6742 May be used in sections : defaults | frontend | listen | backend
6743 no | yes | yes | yes
6744 Arguments :
6745 <search> is the regular expression applied to HTTP headers and to the
6746 request line. This is an extended regular expression. Parenthesis
6747 grouping is supported and no preliminary backslash is required.
6748 Any space or known delimiter must be escaped using a backslash
6749 ('\'). The pattern applies to a full line at a time. The
6750 "reqpass" keyword strictly matches case while "reqipass" ignores
6751 case.
6752
Willy Tarreau5321c422010-01-28 20:35:13 +01006753 <cond> is an optional matching condition built from ACLs. It makes it
6754 possible to ignore this rule when other conditions are not met.
6755
Willy Tarreau303c0352008-01-17 19:01:39 +01006756 A request containing any line which matches extended regular expression
6757 <search> will skip next rules, without assigning any deny or allow verdict.
6758 The test applies both to the request line and to request headers. Keep in
6759 mind that URLs in request line are case-sensitive while header names are not.
6760
6761 It is easier, faster and more powerful to use ACLs to write access policies.
6762 Reqdeny, reqallow and reqpass should be avoided in new designs.
6763
6764 Example :
6765 # refuse *.local, then allow www.*, but ignore "www.private.local"
6766 reqipass ^Host:\ www.private\.local
6767 reqideny ^Host:\ .*\.local
6768 reqiallow ^Host:\ www\.
6769
Willy Tarreau5321c422010-01-28 20:35:13 +01006770 See also: "reqallow", "reqdeny", "block", section 6 about HTTP header
6771 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006772
6773
Willy Tarreau5321c422010-01-28 20:35:13 +01006774reqrep <search> <string> [{if | unless} <cond>]
6775reqirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006776 Replace a regular expression with a string in an HTTP request line
6777 May be used in sections : defaults | frontend | listen | backend
6778 no | yes | yes | yes
6779 Arguments :
6780 <search> is the regular expression applied to HTTP headers and to the
6781 request line. This is an extended regular expression. Parenthesis
6782 grouping is supported and no preliminary backslash is required.
6783 Any space or known delimiter must be escaped using a backslash
6784 ('\'). The pattern applies to a full line at a time. The "reqrep"
6785 keyword strictly matches case while "reqirep" ignores case.
6786
6787 <string> is the complete line to be added. Any space or known delimiter
6788 must be escaped using a backslash ('\'). References to matched
6789 pattern groups are possible using the common \N form, with N
6790 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006791 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006792
Willy Tarreau5321c422010-01-28 20:35:13 +01006793 <cond> is an optional matching condition built from ACLs. It makes it
6794 possible to ignore this rule when other conditions are not met.
6795
Willy Tarreau303c0352008-01-17 19:01:39 +01006796 Any line matching extended regular expression <search> in the request (both
6797 the request line and header lines) will be completely replaced with <string>.
6798 Most common use of this is to rewrite URLs or domain names in "Host" headers.
6799
6800 Header transformations only apply to traffic which passes through HAProxy,
6801 and not to traffic generated by HAProxy, such as health-checks or error
6802 responses. Note that for increased readability, it is suggested to add enough
6803 spaces between the request and the response. Keep in mind that URLs in
6804 request line are case-sensitive while header names are not.
6805
6806 Example :
6807 # replace "/static/" with "/" at the beginning of any request path.
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04006808 reqrep ^([^\ :]*)\ /static/(.*) \1\ /\2
Willy Tarreau303c0352008-01-17 19:01:39 +01006809 # replace "www.mydomain.com" with "www" in the host name.
6810 reqirep ^Host:\ www.mydomain.com Host:\ www
6811
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04006812 See also: "reqadd", "reqdel", "rsprep", "tune.bufsize", section 6 about
6813 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006814
6815
Willy Tarreau5321c422010-01-28 20:35:13 +01006816reqtarpit <search> [{if | unless} <cond>]
6817reqitarpit <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006818 Tarpit an HTTP request containing a line matching a regular expression
6819 May be used in sections : defaults | frontend | listen | backend
6820 no | yes | yes | yes
6821 Arguments :
6822 <search> is the regular expression applied to HTTP headers and to the
6823 request line. This is an extended regular expression. Parenthesis
6824 grouping is supported and no preliminary backslash is required.
6825 Any space or known delimiter must be escaped using a backslash
6826 ('\'). The pattern applies to a full line at a time. The
6827 "reqtarpit" keyword strictly matches case while "reqitarpit"
6828 ignores case.
6829
Willy Tarreau5321c422010-01-28 20:35:13 +01006830 <cond> is an optional matching condition built from ACLs. It makes it
6831 possible to ignore this rule when other conditions are not met.
6832
Willy Tarreau303c0352008-01-17 19:01:39 +01006833 A request containing any line which matches extended regular expression
6834 <search> will be tarpitted, which means that it will connect to nowhere, will
Willy Tarreauced27012008-01-17 20:35:34 +01006835 be kept open for a pre-defined time, then will return an HTTP error 500 so
6836 that the attacker does not suspect it has been tarpitted. The status 500 will
6837 be reported in the logs, but the completion flags will indicate "PT". The
Willy Tarreau303c0352008-01-17 19:01:39 +01006838 delay is defined by "timeout tarpit", or "timeout connect" if the former is
6839 not set.
6840
6841 The goal of the tarpit is to slow down robots attacking servers with
6842 identifiable requests. Many robots limit their outgoing number of connections
6843 and stay connected waiting for a reply which can take several minutes to
6844 come. Depending on the environment and attack, it may be particularly
6845 efficient at reducing the load on the network and firewalls.
6846
Willy Tarreau5321c422010-01-28 20:35:13 +01006847 Examples :
Willy Tarreau303c0352008-01-17 19:01:39 +01006848 # ignore user-agents reporting any flavour of "Mozilla" or "MSIE", but
6849 # block all others.
6850 reqipass ^User-Agent:\.*(Mozilla|MSIE)
6851 reqitarpit ^User-Agent:
6852
Willy Tarreau5321c422010-01-28 20:35:13 +01006853 # block bad guys
6854 acl badguys src 10.1.0.3 172.16.13.20/28
6855 reqitarpit . if badguys
6856
6857 See also: "reqallow", "reqdeny", "reqpass", section 6 about HTTP header
6858 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006859
6860
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02006861retries <value>
6862 Set the number of retries to perform on a server after a connection failure
6863 May be used in sections: defaults | frontend | listen | backend
6864 yes | no | yes | yes
6865 Arguments :
6866 <value> is the number of times a connection attempt should be retried on
6867 a server when a connection either is refused or times out. The
6868 default value is 3.
6869
6870 It is important to understand that this value applies to the number of
6871 connection attempts, not full requests. When a connection has effectively
6872 been established to a server, there will be no more retry.
6873
6874 In order to avoid immediate reconnections to a server which is restarting,
Joseph Lynch726ab712015-05-11 23:25:34 -07006875 a turn-around timer of min("timeout connect", one second) is applied before
6876 a retry occurs.
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02006877
6878 When "option redispatch" is set, the last retry may be performed on another
6879 server even if a cookie references a different server.
6880
6881 See also : "option redispatch"
6882
6883
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006884rspadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01006885 Add a header at the end of the HTTP response
6886 May be used in sections : defaults | frontend | listen | backend
6887 no | yes | yes | yes
6888 Arguments :
6889 <string> is the complete line to be added. Any space or known delimiter
6890 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006891 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006892
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006893 <cond> is an optional matching condition built from ACLs. It makes it
6894 possible to ignore this rule when other conditions are not met.
6895
Willy Tarreau303c0352008-01-17 19:01:39 +01006896 A new line consisting in <string> followed by a line feed will be added after
6897 the last header of an HTTP response.
6898
6899 Header transformations only apply to traffic which passes through HAProxy,
6900 and not to traffic generated by HAProxy, such as health-checks or error
6901 responses.
6902
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006903 See also: "reqadd", section 6 about HTTP header manipulation, and section 7
6904 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006905
6906
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006907rspdel <search> [{if | unless} <cond>]
6908rspidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006909 Delete all headers matching a regular expression in an HTTP response
6910 May be used in sections : defaults | frontend | listen | backend
6911 no | yes | yes | yes
6912 Arguments :
6913 <search> is the regular expression applied to HTTP headers and to the
6914 response line. This is an extended regular expression, so
6915 parenthesis grouping is supported and no preliminary backslash
6916 is required. Any space or known delimiter must be escaped using
6917 a backslash ('\'). The pattern applies to a full line at a time.
6918 The "rspdel" keyword strictly matches case while "rspidel"
6919 ignores case.
6920
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006921 <cond> is an optional matching condition built from ACLs. It makes it
6922 possible to ignore this rule when other conditions are not met.
6923
Willy Tarreau303c0352008-01-17 19:01:39 +01006924 Any header line matching extended regular expression <search> in the response
6925 will be completely deleted. Most common use of this is to remove unwanted
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02006926 and/or sensitive headers or cookies from a response before passing it to the
Willy Tarreau303c0352008-01-17 19:01:39 +01006927 client.
6928
6929 Header transformations only apply to traffic which passes through HAProxy,
6930 and not to traffic generated by HAProxy, such as health-checks or error
6931 responses. Keep in mind that header names are not case-sensitive.
6932
6933 Example :
6934 # remove the Server header from responses
Willy Tarreau5e80e022013-05-25 08:31:25 +02006935 rspidel ^Server:.*
Willy Tarreau303c0352008-01-17 19:01:39 +01006936
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006937 See also: "rspadd", "rsprep", "reqdel", section 6 about HTTP header
6938 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006939
6940
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006941rspdeny <search> [{if | unless} <cond>]
6942rspideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006943 Block an HTTP response if a line matches a regular expression
6944 May be used in sections : defaults | frontend | listen | backend
6945 no | yes | yes | yes
6946 Arguments :
6947 <search> is the regular expression applied to HTTP headers and to the
6948 response line. This is an extended regular expression, so
6949 parenthesis grouping is supported and no preliminary backslash
6950 is required. Any space or known delimiter must be escaped using
6951 a backslash ('\'). The pattern applies to a full line at a time.
6952 The "rspdeny" keyword strictly matches case while "rspideny"
6953 ignores case.
6954
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006955 <cond> is an optional matching condition built from ACLs. It makes it
6956 possible to ignore this rule when other conditions are not met.
6957
Willy Tarreau303c0352008-01-17 19:01:39 +01006958 A response containing any line which matches extended regular expression
6959 <search> will mark the request as denied. The test applies both to the
6960 response line and to response headers. Keep in mind that header names are not
6961 case-sensitive.
6962
6963 Main use of this keyword is to prevent sensitive information leak and to
Willy Tarreauced27012008-01-17 20:35:34 +01006964 block the response before it reaches the client. If a response is denied, it
6965 will be replaced with an HTTP 502 error so that the client never retrieves
6966 any sensitive data.
Willy Tarreau303c0352008-01-17 19:01:39 +01006967
6968 It is easier, faster and more powerful to use ACLs to write access policies.
6969 Rspdeny should be avoided in new designs.
6970
6971 Example :
6972 # Ensure that no content type matching ms-word will leak
6973 rspideny ^Content-type:\.*/ms-word
6974
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006975 See also: "reqdeny", "acl", "block", section 6 about HTTP header manipulation
6976 and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006977
6978
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006979rsprep <search> <string> [{if | unless} <cond>]
6980rspirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006981 Replace a regular expression with a string in an HTTP response line
6982 May be used in sections : defaults | frontend | listen | backend
6983 no | yes | yes | yes
6984 Arguments :
6985 <search> is the regular expression applied to HTTP headers and to the
6986 response line. This is an extended regular expression, so
6987 parenthesis grouping is supported and no preliminary backslash
6988 is required. Any space or known delimiter must be escaped using
6989 a backslash ('\'). The pattern applies to a full line at a time.
6990 The "rsprep" keyword strictly matches case while "rspirep"
6991 ignores case.
6992
6993 <string> is the complete line to be added. Any space or known delimiter
6994 must be escaped using a backslash ('\'). References to matched
6995 pattern groups are possible using the common \N form, with N
6996 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006997 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006998
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006999 <cond> is an optional matching condition built from ACLs. It makes it
7000 possible to ignore this rule when other conditions are not met.
7001
Willy Tarreau303c0352008-01-17 19:01:39 +01007002 Any line matching extended regular expression <search> in the response (both
7003 the response line and header lines) will be completely replaced with
7004 <string>. Most common use of this is to rewrite Location headers.
7005
7006 Header transformations only apply to traffic which passes through HAProxy,
7007 and not to traffic generated by HAProxy, such as health-checks or error
7008 responses. Note that for increased readability, it is suggested to add enough
7009 spaces between the request and the response. Keep in mind that header names
7010 are not case-sensitive.
7011
7012 Example :
7013 # replace "Location: 127.0.0.1:8080" with "Location: www.mydomain.com"
7014 rspirep ^Location:\ 127.0.0.1:8080 Location:\ www.mydomain.com
7015
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007016 See also: "rspadd", "rspdel", "reqrep", section 6 about HTTP header
7017 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007018
7019
David du Colombier486df472011-03-17 10:40:26 +01007020server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007021 Declare a server in a backend
7022 May be used in sections : defaults | frontend | listen | backend
7023 no | no | yes | yes
7024 Arguments :
7025 <name> is the internal name assigned to this server. This name will
Cyril Bonté941a0c62012-10-15 19:44:24 +02007026 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -05007027 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007028
David du Colombier486df472011-03-17 10:40:26 +01007029 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
7030 resolvable hostname is supported, but this name will be resolved
7031 during start-up. Address "0.0.0.0" or "*" has a special meaning.
7032 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +02007033 address as the one from the client connection. This is useful in
7034 transparent proxy architectures where the client's connection is
7035 intercepted and haproxy must forward to the original destination
7036 address. This is more or less what the "transparent" keyword does
7037 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +01007038 to report statistics. Optionally, an address family prefix may be
7039 used before the address to force the family regardless of the
7040 address format, which can be useful to specify a path to a unix
7041 socket with no slash ('/'). Currently supported prefixes are :
7042 - 'ipv4@' -> address is always IPv4
7043 - 'ipv6@' -> address is always IPv6
7044 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02007045 - 'abns@' -> address is in abstract namespace (Linux only)
William Lallemandb2f07452015-05-12 14:27:13 +02007046 You may want to reference some environment variables in the
7047 address parameter, see section 2.3 about environment
7048 variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007049
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007050 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007051 be sent to this port. If unset, the same port the client
7052 connected to will be used. The port may also be prefixed by a "+"
7053 or a "-". In this case, the server's port will be determined by
7054 adding this value to the client's port.
7055
7056 <param*> is a list of parameters for this server. The "server" keywords
7057 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007058 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007059
7060 Examples :
7061 server first 10.1.1.1:1080 cookie first check inter 1000
7062 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +01007063 server transp ipv4@
William Lallemandb2f07452015-05-12 14:27:13 +02007064 server backup "${SRV_BACKUP}:1080" backup
7065 server www1_dc1 "${LAN_DC1}.101:80"
7066 server www1_dc2 "${LAN_DC2}.101:80"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007067
Willy Tarreau55dcaf62015-09-27 15:03:15 +02007068 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
7069 sun_path length is used for the address length. Some other programs
7070 such as socat use the string length only by default. Pass the option
7071 ",unix-tightsocklen=0" to any abstract socket definition in socat to
7072 make it compatible with HAProxy's.
7073
Mark Lamourinec2247f02012-01-04 13:02:01 -05007074 See also: "default-server", "http-send-name-header" and section 5 about
7075 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007076
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007077server-state-file-name [<file>]
7078 Set the server state file to read, load and apply to servers available in
7079 this backend. It only applies when the directive "load-server-state-from-file"
7080 is set to "local". When <file> is not provided or if this directive is not
7081 set, then backend name is used. If <file> starts with a slash '/', then it is
7082 considered as an absolute path. Otherwise, <file> is concatenated to the
7083 global directive "server-state-file-base".
7084
7085 Example: the minimal configuration below would make HAProxy look for the
7086 state server file '/etc/haproxy/states/bk':
7087
7088 global
7089 server-state-file-base /etc/haproxy/states
7090
7091 backend bk
7092 load-server-state-from-file
7093
7094 See also: "server-state-file-base", "load-server-state-from-file", and
7095 "show servers state"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007096
7097source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02007098source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +01007099source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007100 Set the source address for outgoing connections
7101 May be used in sections : defaults | frontend | listen | backend
7102 yes | no | yes | yes
7103 Arguments :
7104 <addr> is the IPv4 address HAProxy will bind to before connecting to a
7105 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +01007106
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007107 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +01007108 the most appropriate address to reach its destination. Optionally
7109 an address family prefix may be used before the address to force
7110 the family regardless of the address format, which can be useful
7111 to specify a path to a unix socket with no slash ('/'). Currently
7112 supported prefixes are :
7113 - 'ipv4@' -> address is always IPv4
7114 - 'ipv6@' -> address is always IPv6
7115 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02007116 - 'abns@' -> address is in abstract namespace (Linux only)
Cyril Bonté307ee1e2015-09-28 23:16:06 +02007117 You may want to reference some environment variables in the
7118 address parameter, see section 2.3 about environment variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007119
7120 <port> is an optional port. It is normally not needed but may be useful
7121 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02007122 the system will select a free port. Note that port ranges are not
7123 supported in the backend. If you want to force port ranges, you
7124 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007125
7126 <addr2> is the IP address to present to the server when connections are
7127 forwarded in full transparent proxy mode. This is currently only
7128 supported on some patched Linux kernels. When this address is
7129 specified, clients connecting to the server will be presented
7130 with this address, while health checks will still use the address
7131 <addr>.
7132
7133 <port2> is the optional port to present to the server when connections
7134 are forwarded in full transparent proxy mode (see <addr2> above).
7135 The default value of zero means the system will select a free
7136 port.
7137
Willy Tarreaubce70882009-09-07 11:51:47 +02007138 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
7139 This is the name of a comma-separated header list which can
7140 contain multiple IP addresses. By default, the last occurrence is
7141 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +01007142 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +02007143 by previous proxy, typically Stunnel. In order to use another
7144 occurrence from the last one, please see the <occ> parameter
7145 below. When the header (or occurrence) is not found, no binding
7146 is performed so that the proxy's default IP address is used. Also
7147 keep in mind that the header name is case insensitive, as for any
7148 HTTP header.
7149
7150 <occ> is the occurrence number of a value to be used in a multi-value
7151 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007152 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +02007153 address. Positive values indicate a position from the first
7154 occurrence, 1 being the first one. Negative values indicate
7155 positions relative to the last one, -1 being the last one. This
7156 is helpful for situations where an X-Forwarded-For header is set
7157 at the entry point of an infrastructure and must be used several
7158 proxy layers away. When this value is not specified, -1 is
7159 assumed. Passing a zero here disables the feature.
7160
Willy Tarreaud53f96b2009-02-04 18:46:54 +01007161 <name> is an optional interface name to which to bind to for outgoing
7162 traffic. On systems supporting this features (currently, only
7163 Linux), this allows one to bind all traffic to the server to
7164 this interface even if it is not the one the system would select
7165 based on routing tables. This should be used with extreme care.
7166 Note that using this option requires root privileges.
7167
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007168 The "source" keyword is useful in complex environments where a specific
7169 address only is allowed to connect to the servers. It may be needed when a
7170 private address must be used through a public gateway for instance, and it is
7171 known that the system cannot determine the adequate source address by itself.
7172
7173 An extension which is available on certain patched Linux kernels may be used
7174 through the "usesrc" optional keyword. It makes it possible to connect to the
7175 servers with an IP address which does not belong to the system itself. This
7176 is called "full transparent proxy mode". For this to work, the destination
7177 servers have to route their traffic back to this address through the machine
7178 running HAProxy, and IP forwarding must generally be enabled on this machine.
7179
7180 In this "full transparent proxy" mode, it is possible to force a specific IP
7181 address to be presented to the servers. This is not much used in fact. A more
7182 common use is to tell HAProxy to present the client's IP address. For this,
7183 there are two methods :
7184
7185 - present the client's IP and port addresses. This is the most transparent
7186 mode, but it can cause problems when IP connection tracking is enabled on
7187 the machine, because a same connection may be seen twice with different
7188 states. However, this solution presents the huge advantage of not
7189 limiting the system to the 64k outgoing address+port couples, because all
7190 of the client ranges may be used.
7191
7192 - present only the client's IP address and select a spare port. This
7193 solution is still quite elegant but slightly less transparent (downstream
7194 firewalls logs will not match upstream's). It also presents the downside
7195 of limiting the number of concurrent connections to the usual 64k ports.
7196 However, since the upstream and downstream ports are different, local IP
7197 connection tracking on the machine will not be upset by the reuse of the
7198 same session.
7199
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007200 This option sets the default source for all servers in the backend. It may
7201 also be specified in a "defaults" section. Finer source address specification
7202 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007203 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007204
Baptiste Assmann91bd3372015-07-17 21:59:42 +02007205 In order to work, "usesrc" requires root privileges.
7206
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007207 Examples :
7208 backend private
7209 # Connect to the servers using our 192.168.1.200 source address
7210 source 192.168.1.200
7211
7212 backend transparent_ssl1
7213 # Connect to the SSL farm from the client's source address
7214 source 192.168.1.200 usesrc clientip
7215
7216 backend transparent_ssl2
7217 # Connect to the SSL farm from the client's source address and port
7218 # not recommended if IP conntrack is present on the local machine.
7219 source 192.168.1.200 usesrc client
7220
7221 backend transparent_ssl3
7222 # Connect to the SSL farm from the client's source address. It
7223 # is more conntrack-friendly.
7224 source 192.168.1.200 usesrc clientip
7225
7226 backend transparent_smtp
7227 # Connect to the SMTP farm from the client's source address/port
7228 # with Tproxy version 4.
7229 source 0.0.0.0 usesrc clientip
7230
Willy Tarreaubce70882009-09-07 11:51:47 +02007231 backend transparent_http
7232 # Connect to the servers using the client's IP as seen by previous
7233 # proxy.
7234 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
7235
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007236 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007237 the Linux kernel on www.balabit.com, the "bind" keyword.
7238
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007239
Willy Tarreau844e3c52008-01-11 16:28:18 +01007240srvtimeout <timeout> (deprecated)
7241 Set the maximum inactivity time on the server side.
7242 May be used in sections : defaults | frontend | listen | backend
7243 yes | no | yes | yes
7244 Arguments :
7245 <timeout> is the timeout value specified in milliseconds by default, but
7246 can be in any other unit if the number is suffixed by the unit,
7247 as explained at the top of this document.
7248
7249 The inactivity timeout applies when the server is expected to acknowledge or
7250 send data. In HTTP mode, this timeout is particularly important to consider
7251 during the first phase of the server's response, when it has to send the
7252 headers, as it directly represents the server's processing time for the
7253 request. To find out what value to put there, it's often good to start with
7254 what would be considered as unacceptable response times, then check the logs
7255 to observe the response time distribution, and adjust the value accordingly.
7256
7257 The value is specified in milliseconds by default, but can be in any other
7258 unit if the number is suffixed by the unit, as specified at the top of this
7259 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
7260 recommended that the client timeout remains equal to the server timeout in
7261 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007262 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01007263 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreaud72758d2010-01-12 10:42:19 +01007264 seconds (eg: 4 or 5 seconds minimum).
Willy Tarreau844e3c52008-01-11 16:28:18 +01007265
7266 This parameter is specific to backends, but can be specified once for all in
7267 "defaults" sections. This is in fact one of the easiest solutions not to
7268 forget about it. An unspecified timeout results in an infinite timeout, which
7269 is not recommended. Such a usage is accepted and works but reports a warning
7270 during startup because it may results in accumulation of expired sessions in
7271 the system if the system's timeouts are not configured either.
7272
7273 This parameter is provided for compatibility but is currently deprecated.
7274 Please use "timeout server" instead.
7275
Willy Tarreauce887fd2012-05-12 12:50:00 +02007276 See also : "timeout server", "timeout tunnel", "timeout client" and
7277 "clitimeout".
Willy Tarreau844e3c52008-01-11 16:28:18 +01007278
7279
Cyril Bonté66c327d2010-10-12 00:14:37 +02007280stats admin { if | unless } <cond>
7281 Enable statistics admin level if/unless a condition is matched
7282 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007283 no | yes | yes | yes
Cyril Bonté66c327d2010-10-12 00:14:37 +02007284
7285 This statement enables the statistics admin level if/unless a condition is
7286 matched.
7287
7288 The admin level allows to enable/disable servers from the web interface. By
7289 default, statistics page is read-only for security reasons.
7290
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007291 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7292 unless you know what you do : memory is not shared between the
7293 processes, which can result in random behaviours.
7294
Cyril Bonté23b39d92011-02-10 22:54:44 +01007295 Currently, the POST request is limited to the buffer size minus the reserved
7296 buffer space, which means that if the list of servers is too long, the
7297 request won't be processed. It is recommended to alter few servers at a
7298 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +02007299
7300 Example :
7301 # statistics admin level only for localhost
7302 backend stats_localhost
7303 stats enable
7304 stats admin if LOCALHOST
7305
7306 Example :
7307 # statistics admin level always enabled because of the authentication
7308 backend stats_auth
7309 stats enable
7310 stats auth admin:AdMiN123
7311 stats admin if TRUE
7312
7313 Example :
7314 # statistics admin level depends on the authenticated user
7315 userlist stats-auth
7316 group admin users admin
7317 user admin insecure-password AdMiN123
7318 group readonly users haproxy
7319 user haproxy insecure-password haproxy
7320
7321 backend stats_auth
7322 stats enable
7323 acl AUTH http_auth(stats-auth)
7324 acl AUTH_ADMIN http_auth_group(stats-auth) admin
7325 stats http-request auth unless AUTH
7326 stats admin if AUTH_ADMIN
7327
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007328 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
7329 "bind-process", section 3.4 about userlists and section 7 about
7330 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +02007331
7332
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007333stats auth <user>:<passwd>
7334 Enable statistics with authentication and grant access to an account
7335 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007336 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007337 Arguments :
7338 <user> is a user name to grant access to
7339
7340 <passwd> is the cleartext password associated to this user
7341
7342 This statement enables statistics with default settings, and restricts access
7343 to declared users only. It may be repeated as many times as necessary to
7344 allow as many users as desired. When a user tries to access the statistics
7345 without a valid account, a "401 Forbidden" response will be returned so that
7346 the browser asks the user to provide a valid user and password. The real
7347 which will be returned to the browser is configurable using "stats realm".
7348
7349 Since the authentication method is HTTP Basic Authentication, the passwords
7350 circulate in cleartext on the network. Thus, it was decided that the
7351 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02007352 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007353
7354 It is also possible to reduce the scope of the proxies which appear in the
7355 report using "stats scope".
7356
7357 Though this statement alone is enough to enable statistics reporting, it is
7358 recommended to set all other settings in order to avoid relying on default
7359 unobvious parameters.
7360
7361 Example :
7362 # public access (limited to this backend only)
7363 backend public_www
7364 server srv1 192.168.0.1:80
7365 stats enable
7366 stats hide-version
7367 stats scope .
7368 stats uri /admin?stats
7369 stats realm Haproxy\ Statistics
7370 stats auth admin1:AdMiN123
7371 stats auth admin2:AdMiN321
7372
7373 # internal monitoring access (unlimited)
7374 backend private_monitoring
7375 stats enable
7376 stats uri /admin?stats
7377 stats refresh 5s
7378
7379 See also : "stats enable", "stats realm", "stats scope", "stats uri"
7380
7381
7382stats enable
7383 Enable statistics reporting with default settings
7384 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007385 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007386 Arguments : none
7387
7388 This statement enables statistics reporting with default settings defined
7389 at build time. Unless stated otherwise, these settings are used :
7390 - stats uri : /haproxy?stats
7391 - stats realm : "HAProxy Statistics"
7392 - stats auth : no authentication
7393 - stats scope : no restriction
7394
7395 Though this statement alone is enough to enable statistics reporting, it is
7396 recommended to set all other settings in order to avoid relying on default
7397 unobvious parameters.
7398
7399 Example :
7400 # public access (limited to this backend only)
7401 backend public_www
7402 server srv1 192.168.0.1:80
7403 stats enable
7404 stats hide-version
7405 stats scope .
7406 stats uri /admin?stats
7407 stats realm Haproxy\ Statistics
7408 stats auth admin1:AdMiN123
7409 stats auth admin2:AdMiN321
7410
7411 # internal monitoring access (unlimited)
7412 backend private_monitoring
7413 stats enable
7414 stats uri /admin?stats
7415 stats refresh 5s
7416
7417 See also : "stats auth", "stats realm", "stats uri"
7418
7419
Willy Tarreaud63335a2010-02-26 12:56:52 +01007420stats hide-version
7421 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007422 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007423 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01007424 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007425
Willy Tarreaud63335a2010-02-26 12:56:52 +01007426 By default, the stats page reports some useful status information along with
7427 the statistics. Among them is HAProxy's version. However, it is generally
7428 considered dangerous to report precise version to anyone, as it can help them
7429 target known weaknesses with specific attacks. The "stats hide-version"
7430 statement removes the version from the statistics report. This is recommended
7431 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007432
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02007433 Though this statement alone is enough to enable statistics reporting, it is
7434 recommended to set all other settings in order to avoid relying on default
7435 unobvious parameters.
7436
Willy Tarreaud63335a2010-02-26 12:56:52 +01007437 Example :
7438 # public access (limited to this backend only)
7439 backend public_www
7440 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02007441 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01007442 stats hide-version
7443 stats scope .
7444 stats uri /admin?stats
7445 stats realm Haproxy\ Statistics
7446 stats auth admin1:AdMiN123
7447 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007448
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007449 # internal monitoring access (unlimited)
7450 backend private_monitoring
7451 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01007452 stats uri /admin?stats
7453 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +01007454
Willy Tarreaud63335a2010-02-26 12:56:52 +01007455 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007456
Willy Tarreau983e01e2010-01-11 18:42:06 +01007457
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02007458stats http-request { allow | deny | auth [realm <realm>] }
7459 [ { if | unless } <condition> ]
7460 Access control for statistics
7461
7462 May be used in sections: defaults | frontend | listen | backend
7463 no | no | yes | yes
7464
7465 As "http-request", these set of options allow to fine control access to
7466 statistics. Each option may be followed by if/unless and acl.
7467 First option with matched condition (or option without condition) is final.
7468 For "deny" a 403 error will be returned, for "allow" normal processing is
7469 performed, for "auth" a 401/407 error code is returned so the client
7470 should be asked to enter a username and password.
7471
7472 There is no fixed limit to the number of http-request statements per
7473 instance.
7474
7475 See also : "http-request", section 3.4 about userlists and section 7
7476 about ACL usage.
7477
7478
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007479stats realm <realm>
7480 Enable statistics and set authentication realm
7481 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007482 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007483 Arguments :
7484 <realm> is the name of the HTTP Basic Authentication realm reported to
7485 the browser. The browser uses it to display it in the pop-up
7486 inviting the user to enter a valid username and password.
7487
7488 The realm is read as a single word, so any spaces in it should be escaped
7489 using a backslash ('\').
7490
7491 This statement is useful only in conjunction with "stats auth" since it is
7492 only related to authentication.
7493
7494 Though this statement alone is enough to enable statistics reporting, it is
7495 recommended to set all other settings in order to avoid relying on default
7496 unobvious parameters.
7497
7498 Example :
7499 # public access (limited to this backend only)
7500 backend public_www
7501 server srv1 192.168.0.1:80
7502 stats enable
7503 stats hide-version
7504 stats scope .
7505 stats uri /admin?stats
7506 stats realm Haproxy\ Statistics
7507 stats auth admin1:AdMiN123
7508 stats auth admin2:AdMiN321
7509
7510 # internal monitoring access (unlimited)
7511 backend private_monitoring
7512 stats enable
7513 stats uri /admin?stats
7514 stats refresh 5s
7515
7516 See also : "stats auth", "stats enable", "stats uri"
7517
7518
7519stats refresh <delay>
7520 Enable statistics with automatic refresh
7521 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007522 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007523 Arguments :
7524 <delay> is the suggested refresh delay, specified in seconds, which will
7525 be returned to the browser consulting the report page. While the
7526 browser is free to apply any delay, it will generally respect it
7527 and refresh the page this every seconds. The refresh interval may
7528 be specified in any other non-default time unit, by suffixing the
7529 unit after the value, as explained at the top of this document.
7530
7531 This statement is useful on monitoring displays with a permanent page
7532 reporting the load balancer's activity. When set, the HTML report page will
7533 include a link "refresh"/"stop refresh" so that the user can select whether
7534 he wants automatic refresh of the page or not.
7535
7536 Though this statement alone is enough to enable statistics reporting, it is
7537 recommended to set all other settings in order to avoid relying on default
7538 unobvious parameters.
7539
7540 Example :
7541 # public access (limited to this backend only)
7542 backend public_www
7543 server srv1 192.168.0.1:80
7544 stats enable
7545 stats hide-version
7546 stats scope .
7547 stats uri /admin?stats
7548 stats realm Haproxy\ Statistics
7549 stats auth admin1:AdMiN123
7550 stats auth admin2:AdMiN321
7551
7552 # internal monitoring access (unlimited)
7553 backend private_monitoring
7554 stats enable
7555 stats uri /admin?stats
7556 stats refresh 5s
7557
7558 See also : "stats auth", "stats enable", "stats realm", "stats uri"
7559
7560
7561stats scope { <name> | "." }
7562 Enable statistics and limit access scope
7563 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007564 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007565 Arguments :
7566 <name> is the name of a listen, frontend or backend section to be
7567 reported. The special name "." (a single dot) designates the
7568 section in which the statement appears.
7569
7570 When this statement is specified, only the sections enumerated with this
7571 statement will appear in the report. All other ones will be hidden. This
7572 statement may appear as many times as needed if multiple sections need to be
7573 reported. Please note that the name checking is performed as simple string
7574 comparisons, and that it is never checked that a give section name really
7575 exists.
7576
7577 Though this statement alone is enough to enable statistics reporting, it is
7578 recommended to set all other settings in order to avoid relying on default
7579 unobvious parameters.
7580
7581 Example :
7582 # public access (limited to this backend only)
7583 backend public_www
7584 server srv1 192.168.0.1:80
7585 stats enable
7586 stats hide-version
7587 stats scope .
7588 stats uri /admin?stats
7589 stats realm Haproxy\ Statistics
7590 stats auth admin1:AdMiN123
7591 stats auth admin2:AdMiN321
7592
7593 # internal monitoring access (unlimited)
7594 backend private_monitoring
7595 stats enable
7596 stats uri /admin?stats
7597 stats refresh 5s
7598
7599 See also : "stats auth", "stats enable", "stats realm", "stats uri"
7600
Willy Tarreaud63335a2010-02-26 12:56:52 +01007601
Willy Tarreauc9705a12010-07-27 20:05:50 +02007602stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +01007603 Enable reporting of a description on the statistics page.
7604 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007605 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01007606
Willy Tarreauc9705a12010-07-27 20:05:50 +02007607 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +01007608 description from global section is automatically used instead.
7609
7610 This statement is useful for users that offer shared services to their
7611 customers, where node or description should be different for each customer.
7612
7613 Though this statement alone is enough to enable statistics reporting, it is
7614 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007615 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +01007616
7617 Example :
7618 # internal monitoring access (unlimited)
7619 backend private_monitoring
7620 stats enable
7621 stats show-desc Master node for Europe, Asia, Africa
7622 stats uri /admin?stats
7623 stats refresh 5s
7624
7625 See also: "show-node", "stats enable", "stats uri" and "description" in
7626 global section.
7627
7628
7629stats show-legends
Willy Tarreaued2119c2014-04-24 22:10:39 +02007630 Enable reporting additional information on the statistics page
7631 May be used in sections : defaults | frontend | listen | backend
7632 yes | yes | yes | yes
7633 Arguments : none
7634
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007635 Enable reporting additional information on the statistics page :
Willy Tarreaud63335a2010-02-26 12:56:52 +01007636 - cap: capabilities (proxy)
7637 - mode: one of tcp, http or health (proxy)
7638 - id: SNMP ID (proxy, socket, server)
7639 - IP (socket, server)
7640 - cookie (backend, server)
7641
7642 Though this statement alone is enough to enable statistics reporting, it is
7643 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007644 unobvious parameters. Default behaviour is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +01007645
7646 See also: "stats enable", "stats uri".
7647
7648
7649stats show-node [ <name> ]
7650 Enable reporting of a host name on the statistics page.
7651 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007652 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01007653 Arguments:
7654 <name> is an optional name to be reported. If unspecified, the
7655 node name from global section is automatically used instead.
7656
7657 This statement is useful for users that offer shared services to their
7658 customers, where node or description might be different on a stats page
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007659 provided for each customer. Default behaviour is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +01007660
7661 Though this statement alone is enough to enable statistics reporting, it is
7662 recommended to set all other settings in order to avoid relying on default
7663 unobvious parameters.
7664
7665 Example:
7666 # internal monitoring access (unlimited)
7667 backend private_monitoring
7668 stats enable
7669 stats show-node Europe-1
7670 stats uri /admin?stats
7671 stats refresh 5s
7672
7673 See also: "show-desc", "stats enable", "stats uri", and "node" in global
7674 section.
7675
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007676
7677stats uri <prefix>
7678 Enable statistics and define the URI prefix to access them
7679 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007680 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007681 Arguments :
7682 <prefix> is the prefix of any URI which will be redirected to stats. This
7683 prefix may contain a question mark ('?') to indicate part of a
7684 query string.
7685
7686 The statistics URI is intercepted on the relayed traffic, so it appears as a
7687 page within the normal application. It is strongly advised to ensure that the
7688 selected URI will never appear in the application, otherwise it will never be
7689 possible to reach it in the application.
7690
7691 The default URI compiled in haproxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007692 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007693 It is generally a good idea to include a question mark in the URI so that
7694 intermediate proxies refrain from caching the results. Also, since any string
7695 beginning with the prefix will be accepted as a stats request, the question
7696 mark helps ensuring that no valid URI will begin with the same words.
7697
7698 It is sometimes very convenient to use "/" as the URI prefix, and put that
7699 statement in a "listen" instance of its own. That makes it easy to dedicate
7700 an address or a port to statistics only.
7701
7702 Though this statement alone is enough to enable statistics reporting, it is
7703 recommended to set all other settings in order to avoid relying on default
7704 unobvious parameters.
7705
7706 Example :
7707 # public access (limited to this backend only)
7708 backend public_www
7709 server srv1 192.168.0.1:80
7710 stats enable
7711 stats hide-version
7712 stats scope .
7713 stats uri /admin?stats
7714 stats realm Haproxy\ Statistics
7715 stats auth admin1:AdMiN123
7716 stats auth admin2:AdMiN321
7717
7718 # internal monitoring access (unlimited)
7719 backend private_monitoring
7720 stats enable
7721 stats uri /admin?stats
7722 stats refresh 5s
7723
7724 See also : "stats auth", "stats enable", "stats realm"
7725
7726
Willy Tarreaud63335a2010-02-26 12:56:52 +01007727stick match <pattern> [table <table>] [{if | unless} <cond>]
7728 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007729 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +01007730 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007731
7732 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02007733 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007734 describes what elements of the incoming request or connection
7735 will be analysed in the hope to find a matching entry in a
7736 stickiness table. This rule is mandatory.
7737
7738 <table> is an optional stickiness table name. If unspecified, the same
7739 backend's table is used. A stickiness table is declared using
7740 the "stick-table" statement.
7741
7742 <cond> is an optional matching condition. It makes it possible to match
7743 on a certain criterion only when other conditions are met (or
7744 not met). For instance, it could be used to match on a source IP
7745 address except when a request passes through a known proxy, in
7746 which case we'd match on a header containing that IP address.
7747
7748 Some protocols or applications require complex stickiness rules and cannot
7749 always simply rely on cookies nor hashing. The "stick match" statement
7750 describes a rule to extract the stickiness criterion from an incoming request
7751 or connection. See section 7 for a complete list of possible patterns and
7752 transformation rules.
7753
7754 The table has to be declared using the "stick-table" statement. It must be of
7755 a type compatible with the pattern. By default it is the one which is present
7756 in the same backend. It is possible to share a table with other backends by
7757 referencing it using the "table" keyword. If another table is referenced,
7758 the server's ID inside the backends are used. By default, all server IDs
7759 start at 1 in each backend, so the server ordering is enough. But in case of
7760 doubt, it is highly recommended to force server IDs using their "id" setting.
7761
7762 It is possible to restrict the conditions where a "stick match" statement
7763 will apply, using "if" or "unless" followed by a condition. See section 7 for
7764 ACL based conditions.
7765
7766 There is no limit on the number of "stick match" statements. The first that
7767 applies and matches will cause the request to be directed to the same server
7768 as was used for the request which created the entry. That way, multiple
7769 matches can be used as fallbacks.
7770
7771 The stick rules are checked after the persistence cookies, so they will not
7772 affect stickiness if a cookie has already been used to select a server. That
7773 way, it becomes very easy to insert cookies and match on IP addresses in
7774 order to maintain stickiness between HTTP and HTTPS.
7775
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007776 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7777 unless you know what you do : memory is not shared between the
7778 processes, which can result in random behaviours.
7779
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007780 Example :
7781 # forward SMTP users to the same server they just used for POP in the
7782 # last 30 minutes
7783 backend pop
7784 mode tcp
7785 balance roundrobin
7786 stick store-request src
7787 stick-table type ip size 200k expire 30m
7788 server s1 192.168.1.1:110
7789 server s2 192.168.1.1:110
7790
7791 backend smtp
7792 mode tcp
7793 balance roundrobin
7794 stick match src table pop
7795 server s1 192.168.1.1:25
7796 server s2 192.168.1.1:25
7797
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007798 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02007799 about ACLs and samples fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007800
7801
7802stick on <pattern> [table <table>] [{if | unless} <condition>]
7803 Define a request pattern to associate a user to a server
7804 May be used in sections : defaults | frontend | listen | backend
7805 no | no | yes | yes
7806
7807 Note : This form is exactly equivalent to "stick match" followed by
7808 "stick store-request", all with the same arguments. Please refer
7809 to both keywords for details. It is only provided as a convenience
7810 for writing more maintainable configurations.
7811
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007812 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7813 unless you know what you do : memory is not shared between the
7814 processes, which can result in random behaviours.
7815
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007816 Examples :
7817 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +01007818 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007819
7820 # ...is strictly equivalent to this one :
7821 stick match src table pop if !localhost
7822 stick store-request src table pop if !localhost
7823
7824
7825 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
7826 # well as HTTP without cookie. Share the same table between both accesses.
7827 backend http
7828 mode http
7829 balance roundrobin
7830 stick on src table https
7831 cookie SRV insert indirect nocache
7832 server s1 192.168.1.1:80 cookie s1
7833 server s2 192.168.1.1:80 cookie s2
7834
7835 backend https
7836 mode tcp
7837 balance roundrobin
7838 stick-table type ip size 200k expire 30m
7839 stick on src
7840 server s1 192.168.1.1:443
7841 server s2 192.168.1.1:443
7842
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007843 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007844
7845
7846stick store-request <pattern> [table <table>] [{if | unless} <condition>]
7847 Define a request pattern used to create an entry in a stickiness table
7848 May be used in sections : defaults | frontend | listen | backend
7849 no | no | yes | yes
7850
7851 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02007852 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007853 describes what elements of the incoming request or connection
7854 will be analysed, extracted and stored in the table once a
7855 server is selected.
7856
7857 <table> is an optional stickiness table name. If unspecified, the same
7858 backend's table is used. A stickiness table is declared using
7859 the "stick-table" statement.
7860
7861 <cond> is an optional storage condition. It makes it possible to store
7862 certain criteria only when some conditions are met (or not met).
7863 For instance, it could be used to store the source IP address
7864 except when the request passes through a known proxy, in which
7865 case we'd store a converted form of a header containing that IP
7866 address.
7867
7868 Some protocols or applications require complex stickiness rules and cannot
7869 always simply rely on cookies nor hashing. The "stick store-request" statement
7870 describes a rule to decide what to extract from the request and when to do
7871 it, in order to store it into a stickiness table for further requests to
7872 match it using the "stick match" statement. Obviously the extracted part must
7873 make sense and have a chance to be matched in a further request. Storing a
7874 client's IP address for instance often makes sense. Storing an ID found in a
7875 URL parameter also makes sense. Storing a source port will almost never make
7876 any sense because it will be randomly matched. See section 7 for a complete
7877 list of possible patterns and transformation rules.
7878
7879 The table has to be declared using the "stick-table" statement. It must be of
7880 a type compatible with the pattern. By default it is the one which is present
7881 in the same backend. It is possible to share a table with other backends by
7882 referencing it using the "table" keyword. If another table is referenced,
7883 the server's ID inside the backends are used. By default, all server IDs
7884 start at 1 in each backend, so the server ordering is enough. But in case of
7885 doubt, it is highly recommended to force server IDs using their "id" setting.
7886
7887 It is possible to restrict the conditions where a "stick store-request"
7888 statement will apply, using "if" or "unless" followed by a condition. This
7889 condition will be evaluated while parsing the request, so any criteria can be
7890 used. See section 7 for ACL based conditions.
7891
7892 There is no limit on the number of "stick store-request" statements, but
7893 there is a limit of 8 simultaneous stores per request or response. This
7894 makes it possible to store up to 8 criteria, all extracted from either the
7895 request or the response, regardless of the number of rules. Only the 8 first
7896 ones which match will be kept. Using this, it is possible to feed multiple
7897 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01007898 another protocol or access method. Using multiple store-request rules with
7899 the same table is possible and may be used to find the best criterion to rely
7900 on, by arranging the rules by decreasing preference order. Only the first
7901 extracted criterion for a given table will be stored. All subsequent store-
7902 request rules referencing the same table will be skipped and their ACLs will
7903 not be evaluated.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007904
7905 The "store-request" rules are evaluated once the server connection has been
7906 established, so that the table will contain the real server that processed
7907 the request.
7908
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007909 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7910 unless you know what you do : memory is not shared between the
7911 processes, which can result in random behaviours.
7912
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007913 Example :
7914 # forward SMTP users to the same server they just used for POP in the
7915 # last 30 minutes
7916 backend pop
7917 mode tcp
7918 balance roundrobin
7919 stick store-request src
7920 stick-table type ip size 200k expire 30m
7921 server s1 192.168.1.1:110
7922 server s2 192.168.1.1:110
7923
7924 backend smtp
7925 mode tcp
7926 balance roundrobin
7927 stick match src table pop
7928 server s1 192.168.1.1:25
7929 server s2 192.168.1.1:25
7930
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007931 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02007932 about ACLs and sample fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007933
7934
Emeric Brun7c6b82e2010-09-24 16:34:28 +02007935stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Emeric Brunf099e792010-09-27 12:05:28 +02007936 size <size> [expire <expire>] [nopurge] [peers <peersect>]
7937 [store <data_type>]*
Godbach64cef792013-12-04 16:08:22 +08007938 Configure the stickiness table for the current section
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007939 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +02007940 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007941
7942 Arguments :
7943 ip a table declared with "type ip" will only store IPv4 addresses.
7944 This form is very compact (about 50 bytes per entry) and allows
7945 very fast entry lookup and stores with almost no overhead. This
7946 is mainly used to store client source IP addresses.
7947
David du Colombier9a6d3c92011-03-17 10:40:24 +01007948 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
7949 This form is very compact (about 60 bytes per entry) and allows
7950 very fast entry lookup and stores with almost no overhead. This
7951 is mainly used to store client source IP addresses.
7952
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007953 integer a table declared with "type integer" will store 32bit integers
7954 which can represent a client identifier found in a request for
7955 instance.
7956
7957 string a table declared with "type string" will store substrings of up
7958 to <len> characters. If the string provided by the pattern
7959 extractor is larger than <len>, it will be truncated before
7960 being stored. During matching, at most <len> characters will be
7961 compared between the string in the table and the extracted
7962 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +02007963 to 32 characters.
7964
7965 binary a table declared with "type binary" will store binary blocks
7966 of <len> bytes. If the block provided by the pattern
7967 extractor is larger than <len>, it will be truncated before
Willy Tarreaube722a22014-06-13 16:31:59 +02007968 being stored. If the block provided by the sample expression
Emeric Brun7c6b82e2010-09-24 16:34:28 +02007969 is shorter than <len>, it will be padded by 0. When not
7970 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007971
7972 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +02007973 "string" type table (See type "string" above). Or the number
7974 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007975 changing this parameter as memory usage will proportionally
7976 increase.
7977
7978 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +01007979 value directly impacts memory usage. Count approximately
7980 50 bytes per entry, plus the size of a string if any. The size
7981 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007982
7983 [nopurge] indicates that we refuse to purge older entries when the table
7984 is full. When not specified and the table is full when haproxy
7985 wants to store an entry in it, it will flush a few of the oldest
7986 entries in order to release some space for the new ones. This is
7987 most often the desired behaviour. In some specific cases, it
7988 be desirable to refuse new entries instead of purging the older
7989 ones. That may be the case when the amount of data to store is
7990 far above the hardware limits and we prefer not to offer access
7991 to new clients than to reject the ones already connected. When
7992 using this parameter, be sure to properly set the "expire"
7993 parameter (see below).
7994
Emeric Brunf099e792010-09-27 12:05:28 +02007995 <peersect> is the name of the peers section to use for replication. Entries
7996 which associate keys to server IDs are kept synchronized with
7997 the remote peers declared in this section. All entries are also
7998 automatically learned from the local peer (old process) during a
7999 soft restart.
8000
Willy Tarreau1abc6732015-05-01 19:21:02 +02008001 NOTE : each peers section may be referenced only by tables
8002 belonging to the same unique process.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008003
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008004 <expire> defines the maximum duration of an entry in the table since it
8005 was last created, refreshed or matched. The expiration delay is
8006 defined using the standard time format, similarly as the various
8007 timeouts. The maximum duration is slightly above 24 days. See
8008 section 2.2 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +02008009 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008010 be removed once full. Be sure not to use the "nopurge" parameter
8011 if not expiration delay is specified.
8012
Willy Tarreau08d5f982010-06-06 13:34:54 +02008013 <data_type> is used to store additional information in the stick-table. This
8014 may be used by ACLs in order to control various criteria related
8015 to the activity of the client matching the stick-table. For each
8016 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +02008017 that the additional data can fit. Several data types may be
8018 stored with an entry. Multiple data types may be specified after
8019 the "store" keyword, as a comma-separated list. Alternatively,
8020 it is possible to repeat the "store" keyword followed by one or
8021 several data types. Except for the "server_id" type which is
8022 automatically detected and enabled, all data types must be
8023 explicitly declared to be stored. If an ACL references a data
8024 type which is not stored, the ACL will simply not match. Some
8025 data types require an argument which must be passed just after
8026 the type between parenthesis. See below for the supported data
8027 types and their arguments.
8028
8029 The data types that can be stored with an entry are the following :
8030 - server_id : this is an integer which holds the numeric ID of the server a
8031 request was assigned to. It is used by the "stick match", "stick store",
8032 and "stick on" rules. It is automatically enabled when referenced.
8033
8034 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
8035 integer which may be used for anything. Most of the time it will be used
8036 to put a special tag on some entries, for instance to note that a
8037 specific behaviour was detected and must be known for future matches.
8038
Willy Tarreauba2ffd12013-05-29 15:54:14 +02008039 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
8040 over a period. It is a positive 32-bit integer integer which may be used
8041 for anything. Just like <gpc0>, it counts events, but instead of keeping
8042 a cumulative count, it maintains the rate at which the counter is
8043 incremented. Most of the time it will be used to measure the frequency of
8044 occurrence of certain events (eg: requests to a specific URL).
8045
Willy Tarreauc9705a12010-07-27 20:05:50 +02008046 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
8047 the absolute number of connections received from clients which matched
8048 this entry. It does not mean the connections were accepted, just that
8049 they were received.
8050
8051 - conn_cur : Current Connections. It is a positive 32-bit integer which
8052 stores the concurrent connection counts for the entry. It is incremented
8053 once an incoming connection matches the entry, and decremented once the
8054 connection leaves. That way it is possible to know at any time the exact
8055 number of concurrent connections for an entry.
8056
8057 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8058 integer parameter <period> which indicates in milliseconds the length
8059 of the period over which the average is measured. It reports the average
8060 incoming connection rate over that period, in connections per period. The
8061 result is an integer which can be matched using ACLs.
8062
8063 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
8064 the absolute number of sessions received from clients which matched this
8065 entry. A session is a connection that was accepted by the layer 4 rules.
8066
8067 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8068 integer parameter <period> which indicates in milliseconds the length
8069 of the period over which the average is measured. It reports the average
8070 incoming session rate over that period, in sessions per period. The
8071 result is an integer which can be matched using ACLs.
8072
8073 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
8074 counts the absolute number of HTTP requests received from clients which
8075 matched this entry. It does not matter whether they are valid requests or
8076 not. Note that this is different from sessions when keep-alive is used on
8077 the client side.
8078
8079 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8080 integer parameter <period> which indicates in milliseconds the length
8081 of the period over which the average is measured. It reports the average
8082 HTTP request rate over that period, in requests per period. The result is
8083 an integer which can be matched using ACLs. It does not matter whether
8084 they are valid requests or not. Note that this is different from sessions
8085 when keep-alive is used on the client side.
8086
8087 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
8088 counts the absolute number of HTTP requests errors induced by clients
8089 which matched this entry. Errors are counted on invalid and truncated
8090 requests, as well as on denied or tarpitted requests, and on failed
8091 authentications. If the server responds with 4xx, then the request is
8092 also counted as an error since it's an error triggered by the client
8093 (eg: vulnerability scan).
8094
8095 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8096 integer parameter <period> which indicates in milliseconds the length
8097 of the period over which the average is measured. It reports the average
8098 HTTP request error rate over that period, in requests per period (see
8099 http_err_cnt above for what is accounted as an error). The result is an
8100 integer which can be matched using ACLs.
8101
8102 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
8103 integer which counts the cumulated amount of bytes received from clients
8104 which matched this entry. Headers are included in the count. This may be
8105 used to limit abuse of upload features on photo or video servers.
8106
8107 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8108 integer parameter <period> which indicates in milliseconds the length
8109 of the period over which the average is measured. It reports the average
8110 incoming bytes rate over that period, in bytes per period. It may be used
8111 to detect users which upload too much and too fast. Warning: with large
8112 uploads, it is possible that the amount of uploaded data will be counted
8113 once upon termination, thus causing spikes in the average transfer speed
8114 instead of having a smooth one. This may partially be smoothed with
8115 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
8116 recommended for better fairness.
8117
8118 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
8119 integer which counts the cumulated amount of bytes sent to clients which
8120 matched this entry. Headers are included in the count. This may be used
8121 to limit abuse of bots sucking the whole site.
8122
8123 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
8124 an integer parameter <period> which indicates in milliseconds the length
8125 of the period over which the average is measured. It reports the average
8126 outgoing bytes rate over that period, in bytes per period. It may be used
8127 to detect users which download too much and too fast. Warning: with large
8128 transfers, it is possible that the amount of transferred data will be
8129 counted once upon termination, thus causing spikes in the average
8130 transfer speed instead of having a smooth one. This may partially be
8131 smoothed with "option contstats" though this is not perfect yet. Use of
8132 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +02008133
Willy Tarreauc00cdc22010-06-06 16:48:26 +02008134 There is only one stick-table per proxy. At the moment of writing this doc,
8135 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008136 to be required, simply create a dummy backend with a stick-table in it and
8137 reference it.
8138
8139 It is important to understand that stickiness based on learning information
8140 has some limitations, including the fact that all learned associations are
8141 lost upon restart. In general it can be good as a complement but not always
8142 as an exclusive stickiness.
8143
Willy Tarreauc9705a12010-07-27 20:05:50 +02008144 Last, memory requirements may be important when storing many data types.
8145 Indeed, storing all indicators above at once in each entry requires 116 bytes
8146 per entry, or 116 MB for a 1-million entries table. This is definitely not
8147 something that can be ignored.
8148
8149 Example:
8150 # Keep track of counters of up to 1 million IP addresses over 5 minutes
8151 # and store a general purpose counter and the average connection rate
8152 # computed over a sliding window of 30 seconds.
8153 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
8154
8155 See also : "stick match", "stick on", "stick store-request", section 2.2
David du Colombiera13d1b92011-03-17 10:40:22 +01008156 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008157
8158
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008159stick store-response <pattern> [table <table>] [{if | unless} <condition>]
8160 Define a request pattern used to create an entry in a stickiness table
8161 May be used in sections : defaults | frontend | listen | backend
8162 no | no | yes | yes
8163
8164 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008165 <pattern> is a sample expression rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008166 describes what elements of the response or connection will
8167 be analysed, extracted and stored in the table once a
8168 server is selected.
8169
8170 <table> is an optional stickiness table name. If unspecified, the same
8171 backend's table is used. A stickiness table is declared using
8172 the "stick-table" statement.
8173
8174 <cond> is an optional storage condition. It makes it possible to store
8175 certain criteria only when some conditions are met (or not met).
8176 For instance, it could be used to store the SSL session ID only
8177 when the response is a SSL server hello.
8178
8179 Some protocols or applications require complex stickiness rules and cannot
8180 always simply rely on cookies nor hashing. The "stick store-response"
8181 statement describes a rule to decide what to extract from the response and
8182 when to do it, in order to store it into a stickiness table for further
8183 requests to match it using the "stick match" statement. Obviously the
8184 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008185 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008186 See section 7 for a complete list of possible patterns and transformation
8187 rules.
8188
8189 The table has to be declared using the "stick-table" statement. It must be of
8190 a type compatible with the pattern. By default it is the one which is present
8191 in the same backend. It is possible to share a table with other backends by
8192 referencing it using the "table" keyword. If another table is referenced,
8193 the server's ID inside the backends are used. By default, all server IDs
8194 start at 1 in each backend, so the server ordering is enough. But in case of
8195 doubt, it is highly recommended to force server IDs using their "id" setting.
8196
8197 It is possible to restrict the conditions where a "stick store-response"
8198 statement will apply, using "if" or "unless" followed by a condition. This
8199 condition will be evaluated while parsing the response, so any criteria can
8200 be used. See section 7 for ACL based conditions.
8201
8202 There is no limit on the number of "stick store-response" statements, but
8203 there is a limit of 8 simultaneous stores per request or response. This
8204 makes it possible to store up to 8 criteria, all extracted from either the
8205 request or the response, regardless of the number of rules. Only the 8 first
8206 ones which match will be kept. Using this, it is possible to feed multiple
8207 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01008208 another protocol or access method. Using multiple store-response rules with
8209 the same table is possible and may be used to find the best criterion to rely
8210 on, by arranging the rules by decreasing preference order. Only the first
8211 extracted criterion for a given table will be stored. All subsequent store-
8212 response rules referencing the same table will be skipped and their ACLs will
8213 not be evaluated. However, even if a store-request rule references a table, a
8214 store-response rule may also use the same table. This means that each table
8215 may learn exactly one element from the request and one element from the
8216 response at once.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008217
8218 The table will contain the real server that processed the request.
8219
8220 Example :
8221 # Learn SSL session ID from both request and response and create affinity.
8222 backend https
8223 mode tcp
8224 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +02008225 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008226 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008227
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008228 acl clienthello req_ssl_hello_type 1
8229 acl serverhello rep_ssl_hello_type 2
8230
8231 # use tcp content accepts to detects ssl client and server hello.
8232 tcp-request inspect-delay 5s
8233 tcp-request content accept if clienthello
8234
8235 # no timeout on response inspect delay by default.
8236 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008237
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008238 # SSL session ID (SSLID) may be present on a client or server hello.
8239 # Its length is coded on 1 byte at offset 43 and its value starts
8240 # at offset 44.
8241
8242 # Match and learn on request if client hello.
8243 stick on payload_lv(43,1) if clienthello
8244
8245 # Learn on response if server hello.
8246 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +02008247
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008248 server s1 192.168.1.1:443
8249 server s2 192.168.1.1:443
8250
8251 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
8252 extraction.
8253
8254
Willy Tarreau938c7fe2014-04-25 14:21:39 +02008255tcp-check connect [params*]
8256 Opens a new connection
8257 May be used in sections: defaults | frontend | listen | backend
8258 no | no | yes | yes
8259
8260 When an application lies on more than a single TCP port or when HAProxy
8261 load-balance many services in a single backend, it makes sense to probe all
8262 the services individually before considering a server as operational.
8263
8264 When there are no TCP port configured on the server line neither server port
8265 directive, then the 'tcp-check connect port <port>' must be the first step
8266 of the sequence.
8267
8268 In a tcp-check ruleset a 'connect' is required, it is also mandatory to start
8269 the ruleset with a 'connect' rule. Purpose is to ensure admin know what they
8270 do.
8271
8272 Parameters :
8273 They are optional and can be used to describe how HAProxy should open and
8274 use the TCP connection.
8275
8276 port if not set, check port or server port is used.
8277 It tells HAProxy where to open the connection to.
8278 <port> must be a valid TCP port source integer, from 1 to 65535.
8279
8280 send-proxy send a PROXY protocol string
8281
8282 ssl opens a ciphered connection
8283
8284 Examples:
8285 # check HTTP and HTTPs services on a server.
8286 # first open port 80 thanks to server line port directive, then
8287 # tcp-check opens port 443, ciphered and run a request on it:
8288 option tcp-check
8289 tcp-check connect
8290 tcp-check send GET\ /\ HTTP/1.0\r\n
8291 tcp-check send Host:\ haproxy.1wt.eu\r\n
8292 tcp-check send \r\n
8293 tcp-check expect rstring (2..|3..)
8294 tcp-check connect port 443 ssl
8295 tcp-check send GET\ /\ HTTP/1.0\r\n
8296 tcp-check send Host:\ haproxy.1wt.eu\r\n
8297 tcp-check send \r\n
8298 tcp-check expect rstring (2..|3..)
8299 server www 10.0.0.1 check port 80
8300
8301 # check both POP and IMAP from a single server:
8302 option tcp-check
8303 tcp-check connect port 110
8304 tcp-check expect string +OK\ POP3\ ready
8305 tcp-check connect port 143
8306 tcp-check expect string *\ OK\ IMAP4\ ready
8307 server mail 10.0.0.1 check
8308
8309 See also : "option tcp-check", "tcp-check send", "tcp-check expect"
8310
8311
8312tcp-check expect [!] <match> <pattern>
8313 Specify data to be collected and analysed during a generic health check
8314 May be used in sections: defaults | frontend | listen | backend
8315 no | no | yes | yes
8316
8317 Arguments :
8318 <match> is a keyword indicating how to look for a specific pattern in the
8319 response. The keyword may be one of "string", "rstring" or
8320 binary.
8321 The keyword may be preceded by an exclamation mark ("!") to negate
8322 the match. Spaces are allowed between the exclamation mark and the
8323 keyword. See below for more details on the supported keywords.
8324
8325 <pattern> is the pattern to look for. It may be a string or a regular
8326 expression. If the pattern contains spaces, they must be escaped
8327 with the usual backslash ('\').
8328 If the match is set to binary, then the pattern must be passed as
8329 a serie of hexadecimal digits in an even number. Each sequence of
8330 two digits will represent a byte. The hexadecimal digits may be
8331 used upper or lower case.
8332
8333
8334 The available matches are intentionally similar to their http-check cousins :
8335
8336 string <string> : test the exact string matches in the response buffer.
8337 A health check response will be considered valid if the
8338 response's buffer contains this exact string. If the
8339 "string" keyword is prefixed with "!", then the response
8340 will be considered invalid if the body contains this
8341 string. This can be used to look for a mandatory pattern
8342 in a protocol response, or to detect a failure when a
8343 specific error appears in a protocol banner.
8344
8345 rstring <regex> : test a regular expression on the response buffer.
8346 A health check response will be considered valid if the
8347 response's buffer matches this expression. If the
8348 "rstring" keyword is prefixed with "!", then the response
8349 will be considered invalid if the body matches the
8350 expression.
8351
8352 binary <hexstring> : test the exact string in its hexadecimal form matches
8353 in the response buffer. A health check response will
8354 be considered valid if the response's buffer contains
8355 this exact hexadecimal string.
8356 Purpose is to match data on binary protocols.
8357
8358 It is important to note that the responses will be limited to a certain size
8359 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
8360 Thus, too large responses may not contain the mandatory pattern when using
8361 "string", "rstring" or binary. If a large response is absolutely required, it
8362 is possible to change the default max size by setting the global variable.
8363 However, it is worth keeping in mind that parsing very large responses can
8364 waste some CPU cycles, especially when regular expressions are used, and that
8365 it is always better to focus the checks on smaller resources. Also, in its
8366 current state, the check will not find any string nor regex past a null
8367 character in the response. Similarly it is not possible to request matching
8368 the null character.
8369
8370 Examples :
8371 # perform a POP check
8372 option tcp-check
8373 tcp-check expect string +OK\ POP3\ ready
8374
8375 # perform an IMAP check
8376 option tcp-check
8377 tcp-check expect string *\ OK\ IMAP4\ ready
8378
8379 # look for the redis master server
8380 option tcp-check
8381 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02008382 tcp-check expect string +PONG
Willy Tarreau938c7fe2014-04-25 14:21:39 +02008383 tcp-check send info\ replication\r\n
8384 tcp-check expect string role:master
8385 tcp-check send QUIT\r\n
8386 tcp-check expect string +OK
8387
8388
8389 See also : "option tcp-check", "tcp-check connect", "tcp-check send",
8390 "tcp-check send-binary", "http-check expect", tune.chksize
8391
8392
8393tcp-check send <data>
8394 Specify a string to be sent as a question during a generic health check
8395 May be used in sections: defaults | frontend | listen | backend
8396 no | no | yes | yes
8397
8398 <data> : the data to be sent as a question during a generic health check
8399 session. For now, <data> must be a string.
8400
8401 Examples :
8402 # look for the redis master server
8403 option tcp-check
8404 tcp-check send info\ replication\r\n
8405 tcp-check expect string role:master
8406
8407 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
8408 "tcp-check send-binary", tune.chksize
8409
8410
8411tcp-check send-binary <hexastring>
8412 Specify an hexa digits string to be sent as a binary question during a raw
8413 tcp health check
8414 May be used in sections: defaults | frontend | listen | backend
8415 no | no | yes | yes
8416
8417 <data> : the data to be sent as a question during a generic health check
8418 session. For now, <data> must be a string.
8419 <hexastring> : test the exact string in its hexadecimal form matches in the
8420 response buffer. A health check response will be considered
8421 valid if the response's buffer contains this exact
8422 hexadecimal string.
8423 Purpose is to send binary data to ask on binary protocols.
8424
8425 Examples :
8426 # redis check in binary
8427 option tcp-check
8428 tcp-check send-binary 50494e470d0a # PING\r\n
8429 tcp-check expect binary 2b504F4e47 # +PONG
8430
8431
8432 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
8433 "tcp-check send", tune.chksize
8434
8435
Willy Tarreaue9656522010-08-17 15:40:09 +02008436tcp-request connection <action> [{if | unless} <condition>]
8437 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +02008438 May be used in sections : defaults | frontend | listen | backend
8439 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +02008440 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02008441 <action> defines the action to perform if the condition applies. See
8442 below.
Willy Tarreau1a687942010-05-23 22:40:30 +02008443
Willy Tarreaue9656522010-08-17 15:40:09 +02008444 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008445
8446 Immediately after acceptance of a new incoming connection, it is possible to
8447 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +02008448 or dropped or have its counters tracked. Those conditions cannot make use of
8449 any data contents because the connection has not been read from yet, and the
8450 buffers are not yet allocated. This is used to selectively and very quickly
8451 accept or drop connections from various sources with a very low overhead. If
8452 some contents need to be inspected in order to take the decision, the
8453 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008454
Willy Tarreaue9656522010-08-17 15:40:09 +02008455 The "tcp-request connection" rules are evaluated in their exact declaration
8456 order. If no rule matches or if there is no rule, the default action is to
8457 accept the incoming connection. There is no specific limit to the number of
8458 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008459
Willy Tarreaua9083d02015-05-08 15:27:59 +02008460 Four types of actions are supported :
Willy Tarreaue9656522010-08-17 15:40:09 +02008461 - accept :
8462 accepts the connection if the condition is true (when used with "if")
8463 or false (when used with "unless"). The first such rule executed ends
8464 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008465
Willy Tarreaue9656522010-08-17 15:40:09 +02008466 - reject :
8467 rejects the connection if the condition is true (when used with "if")
8468 or false (when used with "unless"). The first such rule executed ends
8469 the rules evaluation. Rejected connections do not even become a
8470 session, which is why they are accounted separately for in the stats,
8471 as "denied connections". They are not considered for the session
8472 rate-limit and are not logged either. The reason is that these rules
8473 should only be used to filter extremely high connection rates such as
8474 the ones encountered during a massive DDoS attack. Under these extreme
8475 conditions, the simple action of logging each event would make the
8476 system collapse and would considerably lower the filtering capacity. If
8477 logging is absolutely desired, then "tcp-request content" rules should
8478 be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008479
Willy Tarreau4f0d9192013-06-11 20:40:55 +02008480 - expect-proxy layer4 :
8481 configures the client-facing connection to receive a PROXY protocol
8482 header before any byte is read from the socket. This is equivalent to
8483 having the "accept-proxy" keyword on the "bind" line, except that using
8484 the TCP rule allows the PROXY protocol to be accepted only for certain
8485 IP address ranges using an ACL. This is convenient when multiple layers
8486 of load balancers are passed through by traffic coming from public
8487 hosts.
8488
Willy Tarreau18bf01e2014-06-13 16:18:52 +02008489 - capture <sample> len <length> :
8490 This only applies to "tcp-request content" rules. It captures sample
8491 expression <sample> from the request buffer, and converts it to a
8492 string of at most <len> characters. The resulting string is stored into
8493 the next request "capture" slot, so it will possibly appear next to
8494 some captured HTTP headers. It will then automatically appear in the
8495 logs, and it will be possible to extract it using sample fetch rules to
8496 feed it into headers or anything. The length should be limited given
8497 that this size will be allocated for each capture during the whole
Willy Tarreaua9083d02015-05-08 15:27:59 +02008498 session life. Please check section 7.3 (Fetching samples) and "capture
8499 request header" for more information.
Willy Tarreau18bf01e2014-06-13 16:18:52 +02008500
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008501 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +02008502 enables tracking of sticky counters from current connection. These
Willy Tarreau09448f72014-06-25 18:12:15 +02008503 rules do not stop evaluation and do not change default action. 3 sets
Willy Tarreaue9656522010-08-17 15:40:09 +02008504 of counters may be simultaneously tracked by the same connection. The
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008505 first "track-sc0" rule executed enables tracking of the counters of the
8506 specified table as the first set. The first "track-sc1" rule executed
Willy Tarreaue9656522010-08-17 15:40:09 +02008507 enables tracking of the counters of the specified table as the second
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008508 set. The first "track-sc2" rule executed enables tracking of the
8509 counters of the specified table as the third set. It is a recommended
8510 practice to use the first set of counters for the per-frontend counters
8511 and the second set for the per-backend ones. But this is just a
8512 guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008513
Willy Tarreaue9656522010-08-17 15:40:09 +02008514 These actions take one or two arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008515 <key> is mandatory, and is a sample expression rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +02008516 in section 7.3. It describes what elements of the incoming
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008517 request or connection will be analysed, extracted, combined,
8518 and used to select which table entry to update the counters.
8519 Note that "tcp-request connection" cannot use content-based
8520 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008521
Willy Tarreaue9656522010-08-17 15:40:09 +02008522 <table> is an optional table to be used instead of the default one,
8523 which is the stick-table declared in the current proxy. All
8524 the counters for the matches and updates for the key will
8525 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008526
Willy Tarreaue9656522010-08-17 15:40:09 +02008527 Once a "track-sc*" rule is executed, the key is looked up in the table
8528 and if it is not found, an entry is allocated for it. Then a pointer to
8529 that entry is kept during all the session's life, and this entry's
8530 counters are updated as often as possible, every time the session's
8531 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008532 Counters are only updated for events that happen after the tracking has
8533 been started. For example, connection counters will not be updated when
8534 tracking layer 7 information, since the connection event happens before
8535 layer7 information is extracted.
8536
Willy Tarreaue9656522010-08-17 15:40:09 +02008537 If the entry tracks concurrent connection counters, one connection is
8538 counted for as long as the entry is tracked, and the entry will not
8539 expire during that time. Tracking counters also provides a performance
8540 advantage over just checking the keys, because only one table lookup is
8541 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008542
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02008543 - sc-inc-gpc0(<sc-id>):
8544 The "sc-inc-gpc0" increments the GPC0 counter according to the sticky
8545 counter designated by <sc-id>. If an error occurs, this action silently
8546 fails and the actions evaluation continues.
8547
Thierry FOURNIER236657b2015-08-19 08:25:14 +02008548 - sc-set-gpt0(<sc-id>) <int>:
8549 This action sets the GPT0 tag according to the sticky counter designated
8550 by <sc-id> and the value of <int>. The expected result is a boolean. If
8551 an error occurs, this action silently fails and the actions evaluation
8552 continues.
8553
Willy Tarreau2d392c22015-08-24 01:43:45 +02008554 - "silent-drop" :
8555 This stops the evaluation of the rules and makes the client-facing
8556 connection suddenly disappear using a system-dependant way that tries
8557 to prevent the client from being notified. The effect it then that the
8558 client still sees an established connection while there's none on
8559 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
8560 except that it doesn't use any local resource at all on the machine
8561 running HAProxy. It can resist much higher loads than "tarpit", and
8562 slow down stronger attackers. It is important to undestand the impact
8563 of using this mechanism. All stateful equipments placed between the
8564 client and HAProxy (firewalls, proxies, load balancers) will also keep
8565 the established connection for a long time and may suffer from this
8566 action. On modern Linux systems running with enough privileges, the
8567 TCP_REPAIR socket option is used to block the emission of a TCP
8568 reset. On other systems, the socket's TTL is reduced to 1 so that the
8569 TCP reset doesn't pass the first router, though it's still delivered to
8570 local networks. Do not use it unless you fully understand how it works.
8571
Willy Tarreaue9656522010-08-17 15:40:09 +02008572 Note that the "if/unless" condition is optional. If no condition is set on
8573 the action, it is simply performed unconditionally. That can be useful for
8574 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008575
Willy Tarreaue9656522010-08-17 15:40:09 +02008576 Example: accept all connections from white-listed hosts, reject too fast
8577 connection without counting them, and track accepted connections.
8578 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008579
Willy Tarreaue9656522010-08-17 15:40:09 +02008580 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008581 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008582 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008583
Willy Tarreaue9656522010-08-17 15:40:09 +02008584 Example: accept all connections from white-listed hosts, count all other
8585 connections and reject too fast ones. This results in abusive ones
8586 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008587
Willy Tarreaue9656522010-08-17 15:40:09 +02008588 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008589 tcp-request connection track-sc0 src
8590 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008591
Willy Tarreau4f0d9192013-06-11 20:40:55 +02008592 Example: enable the PROXY protocol for traffic coming from all known proxies.
8593
8594 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
8595
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008596 See section 7 about ACL usage.
8597
Willy Tarreaue9656522010-08-17 15:40:09 +02008598 See also : "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008599
8600
Willy Tarreaue9656522010-08-17 15:40:09 +02008601tcp-request content <action> [{if | unless} <condition>]
8602 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +02008603 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02008604 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +02008605 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02008606 <action> defines the action to perform if the condition applies. See
8607 below.
Willy Tarreau62644772008-07-16 18:36:06 +02008608
Willy Tarreaue9656522010-08-17 15:40:09 +02008609 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +02008610
Willy Tarreaue9656522010-08-17 15:40:09 +02008611 A request's contents can be analysed at an early stage of request processing
8612 called "TCP content inspection". During this stage, ACL-based rules are
8613 evaluated every time the request contents are updated, until either an
8614 "accept" or a "reject" rule matches, or the TCP request inspection delay
8615 expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +02008616
Willy Tarreaue9656522010-08-17 15:40:09 +02008617 The first difference between these rules and "tcp-request connection" rules
8618 is that "tcp-request content" rules can make use of contents to take a
8619 decision. Most often, these decisions will consider a protocol recognition or
8620 validity. The second difference is that content-based rules can be used in
Willy Tarreauf3338342014-01-28 21:40:28 +01008621 both frontends and backends. In case of HTTP keep-alive with the client, all
8622 tcp-request content rules are evaluated again, so haproxy keeps a record of
8623 what sticky counters were assigned by a "tcp-request connection" versus a
8624 "tcp-request content" rule, and flushes all the content-related ones after
8625 processing an HTTP request, so that they may be evaluated again by the rules
8626 being evaluated again for the next request. This is of particular importance
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008627 when the rule tracks some L7 information or when it is conditioned by an
Willy Tarreauf3338342014-01-28 21:40:28 +01008628 L7-based ACL, since tracking may change between requests.
Willy Tarreau62644772008-07-16 18:36:06 +02008629
Willy Tarreaue9656522010-08-17 15:40:09 +02008630 Content-based rules are evaluated in their exact declaration order. If no
8631 rule matches or if there is no rule, the default action is to accept the
8632 contents. There is no specific limit to the number of rules which may be
8633 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +02008634
Thierry FOURNIER236657b2015-08-19 08:25:14 +02008635 Several types of actions are supported :
Willy Tarreau18bf01e2014-06-13 16:18:52 +02008636 - accept : the request is accepted
8637 - reject : the request is rejected and the connection is closed
8638 - capture : the specified sample expression is captured
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008639 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02008640 - sc-inc-gpc0(<sc-id>)
Thierry FOURNIER236657b2015-08-19 08:25:14 +02008641 - set-gpt0(<sc-id>) <int>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02008642 - set-var(<var-name>) <expr>
Willy Tarreau2d392c22015-08-24 01:43:45 +02008643 - silent-drop
Willy Tarreau62644772008-07-16 18:36:06 +02008644
Willy Tarreaue9656522010-08-17 15:40:09 +02008645 They have the same meaning as their counter-parts in "tcp-request connection"
8646 so please refer to that section for a complete description.
Willy Tarreau62644772008-07-16 18:36:06 +02008647
Willy Tarreauf3338342014-01-28 21:40:28 +01008648 While there is nothing mandatory about it, it is recommended to use the
8649 track-sc0 in "tcp-request connection" rules, track-sc1 for "tcp-request
8650 content" rules in the frontend, and track-sc2 for "tcp-request content"
8651 rules in the backend, because that makes the configuration more readable
8652 and easier to troubleshoot, but this is just a guideline and all counters
8653 may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +02008654
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008655 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +02008656 the action, it is simply performed unconditionally. That can be useful for
8657 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +02008658
Willy Tarreaue9656522010-08-17 15:40:09 +02008659 It is perfectly possible to match layer 7 contents with "tcp-request content"
Willy Tarreauc0239e02012-04-16 14:42:55 +02008660 rules, since HTTP-specific ACL matches are able to preliminarily parse the
8661 contents of a buffer before extracting the required data. If the buffered
8662 contents do not parse as a valid HTTP message, then the ACL does not match.
8663 The parser which is involved there is exactly the same as for all other HTTP
Willy Tarreauf3338342014-01-28 21:40:28 +01008664 processing, so there is no risk of parsing something differently. In an HTTP
8665 backend connected to from an HTTP frontend, it is guaranteed that HTTP
8666 contents will always be immediately present when the rule is evaluated first.
Willy Tarreau62644772008-07-16 18:36:06 +02008667
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008668 Tracking layer7 information is also possible provided that the information
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02008669 are present when the rule is processed. The rule processing engine is able to
8670 wait until the inspect delay expires when the data to be tracked is not yet
8671 available.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008672
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02008673 The "set-var" is used to set the content of a variable. The variable is
8674 declared inline.
8675
8676 <var-name> The name of the variable starts by an indication about its scope.
8677 The allowed scopes are:
8678 "sess" : the variable is shared with all the session,
8679 "txn" : the variable is shared with all the transaction
8680 (request and response)
8681 "req" : the variable is shared only during the request
8682 processing
8683 "res" : the variable is shared only during the response
8684 processing.
8685 This prefix is followed by a name. The separator is a '.'.
8686 The name may only contain characters 'a-z', 'A-Z', '0-9' and '_'.
8687
8688 <expr> Is a standard HAProxy expression formed by a sample-fetch
8689 followed by some converters.
8690
8691 Example:
8692
8693 tcp-request content set-var(sess.my_var) src
8694
Willy Tarreau62644772008-07-16 18:36:06 +02008695 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +02008696 # Accept HTTP requests containing a Host header saying "example.com"
8697 # and reject everything else.
8698 acl is_host_com hdr(Host) -i example.com
8699 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +02008700 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +02008701 tcp-request content reject
8702
8703 Example:
Willy Tarreau62644772008-07-16 18:36:06 +02008704 # reject SMTP connection if client speaks first
8705 tcp-request inspect-delay 30s
8706 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008707 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +02008708
8709 # Forward HTTPS connection only if client speaks
8710 tcp-request inspect-delay 30s
8711 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008712 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +02008713 tcp-request content reject
8714
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008715 Example:
8716 # Track the last IP from X-Forwarded-For
8717 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02008718 tcp-request content track-sc0 hdr(x-forwarded-for,-1)
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008719
8720 Example:
8721 # track request counts per "base" (concatenation of Host+URL)
8722 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02008723 tcp-request content track-sc0 base table req-rate
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008724
Willy Tarreaue9656522010-08-17 15:40:09 +02008725 Example: track per-frontend and per-backend counters, block abusers at the
8726 frontend when the backend detects abuse.
8727
8728 frontend http
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008729 # Use General Purpose Couter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +02008730 # protecting all our sites
8731 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008732 tcp-request connection track-sc0 src
8733 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +02008734 ...
8735 use_backend http_dynamic if { path_end .php }
8736
8737 backend http_dynamic
8738 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008739 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +02008740 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008741 acl click_too_fast sc1_http_req_rate gt 10
8742 acl mark_as_abuser sc0_inc_gpc0 gt 0
8743 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +02008744 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +02008745
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008746 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +02008747
Willy Tarreaue9656522010-08-17 15:40:09 +02008748 See also : "tcp-request connection", "tcp-request inspect-delay"
Willy Tarreau62644772008-07-16 18:36:06 +02008749
8750
8751tcp-request inspect-delay <timeout>
8752 Set the maximum allowed time to wait for data during content inspection
8753 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02008754 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +02008755 Arguments :
8756 <timeout> is the timeout value specified in milliseconds by default, but
8757 can be in any other unit if the number is suffixed by the unit,
8758 as explained at the top of this document.
8759
8760 People using haproxy primarily as a TCP relay are often worried about the
8761 risk of passing any type of protocol to a server without any analysis. In
8762 order to be able to analyze the request contents, we must first withhold
8763 the data then analyze them. This statement simply enables withholding of
8764 data for at most the specified amount of time.
8765
Willy Tarreaufb356202010-08-03 14:02:05 +02008766 TCP content inspection applies very early when a connection reaches a
8767 frontend, then very early when the connection is forwarded to a backend. This
8768 means that a connection may experience a first delay in the frontend and a
8769 second delay in the backend if both have tcp-request rules.
8770
Willy Tarreau62644772008-07-16 18:36:06 +02008771 Note that when performing content inspection, haproxy will evaluate the whole
8772 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008773 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +02008774 a last check is performed upon expiration, this time considering that the
Willy Tarreaud869b242009-03-15 14:43:58 +01008775 contents are definitive. If no delay is set, haproxy will not wait at all
8776 and will immediately apply a verdict based on the available information.
8777 Obviously this is unlikely to be very useful and might even be racy, so such
8778 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +02008779
8780 As soon as a rule matches, the request is released and continues as usual. If
8781 the timeout is reached and no rule matches, the default policy will be to let
8782 it pass through unaffected.
8783
8784 For most protocols, it is enough to set it to a few seconds, as most clients
8785 send the full request immediately upon connection. Add 3 or more seconds to
8786 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +01008787 to use large values, for instance to ensure that the client never talks
Willy Tarreau62644772008-07-16 18:36:06 +02008788 before the server (eg: SMTP), or to wait for a client to talk before passing
8789 data to the server (eg: SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +02008790 least the inspection delay, otherwise it will expire first. If the client
8791 closes the connection or if the buffer is full, the delay immediately expires
8792 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +02008793
Willy Tarreau55165fe2009-05-10 12:02:55 +02008794 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +02008795 "timeout client".
8796
8797
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008798tcp-response content <action> [{if | unless} <condition>]
8799 Perform an action on a session response depending on a layer 4-7 condition
8800 May be used in sections : defaults | frontend | listen | backend
8801 no | no | yes | yes
8802 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02008803 <action> defines the action to perform if the condition applies. See
8804 below.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008805
8806 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
8807
8808 Response contents can be analysed at an early stage of response processing
8809 called "TCP content inspection". During this stage, ACL-based rules are
8810 evaluated every time the response contents are updated, until either an
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02008811 "accept", "close" or a "reject" rule matches, or a TCP response inspection
8812 delay is set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008813
8814 Most often, these decisions will consider a protocol recognition or validity.
8815
8816 Content-based rules are evaluated in their exact declaration order. If no
8817 rule matches or if there is no rule, the default action is to accept the
8818 contents. There is no specific limit to the number of rules which may be
8819 inserted.
8820
Thierry FOURNIER236657b2015-08-19 08:25:14 +02008821 Several types of actions are supported :
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008822 - accept :
8823 accepts the response if the condition is true (when used with "if")
8824 or false (when used with "unless"). The first such rule executed ends
8825 the rules evaluation.
8826
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02008827 - close :
8828 immediately closes the connection with the server if the condition is
8829 true (when used with "if"), or false (when used with "unless"). The
8830 first such rule executed ends the rules evaluation. The main purpose of
8831 this action is to force a connection to be finished between a client
8832 and a server after an exchange when the application protocol expects
8833 some long time outs to elapse first. The goal is to eliminate idle
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008834 connections which take significant resources on servers with certain
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02008835 protocols.
8836
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008837 - reject :
8838 rejects the response if the condition is true (when used with "if")
8839 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -04008840 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008841
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02008842 - set-var(<var-name>) <expr>
8843 Sets a variable.
8844
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02008845 - sc-inc-gpc0(<sc-id>):
8846 This action increments the GPC0 counter according to the sticky
8847 counter designated by <sc-id>. If an error occurs, this action fails
8848 silently and the actions evaluation continues.
8849
Thierry FOURNIER236657b2015-08-19 08:25:14 +02008850 - sc-set-gpt0(<sc-id>) <int> :
8851 This action sets the GPT0 tag according to the sticky counter designated
8852 by <sc-id> and the value of <int>. The expected result is a boolean. If
8853 an error occurs, this action silently fails and the actions evaluation
8854 continues.
8855
Willy Tarreau2d392c22015-08-24 01:43:45 +02008856 - "silent-drop" :
8857 This stops the evaluation of the rules and makes the client-facing
8858 connection suddenly disappear using a system-dependant way that tries
8859 to prevent the client from being notified. The effect it then that the
8860 client still sees an established connection while there's none on
8861 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
8862 except that it doesn't use any local resource at all on the machine
8863 running HAProxy. It can resist much higher loads than "tarpit", and
8864 slow down stronger attackers. It is important to undestand the impact
8865 of using this mechanism. All stateful equipments placed between the
8866 client and HAProxy (firewalls, proxies, load balancers) will also keep
8867 the established connection for a long time and may suffer from this
8868 action. On modern Linux systems running with enough privileges, the
8869 TCP_REPAIR socket option is used to block the emission of a TCP
8870 reset. On other systems, the socket's TTL is reduced to 1 so that the
8871 TCP reset doesn't pass the first router, though it's still delivered to
8872 local networks. Do not use it unless you fully understand how it works.
8873
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008874 Note that the "if/unless" condition is optional. If no condition is set on
8875 the action, it is simply performed unconditionally. That can be useful for
8876 for changing the default action to a reject.
8877
Jamie Gloudonaaa21002012-08-25 00:18:33 -04008878 It is perfectly possible to match layer 7 contents with "tcp-response
8879 content" rules, but then it is important to ensure that a full response has
8880 been buffered, otherwise no contents will match. In order to achieve this,
8881 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008882 period.
8883
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02008884 The "set-var" is used to set the content of a variable. The variable is
8885 declared inline.
8886
8887 <var-name> The name of the variable starts by an indication about its scope.
8888 The allowed scopes are:
8889 "sess" : the variable is shared with all the session,
8890 "txn" : the variable is shared with all the transaction
8891 (request and response)
8892 "req" : the variable is shared only during the request
8893 processing
8894 "res" : the variable is shared only during the response
8895 processing.
8896 This prefix is followed by a name. The separator is a '.'.
8897 The name may only contain characters 'a-z', 'A-Z', '0-9' and '_'.
8898
8899 <expr> Is a standard HAProxy expression formed by a sample-fetch
8900 followed by some converters.
8901
8902 Example:
8903
8904 tcp-request content set-var(sess.my_var) src
8905
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008906 See section 7 about ACL usage.
8907
8908 See also : "tcp-request content", "tcp-response inspect-delay"
8909
8910
8911tcp-response inspect-delay <timeout>
8912 Set the maximum allowed time to wait for a response during content inspection
8913 May be used in sections : defaults | frontend | listen | backend
8914 no | no | yes | yes
8915 Arguments :
8916 <timeout> is the timeout value specified in milliseconds by default, but
8917 can be in any other unit if the number is suffixed by the unit,
8918 as explained at the top of this document.
8919
8920 See also : "tcp-response content", "tcp-request inspect-delay".
8921
8922
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01008923timeout check <timeout>
8924 Set additional check timeout, but only after a connection has been already
8925 established.
8926
8927 May be used in sections: defaults | frontend | listen | backend
8928 yes | no | yes | yes
8929 Arguments:
8930 <timeout> is the timeout value specified in milliseconds by default, but
8931 can be in any other unit if the number is suffixed by the unit,
8932 as explained at the top of this document.
8933
8934 If set, haproxy uses min("timeout connect", "inter") as a connect timeout
8935 for check and "timeout check" as an additional read timeout. The "min" is
8936 used so that people running with *very* long "timeout connect" (eg. those
8937 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +01008938 (Please also note that there is no valid reason to have such long connect
8939 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
8940 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01008941
8942 If "timeout check" is not set haproxy uses "inter" for complete check
8943 timeout (connect + read) exactly like all <1.3.15 version.
8944
8945 In most cases check request is much simpler and faster to handle than normal
8946 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +01008947 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01008948
8949 This parameter is specific to backends, but can be specified once for all in
8950 "defaults" sections. This is in fact one of the easiest solutions not to
8951 forget about it.
8952
Willy Tarreau41a340d2008-01-22 12:25:31 +01008953 See also: "timeout connect", "timeout queue", "timeout server",
8954 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01008955
8956
Willy Tarreau0ba27502007-12-24 16:55:16 +01008957timeout client <timeout>
8958timeout clitimeout <timeout> (deprecated)
8959 Set the maximum inactivity time on the client side.
8960 May be used in sections : defaults | frontend | listen | backend
8961 yes | yes | yes | no
8962 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01008963 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01008964 can be in any other unit if the number is suffixed by the unit,
8965 as explained at the top of this document.
8966
8967 The inactivity timeout applies when the client is expected to acknowledge or
8968 send data. In HTTP mode, this timeout is particularly important to consider
8969 during the first phase, when the client sends the request, and during the
8970 response while it is reading data sent by the server. The value is specified
8971 in milliseconds by default, but can be in any other unit if the number is
8972 suffixed by the unit, as specified at the top of this document. In TCP mode
8973 (and to a lesser extent, in HTTP mode), it is highly recommended that the
8974 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01008975 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01008976 losses by specifying timeouts that are slightly above multiples of 3 seconds
Willy Tarreauce887fd2012-05-12 12:50:00 +02008977 (eg: 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
8978 sessions (eg: WebSocket and HTTP), it's worth considering "timeout tunnel",
Willy Tarreau05cdd962014-05-10 14:30:07 +02008979 which overrides "timeout client" and "timeout server" for tunnels, as well as
8980 "timeout client-fin" for half-closed connections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01008981
8982 This parameter is specific to frontends, but can be specified once for all in
8983 "defaults" sections. This is in fact one of the easiest solutions not to
8984 forget about it. An unspecified timeout results in an infinite timeout, which
8985 is not recommended. Such a usage is accepted and works but reports a warning
8986 during startup because it may results in accumulation of expired sessions in
8987 the system if the system's timeouts are not configured either.
8988
8989 This parameter replaces the old, deprecated "clitimeout". It is recommended
8990 to use it to write new configurations. The form "timeout clitimeout" is
8991 provided only by backwards compatibility but its use is strongly discouraged.
8992
Willy Tarreauce887fd2012-05-12 12:50:00 +02008993 See also : "clitimeout", "timeout server", "timeout tunnel".
Willy Tarreau0ba27502007-12-24 16:55:16 +01008994
8995
Willy Tarreau05cdd962014-05-10 14:30:07 +02008996timeout client-fin <timeout>
8997 Set the inactivity timeout on the client side for half-closed connections.
8998 May be used in sections : defaults | frontend | listen | backend
8999 yes | yes | yes | no
9000 Arguments :
9001 <timeout> is the timeout value specified in milliseconds by default, but
9002 can be in any other unit if the number is suffixed by the unit,
9003 as explained at the top of this document.
9004
9005 The inactivity timeout applies when the client is expected to acknowledge or
9006 send data while one direction is already shut down. This timeout is different
9007 from "timeout client" in that it only applies to connections which are closed
9008 in one direction. This is particularly useful to avoid keeping connections in
9009 FIN_WAIT state for too long when clients do not disconnect cleanly. This
9010 problem is particularly common long connections such as RDP or WebSocket.
9011 Note that this timeout can override "timeout tunnel" when a connection shuts
9012 down in one direction.
9013
9014 This parameter is specific to frontends, but can be specified once for all in
9015 "defaults" sections. By default it is not set, so half-closed connections
9016 will use the other timeouts (timeout.client or timeout.tunnel).
9017
9018 See also : "timeout client", "timeout server-fin", and "timeout tunnel".
9019
9020
Willy Tarreau0ba27502007-12-24 16:55:16 +01009021timeout connect <timeout>
9022timeout contimeout <timeout> (deprecated)
9023 Set the maximum time to wait for a connection attempt to a server to succeed.
9024 May be used in sections : defaults | frontend | listen | backend
9025 yes | no | yes | yes
9026 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01009027 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01009028 can be in any other unit if the number is suffixed by the unit,
9029 as explained at the top of this document.
9030
9031 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01009032 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01009033 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01009034 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009035 connect timeout also presets both queue and tarpit timeouts to the same value
9036 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +01009037
9038 This parameter is specific to backends, but can be specified once for all in
9039 "defaults" sections. This is in fact one of the easiest solutions not to
9040 forget about it. An unspecified timeout results in an infinite timeout, which
9041 is not recommended. Such a usage is accepted and works but reports a warning
9042 during startup because it may results in accumulation of failed sessions in
9043 the system if the system's timeouts are not configured either.
9044
9045 This parameter replaces the old, deprecated "contimeout". It is recommended
9046 to use it to write new configurations. The form "timeout contimeout" is
9047 provided only by backwards compatibility but its use is strongly discouraged.
9048
Willy Tarreau41a340d2008-01-22 12:25:31 +01009049 See also: "timeout check", "timeout queue", "timeout server", "contimeout",
9050 "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +01009051
9052
Willy Tarreaub16a5742010-01-10 14:46:16 +01009053timeout http-keep-alive <timeout>
9054 Set the maximum allowed time to wait for a new HTTP request to appear
9055 May be used in sections : defaults | frontend | listen | backend
9056 yes | yes | yes | yes
9057 Arguments :
9058 <timeout> is the timeout value specified in milliseconds by default, but
9059 can be in any other unit if the number is suffixed by the unit,
9060 as explained at the top of this document.
9061
9062 By default, the time to wait for a new request in case of keep-alive is set
9063 by "timeout http-request". However this is not always convenient because some
9064 people want very short keep-alive timeouts in order to release connections
9065 faster, and others prefer to have larger ones but still have short timeouts
9066 once the request has started to present itself.
9067
9068 The "http-keep-alive" timeout covers these needs. It will define how long to
9069 wait for a new HTTP request to start coming after a response was sent. Once
9070 the first byte of request has been seen, the "http-request" timeout is used
9071 to wait for the complete request to come. Note that empty lines prior to a
9072 new request do not refresh the timeout and are not counted as a new request.
9073
9074 There is also another difference between the two timeouts : when a connection
9075 expires during timeout http-keep-alive, no error is returned, the connection
9076 just closes. If the connection expires in "http-request" while waiting for a
9077 connection to complete, a HTTP 408 error is returned.
9078
9079 In general it is optimal to set this value to a few tens to hundreds of
9080 milliseconds, to allow users to fetch all objects of a page at once but
9081 without waiting for further clicks. Also, if set to a very small value (eg:
9082 1 millisecond) it will probably only accept pipelined requests but not the
9083 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +02009084 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +01009085
9086 If this parameter is not set, the "http-request" timeout applies, and if both
9087 are not set, "timeout client" still applies at the lower level. It should be
9088 set in the frontend to take effect, unless the frontend is in TCP mode, in
9089 which case the HTTP backend's timeout will be used.
9090
9091 See also : "timeout http-request", "timeout client".
9092
9093
Willy Tarreau036fae02008-01-06 13:24:40 +01009094timeout http-request <timeout>
9095 Set the maximum allowed time to wait for a complete HTTP request
9096 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +02009097 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +01009098 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01009099 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +01009100 can be in any other unit if the number is suffixed by the unit,
9101 as explained at the top of this document.
9102
9103 In order to offer DoS protection, it may be required to lower the maximum
9104 accepted time to receive a complete HTTP request without affecting the client
9105 timeout. This helps protecting against established connections on which
9106 nothing is sent. The client timeout cannot offer a good protection against
9107 this abuse because it is an inactivity timeout, which means that if the
9108 attacker sends one character every now and then, the timeout will not
9109 trigger. With the HTTP request timeout, no matter what speed the client
Willy Tarreau2705a612014-05-23 17:38:34 +02009110 types, the request will be aborted if it does not complete in time. When the
9111 timeout expires, an HTTP 408 response is sent to the client to inform it
9112 about the problem, and the connection is closed. The logs will report
9113 termination codes "cR". Some recent browsers are having problems with this
9114 standard, well-documented behaviour, so it might be needed to hide the 408
Willy Tarreau0f228a02015-05-01 15:37:53 +02009115 code using "option http-ignore-probes" or "errorfile 408 /dev/null". See
9116 more details in the explanations of the "cR" termination code in section 8.5.
Willy Tarreau036fae02008-01-06 13:24:40 +01009117
9118 Note that this timeout only applies to the header part of the request, and
9119 not to any data. As soon as the empty line is received, this timeout is not
Willy Tarreaub16a5742010-01-10 14:46:16 +01009120 used anymore. It is used again on keep-alive connections to wait for a second
9121 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +01009122
9123 Generally it is enough to set it to a few seconds, as most clients send the
9124 full request immediately upon connection. Add 3 or more seconds to cover TCP
9125 retransmits but that's all. Setting it to very low values (eg: 50 ms) will
9126 generally work on local networks as long as there are no packet losses. This
9127 will prevent people from sending bare HTTP requests using telnet.
9128
9129 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +02009130 chunk of the incoming request. It should be set in the frontend to take
9131 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
9132 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +01009133
Willy Tarreau0f228a02015-05-01 15:37:53 +02009134 See also : "errorfile", "http-ignore-probes", "timeout http-keep-alive", and
9135 "timeout client".
Willy Tarreau036fae02008-01-06 13:24:40 +01009136
Willy Tarreau844e3c52008-01-11 16:28:18 +01009137
9138timeout queue <timeout>
9139 Set the maximum time to wait in the queue for a connection slot to be free
9140 May be used in sections : defaults | frontend | listen | backend
9141 yes | no | yes | yes
9142 Arguments :
9143 <timeout> is the timeout value specified in milliseconds by default, but
9144 can be in any other unit if the number is suffixed by the unit,
9145 as explained at the top of this document.
9146
9147 When a server's maxconn is reached, connections are left pending in a queue
9148 which may be server-specific or global to the backend. In order not to wait
9149 indefinitely, a timeout is applied to requests pending in the queue. If the
9150 timeout is reached, it is considered that the request will almost never be
9151 served, so it is dropped and a 503 error is returned to the client.
9152
9153 The "timeout queue" statement allows to fix the maximum time for a request to
9154 be left pending in a queue. If unspecified, the same value as the backend's
9155 connection timeout ("timeout connect") is used, for backwards compatibility
9156 with older versions with no "timeout queue" parameter.
9157
9158 See also : "timeout connect", "contimeout".
9159
9160
9161timeout server <timeout>
9162timeout srvtimeout <timeout> (deprecated)
9163 Set the maximum inactivity time on the server side.
9164 May be used in sections : defaults | frontend | listen | backend
9165 yes | no | yes | yes
9166 Arguments :
9167 <timeout> is the timeout value specified in milliseconds by default, but
9168 can be in any other unit if the number is suffixed by the unit,
9169 as explained at the top of this document.
9170
9171 The inactivity timeout applies when the server is expected to acknowledge or
9172 send data. In HTTP mode, this timeout is particularly important to consider
9173 during the first phase of the server's response, when it has to send the
9174 headers, as it directly represents the server's processing time for the
9175 request. To find out what value to put there, it's often good to start with
9176 what would be considered as unacceptable response times, then check the logs
9177 to observe the response time distribution, and adjust the value accordingly.
9178
9179 The value is specified in milliseconds by default, but can be in any other
9180 unit if the number is suffixed by the unit, as specified at the top of this
9181 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
9182 recommended that the client timeout remains equal to the server timeout in
9183 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01009184 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01009185 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreauce887fd2012-05-12 12:50:00 +02009186 seconds (eg: 4 or 5 seconds minimum). If some long-lived sessions are mixed
9187 with short-lived sessions (eg: WebSocket and HTTP), it's worth considering
9188 "timeout tunnel", which overrides "timeout client" and "timeout server" for
9189 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009190
9191 This parameter is specific to backends, but can be specified once for all in
9192 "defaults" sections. This is in fact one of the easiest solutions not to
9193 forget about it. An unspecified timeout results in an infinite timeout, which
9194 is not recommended. Such a usage is accepted and works but reports a warning
9195 during startup because it may results in accumulation of expired sessions in
9196 the system if the system's timeouts are not configured either.
9197
9198 This parameter replaces the old, deprecated "srvtimeout". It is recommended
9199 to use it to write new configurations. The form "timeout srvtimeout" is
9200 provided only by backwards compatibility but its use is strongly discouraged.
9201
Willy Tarreauce887fd2012-05-12 12:50:00 +02009202 See also : "srvtimeout", "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +01009203
Willy Tarreau05cdd962014-05-10 14:30:07 +02009204
9205timeout server-fin <timeout>
9206 Set the inactivity timeout on the server side for half-closed connections.
9207 May be used in sections : defaults | frontend | listen | backend
9208 yes | no | yes | yes
9209 Arguments :
9210 <timeout> is the timeout value specified in milliseconds by default, but
9211 can be in any other unit if the number is suffixed by the unit,
9212 as explained at the top of this document.
9213
9214 The inactivity timeout applies when the server is expected to acknowledge or
9215 send data while one direction is already shut down. This timeout is different
9216 from "timeout server" in that it only applies to connections which are closed
9217 in one direction. This is particularly useful to avoid keeping connections in
9218 FIN_WAIT state for too long when a remote server does not disconnect cleanly.
9219 This problem is particularly common long connections such as RDP or WebSocket.
9220 Note that this timeout can override "timeout tunnel" when a connection shuts
9221 down in one direction. This setting was provided for completeness, but in most
9222 situations, it should not be needed.
9223
9224 This parameter is specific to backends, but can be specified once for all in
9225 "defaults" sections. By default it is not set, so half-closed connections
9226 will use the other timeouts (timeout.server or timeout.tunnel).
9227
9228 See also : "timeout client-fin", "timeout server", and "timeout tunnel".
9229
Willy Tarreau844e3c52008-01-11 16:28:18 +01009230
9231timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +01009232 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +01009233 May be used in sections : defaults | frontend | listen | backend
9234 yes | yes | yes | yes
9235 Arguments :
9236 <timeout> is the tarpit duration specified in milliseconds by default, but
9237 can be in any other unit if the number is suffixed by the unit,
9238 as explained at the top of this document.
9239
9240 When a connection is tarpitted using "reqtarpit", it is maintained open with
9241 no activity for a certain amount of time, then closed. "timeout tarpit"
9242 defines how long it will be maintained open.
9243
9244 The value is specified in milliseconds by default, but can be in any other
9245 unit if the number is suffixed by the unit, as specified at the top of this
9246 document. If unspecified, the same value as the backend's connection timeout
9247 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +01009248 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009249
9250 See also : "timeout connect", "contimeout".
9251
9252
Willy Tarreauce887fd2012-05-12 12:50:00 +02009253timeout tunnel <timeout>
9254 Set the maximum inactivity time on the client and server side for tunnels.
9255 May be used in sections : defaults | frontend | listen | backend
9256 yes | no | yes | yes
9257 Arguments :
9258 <timeout> is the timeout value specified in milliseconds by default, but
9259 can be in any other unit if the number is suffixed by the unit,
9260 as explained at the top of this document.
9261
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009262 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +02009263 between a client and a server, and the connection remains inactive in both
9264 directions. This timeout supersedes both the client and server timeouts once
9265 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
9266 analyser remains attached to either connection (eg: tcp content rules are
9267 accepted). In HTTP, this timeout is used when a connection is upgraded (eg:
9268 when switching to the WebSocket protocol, or forwarding a CONNECT request
9269 to a proxy), or after the first response when no keepalive/close option is
9270 specified.
9271
Willy Tarreau05cdd962014-05-10 14:30:07 +02009272 Since this timeout is usually used in conjunction with long-lived connections,
9273 it usually is a good idea to also set "timeout client-fin" to handle the
9274 situation where a client suddenly disappears from the net and does not
9275 acknowledge a close, or sends a shutdown and does not acknowledge pending
9276 data anymore. This can happen in lossy networks where firewalls are present,
9277 and is detected by the presence of large amounts of sessions in a FIN_WAIT
9278 state.
9279
Willy Tarreauce887fd2012-05-12 12:50:00 +02009280 The value is specified in milliseconds by default, but can be in any other
9281 unit if the number is suffixed by the unit, as specified at the top of this
9282 document. Whatever the expected normal idle time, it is a good practice to
9283 cover at least one or several TCP packet losses by specifying timeouts that
9284 are slightly above multiples of 3 seconds (eg: 4 or 5 seconds minimum).
9285
9286 This parameter is specific to backends, but can be specified once for all in
9287 "defaults" sections. This is in fact one of the easiest solutions not to
9288 forget about it.
9289
9290 Example :
9291 defaults http
9292 option http-server-close
9293 timeout connect 5s
9294 timeout client 30s
Willy Tarreau05cdd962014-05-10 14:30:07 +02009295 timeout client-fin 30s
Willy Tarreauce887fd2012-05-12 12:50:00 +02009296 timeout server 30s
9297 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
9298
Willy Tarreau05cdd962014-05-10 14:30:07 +02009299 See also : "timeout client", "timeout client-fin", "timeout server".
Willy Tarreauce887fd2012-05-12 12:50:00 +02009300
9301
Willy Tarreau844e3c52008-01-11 16:28:18 +01009302transparent (deprecated)
9303 Enable client-side transparent proxying
9304 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01009305 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01009306 Arguments : none
9307
9308 This keyword was introduced in order to provide layer 7 persistence to layer
9309 3 load balancers. The idea is to use the OS's ability to redirect an incoming
9310 connection for a remote address to a local process (here HAProxy), and let
9311 this process know what address was initially requested. When this option is
9312 used, sessions without cookies will be forwarded to the original destination
9313 IP address of the incoming request (which should match that of another
9314 equipment), while requests with cookies will still be forwarded to the
9315 appropriate server.
9316
9317 The "transparent" keyword is deprecated, use "option transparent" instead.
9318
9319 Note that contrary to a common belief, this option does NOT make HAProxy
9320 present the client's IP to the server when establishing the connection.
9321
Willy Tarreau844e3c52008-01-11 16:28:18 +01009322 See also: "option transparent"
9323
William Lallemanda73203e2012-03-12 12:48:57 +01009324unique-id-format <string>
9325 Generate a unique ID for each request.
9326 May be used in sections : defaults | frontend | listen | backend
9327 yes | yes | yes | no
9328 Arguments :
9329 <string> is a log-format string.
9330
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009331 This keyword creates a ID for each request using the custom log format. A
9332 unique ID is useful to trace a request passing through many components of
9333 a complex infrastructure. The newly created ID may also be logged using the
9334 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +01009335
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009336 The format should be composed from elements that are guaranteed to be
9337 unique when combined together. For instance, if multiple haproxy instances
9338 are involved, it might be important to include the node name. It is often
9339 needed to log the incoming connection's source and destination addresses
9340 and ports. Note that since multiple requests may be performed over the same
9341 connection, including a request counter may help differentiate them.
9342 Similarly, a timestamp may protect against a rollover of the counter.
9343 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +01009344
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009345 It is recommended to use hexadecimal notation for many fields since it
9346 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +01009347
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009348 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01009349
Julien Vehentf21be322014-03-07 08:27:34 -05009350 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +01009351
9352 will generate:
9353
9354 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
9355
9356 See also: "unique-id-header"
9357
9358unique-id-header <name>
9359 Add a unique ID header in the HTTP request.
9360 May be used in sections : defaults | frontend | listen | backend
9361 yes | yes | yes | no
9362 Arguments :
9363 <name> is the name of the header.
9364
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009365 Add a unique-id header in the HTTP request sent to the server, using the
9366 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +01009367
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009368 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01009369
Julien Vehentf21be322014-03-07 08:27:34 -05009370 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +01009371 unique-id-header X-Unique-ID
9372
9373 will generate:
9374
9375 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
9376
9377 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +01009378
Willy Tarreauf51658d2014-04-23 01:21:56 +02009379use_backend <backend> [{if | unless} <condition>]
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02009380 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009381 May be used in sections : defaults | frontend | listen | backend
9382 no | yes | yes | no
9383 Arguments :
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01009384 <backend> is the name of a valid backend or "listen" section, or a
9385 "log-format" string resolving to a backend name.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009386
Willy Tarreauf51658d2014-04-23 01:21:56 +02009387 <condition> is a condition composed of ACLs, as described in section 7. If
9388 it is omitted, the rule is unconditionally applied.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009389
9390 When doing content-switching, connections arrive on a frontend and are then
9391 dispatched to various backends depending on a number of conditions. The
9392 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02009393 "use_backend" keyword. While it is normally used with HTTP processing, it can
9394 also be used in pure TCP, either without content using stateless ACLs (eg:
9395 source address validation) or combined with a "tcp-request" rule to wait for
9396 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009397
9398 There may be as many "use_backend" rules as desired. All of these rules are
9399 evaluated in their declaration order, and the first one which matches will
9400 assign the backend.
9401
9402 In the first form, the backend will be used if the condition is met. In the
9403 second form, the backend will be used if the condition is not met. If no
9404 condition is valid, the backend defined with "default_backend" will be used.
9405 If no default backend is defined, either the servers in the same section are
9406 used (in case of a "listen" section) or, in case of a frontend, no server is
9407 used and a 503 service unavailable response is returned.
9408
Willy Tarreau51aecc72009-07-12 09:47:04 +02009409 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009410 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +02009411 and backend processing will immediately follow, or the backend will wait for
9412 a complete HTTP request to get in. This feature is useful when a frontend
9413 must decode several protocols on a unique port, one of them being HTTP.
9414
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01009415 When <backend> is a simple name, it is resolved at configuration time, and an
9416 error is reported if the specified backend does not exist. If <backend> is
9417 a log-format string instead, no check may be done at configuration time, so
9418 the backend name is resolved dynamically at run time. If the resulting
9419 backend name does not correspond to any valid backend, no other rule is
9420 evaluated, and the default_backend directive is applied instead. Note that
9421 when using dynamic backend names, it is highly recommended to use a prefix
9422 that no other backend uses in order to ensure that an unauthorized backend
9423 cannot be forced from the request.
9424
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009425 It is worth mentioning that "use_backend" rules with an explicit name are
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01009426 used to detect the association between frontends and backends to compute the
9427 backend's "fullconn" setting. This cannot be done for dynamic names.
9428
9429 See also: "default_backend", "tcp-request", "fullconn", "log-format", and
9430 section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +01009431
Willy Tarreau036fae02008-01-06 13:24:40 +01009432
Willy Tarreau4a5cade2012-04-05 21:09:48 +02009433use-server <server> if <condition>
9434use-server <server> unless <condition>
9435 Only use a specific server if/unless an ACL-based condition is matched.
9436 May be used in sections : defaults | frontend | listen | backend
9437 no | no | yes | yes
9438 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009439 <server> is the name of a valid server in the same backend section.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02009440
9441 <condition> is a condition composed of ACLs, as described in section 7.
9442
9443 By default, connections which arrive to a backend are load-balanced across
9444 the available servers according to the configured algorithm, unless a
9445 persistence mechanism such as a cookie is used and found in the request.
9446
9447 Sometimes it is desirable to forward a particular request to a specific
9448 server without having to declare a dedicated backend for this server. This
9449 can be achieved using the "use-server" rules. These rules are evaluated after
9450 the "redirect" rules and before evaluating cookies, and they have precedence
9451 on them. There may be as many "use-server" rules as desired. All of these
9452 rules are evaluated in their declaration order, and the first one which
9453 matches will assign the server.
9454
9455 If a rule designates a server which is down, and "option persist" is not used
9456 and no force-persist rule was validated, it is ignored and evaluation goes on
9457 with the next rules until one matches.
9458
9459 In the first form, the server will be used if the condition is met. In the
9460 second form, the server will be used if the condition is not met. If no
9461 condition is valid, the processing continues and the server will be assigned
9462 according to other persistence mechanisms.
9463
9464 Note that even if a rule is matched, cookie processing is still performed but
9465 does not assign the server. This allows prefixed cookies to have their prefix
9466 stripped.
9467
9468 The "use-server" statement works both in HTTP and TCP mode. This makes it
9469 suitable for use with content-based inspection. For instance, a server could
9470 be selected in a farm according to the TLS SNI field. And if these servers
9471 have their weight set to zero, they will not be used for other traffic.
9472
9473 Example :
9474 # intercept incoming TLS requests based on the SNI field
9475 use-server www if { req_ssl_sni -i www.example.com }
9476 server www 192.168.0.1:443 weight 0
9477 use-server mail if { req_ssl_sni -i mail.example.com }
9478 server mail 192.168.0.1:587 weight 0
9479 use-server imap if { req_ssl_sni -i imap.example.com }
9480 server mail 192.168.0.1:993 weight 0
9481 # all the rest is forwarded to this server
9482 server default 192.168.0.2:443 check
9483
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009484 See also: "use_backend", section 5 about server and section 7 about ACLs.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02009485
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009486
94875. Bind and Server options
9488--------------------------
9489
9490The "bind", "server" and "default-server" keywords support a number of settings
9491depending on some build options and on the system HAProxy was built on. These
9492settings generally each consist in one word sometimes followed by a value,
9493written on the same line as the "bind" or "server" line. All these options are
9494described in this section.
9495
9496
94975.1. Bind options
9498-----------------
9499
9500The "bind" keyword supports a certain number of settings which are all passed
9501as arguments on the same line. The order in which those arguments appear makes
9502no importance, provided that they appear after the bind address. All of these
9503parameters are optional. Some of them consist in a single words (booleans),
9504while other ones expect a value after them. In this case, the value must be
9505provided immediately after the setting name.
9506
9507The currently supported settings are the following ones.
9508
9509accept-proxy
9510 Enforces the use of the PROXY protocol over any connection accepted by any of
Willy Tarreau77992672014-06-14 11:06:17 +02009511 the sockets declared on the same line. Versions 1 and 2 of the PROXY protocol
9512 are supported and correctly detected. The PROXY protocol dictates the layer
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009513 3/4 addresses of the incoming connection to be used everywhere an address is
9514 used, with the only exception of "tcp-request connection" rules which will
9515 only see the real connection address. Logs will reflect the addresses
9516 indicated in the protocol, unless it is violated, in which case the real
9517 address will still be used. This keyword combined with support from external
9518 components can be used as an efficient and reliable alternative to the
9519 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +02009520 usable. See also "tcp-request connection expect-proxy" for a finer-grained
9521 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009522
Willy Tarreauab861d32013-04-02 02:30:41 +02009523alpn <protocols>
9524 This enables the TLS ALPN extension and advertises the specified protocol
9525 list as supported on top of ALPN. The protocol list consists in a comma-
9526 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
9527 quotes). This requires that the SSL library is build with support for TLS
9528 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
9529 initial NPN extension.
9530
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009531backlog <backlog>
9532 Sets the socket's backlog to this value. If unspecified, the frontend's
9533 backlog is used instead, which generally defaults to the maxconn value.
9534
Emeric Brun7fb34422012-09-28 15:26:15 +02009535ecdhe <named curve>
9536 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +01009537 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
9538 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +02009539
Emeric Brunfd33a262012-10-11 16:28:27 +02009540ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +02009541 This setting is only available when support for OpenSSL was built in. It
9542 designates a PEM file from which to load CA certificates used to verify
9543 client's certificate.
9544
Emeric Brunb6dc9342012-09-28 17:55:37 +02009545ca-ignore-err [all|<errorID>,...]
9546 This setting is only available when support for OpenSSL was built in.
9547 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
9548 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
9549 error is ignored.
9550
Christopher Faulet31af49d2015-06-09 17:29:50 +02009551ca-sign-file <cafile>
9552 This setting is only available when support for OpenSSL was built in. It
9553 designates a PEM file containing both the CA certificate and the CA private
9554 key used to create and sign server's certificates. This is a mandatory
9555 setting when the dynamic generation of certificates is enabled. See
9556 'generate-certificates' for details.
9557
9558ca-sign-passphrase <passphrase>
9559 This setting is only available when support for OpenSSL was built in. It is
9560 the CA private key passphrase. This setting is optional and used only when
9561 the dynamic generation of certificates is enabled. See
9562 'generate-certificates' for details.
9563
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009564ciphers <ciphers>
9565 This setting is only available when support for OpenSSL was built in. It sets
9566 the string describing the list of cipher algorithms ("cipher suite") that are
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009567 negotiated during the SSL/TLS handshake. The format of the string is defined
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009568 in "man 1 ciphers" from OpenSSL man pages, and can be for instance a string
9569 such as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes).
9570
Emeric Brunfd33a262012-10-11 16:28:27 +02009571crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +02009572 This setting is only available when support for OpenSSL was built in. It
9573 designates a PEM file from which to load certificate revocation list used
9574 to verify client's certificate.
9575
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009576crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +00009577 This setting is only available when support for OpenSSL was built in. It
9578 designates a PEM file containing both the required certificates and any
9579 associated private keys. This file can be built by concatenating multiple
9580 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
9581 requires an intermediate certificate, this can also be concatenated into this
9582 file.
9583
9584 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
9585 are loaded.
9586
9587 If a directory name is used instead of a PEM file, then all files found in
Cyril Bonté3180f7b2015-01-25 00:16:08 +01009588 that directory will be loaded in alphabetic order unless their name ends with
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +01009589 '.issuer', '.ocsp' or '.sctl' (reserved extensions). This directive may be
9590 specified multiple times in order to load certificates from multiple files or
9591 directories. The certificates will be presented to clients who provide a
9592 valid TLS Server Name Indication field matching one of their CN or alt
9593 subjects. Wildcards are supported, where a wildcard character '*' is used
9594 instead of the first hostname component (eg: *.example.org matches
9595 www.example.org but not www.sub.example.org).
Alex Davies0fbf0162013-03-02 16:04:50 +00009596
9597 If no SNI is provided by the client or if the SSL library does not support
9598 TLS extensions, or if the client provides an SNI hostname which does not
9599 match any certificate, then the first loaded certificate will be presented.
9600 This means that when loading certificates from a directory, it is highly
Cyril Bonté3180f7b2015-01-25 00:16:08 +01009601 recommended to load the default one first as a file or to ensure that it will
9602 always be the first one in the directory.
Alex Davies0fbf0162013-03-02 16:04:50 +00009603
Emeric Brune032bfa2012-09-28 13:01:45 +02009604 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009605
Alex Davies0fbf0162013-03-02 16:04:50 +00009606 Some CAs (such as Godaddy) offer a drop down list of server types that do not
9607 include HAProxy when obtaining a certificate. If this happens be sure to
Godbach8bf60a12014-04-21 21:42:41 +08009608 choose a webserver that the CA believes requires an intermediate CA (for
Alex Davies0fbf0162013-03-02 16:04:50 +00009609 Godaddy, selection Apache Tomcat will get the correct bundle, but many
9610 others, e.g. nginx, result in a wrong bundle that will not work for some
9611 clients).
9612
Emeric Brun4147b2e2014-06-16 18:36:30 +02009613 For each PEM file, haproxy checks for the presence of file at the same path
9614 suffixed by ".ocsp". If such file is found, support for the TLS Certificate
9615 Status Request extension (also known as "OCSP stapling") is automatically
9616 enabled. The content of this file is optional. If not empty, it must contain
9617 a valid OCSP Response in DER format. In order to be valid an OCSP Response
9618 must comply with the following rules: it has to indicate a good status,
9619 it has to be a single response for the certificate of the PEM file, and it
9620 has to be valid at the moment of addition. If these rules are not respected
9621 the OCSP Response is ignored and a warning is emitted. In order to identify
9622 which certificate an OCSP Response applies to, the issuer's certificate is
9623 necessary. If the issuer's certificate is not found in the PEM file, it will
9624 be loaded from a file at the same path as the PEM file suffixed by ".issuer"
9625 if it exists otherwise it will fail with an error.
9626
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +01009627 For each PEM file, haproxy also checks for the presence of file at the same
9628 path suffixed by ".sctl". If such file is found, support for Certificate
9629 Transparency (RFC6962) TLS extension is enabled. The file must contain a
9630 valid Signed Certificate Timestamp List, as described in RFC. File is parsed
9631 to check basic syntax, but no signatures are verified.
9632
Emeric Brunb6dc9342012-09-28 17:55:37 +02009633crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +00009634 This setting is only available when support for OpenSSL was built in. Sets a
9635 comma separated list of errorIDs to ignore during verify at depth == 0. If
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009636 set to 'all', all errors are ignored. SSL handshake is not aborted if an error
Alex Davies0fbf0162013-03-02 16:04:50 +00009637 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +02009638
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009639crt-list <file>
9640 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02009641 designates a list of PEM file with an optional list of SNI filter per
9642 certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009643
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02009644 <crtfile> [[!]<snifilter> ...]
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009645
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02009646 Wildcards are supported in the SNI filter. Negative filter are also supported,
9647 only useful in combination with a wildcard filter to exclude a particular SNI.
9648 The certificates will be presented to clients who provide a valid TLS Server
9649 Name Indication field matching one of the SNI filters. If no SNI filter is
9650 specified, the CN and alt subjects are used. This directive may be specified
9651 multiple times. See the "crt" option for more information. The default
9652 certificate is still needed to meet OpenSSL expectations. If it is not used,
9653 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009654
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009655defer-accept
9656 Is an optional keyword which is supported only on certain Linux kernels. It
9657 states that a connection will only be accepted once some data arrive on it,
9658 or at worst after the first retransmit. This should be used only on protocols
9659 for which the client talks first (eg: HTTP). It can slightly improve
9660 performance by ensuring that most of the request is already available when
9661 the connection is accepted. On the other hand, it will not be able to detect
9662 connections which don't talk. It is important to note that this option is
9663 broken in all kernels up to 2.6.31, as the connection is never accepted until
9664 the client talks. This can cause issues with front firewalls which would see
9665 an established connection while the proxy will only see it in SYN_RECV. This
9666 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
9667
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009668force-sslv3
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009669 This option enforces use of SSLv3 only on SSL connections instantiated from
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009670 this listener. SSLv3 is generally less expensive than the TLS counterparts
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009671 for high connection rates. This option is also available on global statement
9672 "ssl-default-bind-options". See also "no-tlsv*" and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009673
9674force-tlsv10
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009675 This option enforces use of TLSv1.0 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009676 this listener. This option is also available on global statement
9677 "ssl-default-bind-options". See also "no-tlsv*" and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009678
9679force-tlsv11
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009680 This option enforces use of TLSv1.1 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009681 this listener. This option is also available on global statement
9682 "ssl-default-bind-options". See also "no-tlsv*", and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009683
9684force-tlsv12
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009685 This option enforces use of TLSv1.2 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009686 this listener. This option is also available on global statement
9687 "ssl-default-bind-options". See also "no-tlsv*", and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009688
Christopher Faulet31af49d2015-06-09 17:29:50 +02009689generate-certificates
9690 This setting is only available when support for OpenSSL was built in. It
9691 enables the dynamic SSL certificates generation. A CA certificate and its
9692 private key are necessary (see 'ca-sign-file'). When HAProxy is configured as
9693 a transparent forward proxy, SSL requests generate errors because of a common
9694 name mismatch on the certificate presented to the client. With this option
9695 enabled, HAProxy will try to forge a certificate using the SNI hostname
9696 indicated by the client. This is done only if no certificate matches the SNI
9697 hostname (see 'crt-list'). If an error occurs, the default certificate is
9698 used, else the 'strict-sni' option is set.
9699 It can also be used when HAProxy is configured as a reverse proxy to ease the
9700 deployment of an architecture with many backends.
9701
9702 Creating a SSL certificate is an expensive operation, so a LRU cache is used
9703 to store forged certificates (see 'tune.ssl.ssl-ctx-cache-size'). It
9704 increases the HAProxy's memroy footprint to reduce latency when the same
9705 certificate is used many times.
9706
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009707gid <gid>
9708 Sets the group of the UNIX sockets to the designated system gid. It can also
9709 be set by default in the global section's "unix-bind" statement. Note that
9710 some platforms simply ignore this. This setting is equivalent to the "group"
9711 setting except that the group ID is used instead of its name. This setting is
9712 ignored by non UNIX sockets.
9713
9714group <group>
9715 Sets the group of the UNIX sockets to the designated system group. It can
9716 also be set by default in the global section's "unix-bind" statement. Note
9717 that some platforms simply ignore this. This setting is equivalent to the
9718 "gid" setting except that the group name is used instead of its gid. This
9719 setting is ignored by non UNIX sockets.
9720
9721id <id>
9722 Fixes the socket ID. By default, socket IDs are automatically assigned, but
9723 sometimes it is more convenient to fix them to ease monitoring. This value
9724 must be strictly positive and unique within the listener/frontend. This
9725 option can only be used when defining only a single socket.
9726
9727interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +01009728 Restricts the socket to a specific interface. When specified, only packets
9729 received from that particular interface are processed by the socket. This is
9730 currently only supported on Linux. The interface must be a primary system
9731 interface, not an aliased interface. It is also possible to bind multiple
9732 frontends to the same address if they are bound to different interfaces. Note
9733 that binding to a network interface requires root privileges. This parameter
9734 is only compatible with TCPv4/TCPv6 sockets.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009735
Willy Tarreauabb175f2012-09-24 12:43:26 +02009736level <level>
9737 This setting is used with the stats sockets only to restrict the nature of
9738 the commands that can be issued on the socket. It is ignored by other
9739 sockets. <level> can be one of :
9740 - "user" is the least privileged level ; only non-sensitive stats can be
9741 read, and no change is allowed. It would make sense on systems where it
9742 is not easy to restrict access to the socket.
9743 - "operator" is the default level and fits most common uses. All data can
9744 be read, and only non-sensitive changes are permitted (eg: clear max
9745 counters).
9746 - "admin" should be used with care, as everything is permitted (eg: clear
9747 all counters).
9748
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009749maxconn <maxconn>
9750 Limits the sockets to this number of concurrent connections. Extraneous
9751 connections will remain in the system's backlog until a connection is
9752 released. If unspecified, the limit will be the same as the frontend's
9753 maxconn. Note that in case of port ranges or multiple addresses, the same
9754 value will be applied to each socket. This setting enables different
9755 limitations on expensive sockets, for instance SSL entries which may easily
9756 eat all memory.
9757
9758mode <mode>
9759 Sets the octal mode used to define access permissions on the UNIX socket. It
9760 can also be set by default in the global section's "unix-bind" statement.
9761 Note that some platforms simply ignore this. This setting is ignored by non
9762 UNIX sockets.
9763
9764mss <maxseg>
9765 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
9766 connections. This can be used to force a lower MSS for certain specific
9767 ports, for instance for connections passing through a VPN. Note that this
9768 relies on a kernel feature which is theoretically supported under Linux but
9769 was buggy in all versions prior to 2.6.28. It may or may not work on other
9770 operating systems. It may also not change the advertised value but change the
9771 effective size of outgoing segments. The commonly advertised value for TCPv4
9772 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
9773 positive, it will be used as the advertised MSS. If it is negative, it will
9774 indicate by how much to reduce the incoming connection's advertised MSS for
9775 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
9776
9777name <name>
9778 Sets an optional name for these sockets, which will be reported on the stats
9779 page.
9780
9781nice <nice>
9782 Sets the 'niceness' of connections initiated from the socket. Value must be
9783 in the range -1024..1024 inclusive, and defaults to zero. Positive values
9784 means that such connections are more friendly to others and easily offer
9785 their place in the scheduler. On the opposite, negative values mean that
9786 connections want to run with a higher priority than others. The difference
9787 only happens under high loads when the system is close to saturation.
9788 Negative values are appropriate for low-latency or administration services,
9789 and high values are generally recommended for CPU intensive tasks such as SSL
9790 processing or bulk transfers which are less sensible to latency. For example,
9791 it may make sense to use a positive value for an SMTP socket and a negative
9792 one for an RDP socket.
9793
Emeric Brun9b3009b2012-10-05 11:55:06 +02009794no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009795 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009796 disables support for SSLv3 on any sockets instantiated from the listener when
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009797 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009798 be enabled using any configuration option. This option is also available on
9799 global statement "ssl-default-bind-options". See also "force-tls*",
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009800 and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009801
Emeric Brun90ad8722012-10-02 14:00:59 +02009802no-tls-tickets
9803 This setting is only available when support for OpenSSL was built in. It
9804 disables the stateless session resumption (RFC 5077 TLS Ticket
9805 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009806 session resumption is more expensive in CPU usage. This option is also
9807 available on global statement "ssl-default-bind-options".
Emeric Brun90ad8722012-10-02 14:00:59 +02009808
Emeric Brun9b3009b2012-10-05 11:55:06 +02009809no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009810 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009811 disables support for TLSv1.0 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009812 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009813 cannot be enabled using any configuration option. This option is also
9814 available on global statement "ssl-default-bind-options". See also
9815 "force-tlsv*", and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009816
Emeric Brun9b3009b2012-10-05 11:55:06 +02009817no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +02009818 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009819 disables support for TLSv1.1 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009820 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009821 cannot be enabled using any configuration option. This option is also
9822 available on global statement "ssl-default-bind-options". See also
9823 "force-tlsv*", and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +02009824
Emeric Brun9b3009b2012-10-05 11:55:06 +02009825no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +02009826 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009827 disables support for TLSv1.2 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009828 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009829 cannot be enabled using any configuration option. This option is also
9830 available on global statement "ssl-default-bind-options". See also
9831 "force-tlsv*", and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +02009832
Willy Tarreau6c9a3d52012-10-18 18:57:14 +02009833npn <protocols>
9834 This enables the NPN TLS extension and advertises the specified protocol list
9835 as supported on top of NPN. The protocol list consists in a comma-delimited
9836 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
9837 This requires that the SSL library is build with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +02009838 enabled (check with haproxy -vv). Note that the NPN extension has been
9839 replaced with the ALPN extension (see the "alpn" keyword).
Willy Tarreau6c9a3d52012-10-18 18:57:14 +02009840
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02009841process [ all | odd | even | <number 1-64>[-<number 1-64>] ]
9842 This restricts the list of processes on which this listener is allowed to
9843 run. It does not enforce any process but eliminates those which do not match.
9844 If the frontend uses a "bind-process" setting, the intersection between the
9845 two is applied. If in the end the listener is not allowed to run on any
9846 remaining process, a warning is emitted, and the listener will either run on
9847 the first process of the listener if a single process was specified, or on
9848 all of its processes if multiple processes were specified. For the unlikely
Willy Tarreauae302532014-05-07 19:22:24 +02009849 case where several ranges are needed, this directive may be repeated. The
9850 main purpose of this directive is to be used with the stats sockets and have
9851 one different socket per process. The second purpose is to have multiple bind
9852 lines sharing the same IP:port but not the same process in a listener, so
9853 that the system can distribute the incoming connections into multiple queues
9854 and allow a smoother inter-process load balancing. Currently Linux 3.9 and
9855 above is known for supporting this. See also "bind-process" and "nbproc".
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02009856
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009857ssl
9858 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009859 enables SSL deciphering on connections instantiated from this listener. A
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009860 certificate is necessary (see "crt" above). All contents in the buffers will
9861 appear in clear text, so that ACLs and HTTP processing will only have access
9862 to deciphered contents.
9863
Emmanuel Hocdet65623372013-01-24 17:17:15 +01009864strict-sni
9865 This setting is only available when support for OpenSSL was built in. The
9866 SSL/TLS negotiation is allow only if the client provided an SNI which match
9867 a certificate. The default certificate is not used.
9868 See the "crt" option for more information.
9869
Willy Tarreau2af207a2015-02-04 00:45:58 +01009870tcp-ut <delay>
9871 Sets the TCP User Timeout for all incoming connections instanciated from this
9872 listening socket. This option is available on Linux since version 2.6.37. It
9873 allows haproxy to configure a timeout for sockets which contain data not
9874 receiving an acknoledgement for the configured delay. This is especially
9875 useful on long-lived connections experiencing long idle periods such as
9876 remote terminals or database connection pools, where the client and server
9877 timeouts must remain high to allow a long period of idle, but where it is
9878 important to detect that the client has disappeared in order to release all
9879 resources associated with its connection (and the server's session). The
9880 argument is a delay expressed in milliseconds by default. This only works
9881 for regular TCP connections, and is ignored for other protocols.
9882
Willy Tarreau1c862c52012-10-05 16:21:00 +02009883tfo
Lukas Tribus0defb902013-02-13 23:35:39 +01009884 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +02009885 enables TCP Fast Open on the listening socket, which means that clients which
9886 support this feature will be able to send a request and receive a response
9887 during the 3-way handshake starting from second connection, thus saving one
9888 round-trip after the first connection. This only makes sense with protocols
9889 that use high connection rates and where each round trip matters. This can
9890 possibly cause issues with many firewalls which do not accept data on SYN
9891 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +02009892 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
9893 need to build HAProxy with USE_TFO=1 if your libc doesn't define
9894 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +02009895
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +01009896tls-ticket-keys <keyfile>
9897 Sets the TLS ticket keys file to load the keys from. The keys need to be 48
9898 bytes long, encoded with base64 (ex. openssl rand -base64 48). Number of keys
9899 is specified by the TLS_TICKETS_NO build option (default 3) and at least as
9900 many keys need to be present in the file. Last TLS_TICKETS_NO keys will be
9901 used for decryption and the penultimate one for encryption. This enables easy
9902 key rotation by just appending new key to the file and reloading the process.
9903 Keys must be periodically rotated (ex. every 12h) or Perfect Forward Secrecy
9904 is compromised. It is also a good idea to keep the keys off any permanent
9905 storage such as hard drives (hint: use tmpfs and don't swap those files).
9906 Lifetime hint can be changed using tune.ssl.timeout.
9907
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009908transparent
9909 Is an optional keyword which is supported only on certain Linux kernels. It
9910 indicates that the addresses will be bound even if they do not belong to the
9911 local machine, and that packets targeting any of these addresses will be
9912 intercepted just as if the addresses were locally configured. This normally
9913 requires that IP forwarding is enabled. Caution! do not use this with the
9914 default address '*', as it would redirect any traffic for the specified port.
9915 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
9916 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
9917 kernel version. Some distribution kernels include backports of the feature,
9918 so check for support with your vendor.
9919
Willy Tarreau77e3af92012-11-24 15:07:23 +01009920v4v6
9921 Is an optional keyword which is supported only on most recent systems
9922 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
9923 and IPv6 when it uses the default address. Doing so is sometimes necessary
9924 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009925 sockets, and is overridden by the "v6only" option.
Willy Tarreau77e3af92012-11-24 15:07:23 +01009926
Willy Tarreau9b6700f2012-11-24 11:55:28 +01009927v6only
9928 Is an optional keyword which is supported only on most recent systems
9929 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
9930 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +01009931 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
9932 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +01009933
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009934uid <uid>
9935 Sets the owner of the UNIX sockets to the designated system uid. It can also
9936 be set by default in the global section's "unix-bind" statement. Note that
9937 some platforms simply ignore this. This setting is equivalent to the "user"
9938 setting except that the user numeric ID is used instead of its name. This
9939 setting is ignored by non UNIX sockets.
9940
9941user <user>
9942 Sets the owner of the UNIX sockets to the designated system user. It can also
9943 be set by default in the global section's "unix-bind" statement. Note that
9944 some platforms simply ignore this. This setting is equivalent to the "uid"
9945 setting except that the user name is used instead of its uid. This setting is
9946 ignored by non UNIX sockets.
9947
Emeric Brun1a073b42012-09-28 17:07:34 +02009948verify [none|optional|required]
9949 This setting is only available when support for OpenSSL was built in. If set
9950 to 'none', client certificate is not requested. This is the default. In other
9951 cases, a client certificate is requested. If the client does not provide a
9952 certificate after the request and if 'verify' is set to 'required', then the
9953 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +02009954 certificate provided by the client is always verified using CAs from
9955 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
9956 is aborted, regardless of the 'verify' option, unless the error code exactly
9957 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02009958
Willy Tarreaub6205fd2012-09-24 12:27:33 +020099595.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +01009960------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02009961
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01009962The "server" and "default-server" keywords support a certain number of settings
9963which are all passed as arguments on the server line. The order in which those
9964arguments appear does not count, and they are all optional. Some of those
9965settings are single words (booleans) while others expect one or several values
9966after them. In this case, the values must immediately follow the setting name.
9967Except default-server, all those settings must be specified after the server's
9968address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +02009969
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009970 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01009971 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +02009972
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009973The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +01009974
Willy Tarreauceb4ac92012-04-28 00:41:46 +02009975addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009976 Using the "addr" parameter, it becomes possible to use a different IP address
9977 to send health-checks. On some servers, it may be desirable to dedicate an IP
9978 address to specific component able to perform complex tests which are more
9979 suitable to health-checks than the application. This parameter is ignored if
9980 the "check" parameter is not set. See also the "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +02009981
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009982 Supported in default-server: No
9983
Simon Hormand60d6912013-11-25 10:46:36 +09009984agent-check
9985 Enable an auxiliary agent check which is run independently of a regular
Willy Tarreau81f5d942013-12-09 20:51:51 +01009986 health check. An agent health check is performed by making a TCP connection
9987 to the port set by the "agent-port" parameter and reading an ASCII string.
9988 The string is made of a series of words delimited by spaces, tabs or commas
9989 in any order, optionally terminated by '\r' and/or '\n', each consisting of :
Simon Hormand60d6912013-11-25 10:46:36 +09009990
Willy Tarreau81f5d942013-12-09 20:51:51 +01009991 - An ASCII representation of a positive integer percentage, e.g. "75%".
Simon Hormand60d6912013-11-25 10:46:36 +09009992 Values in this format will set the weight proportional to the initial
Willy Tarreauc5af3a62014-10-07 15:27:33 +02009993 weight of a server as configured when haproxy starts. Note that a zero
9994 weight is reported on the stats page as "DRAIN" since it has the same
9995 effect on the server (it's removed from the LB farm).
Simon Hormand60d6912013-11-25 10:46:36 +09009996
Willy Tarreau81f5d942013-12-09 20:51:51 +01009997 - The word "ready". This will turn the server's administrative state to the
9998 READY mode, thus cancelling any DRAIN or MAINT state
Simon Hormand60d6912013-11-25 10:46:36 +09009999
Willy Tarreau81f5d942013-12-09 20:51:51 +010010000 - The word "drain". This will turn the server's administrative state to the
10001 DRAIN mode, thus it will not accept any new connections other than those
10002 that are accepted via persistence.
Simon Hormand60d6912013-11-25 10:46:36 +090010003
Willy Tarreau81f5d942013-12-09 20:51:51 +010010004 - The word "maint". This will turn the server's administrative state to the
10005 MAINT mode, thus it will not accept any new connections at all, and health
10006 checks will be stopped.
Simon Hormand60d6912013-11-25 10:46:36 +090010007
Willy Tarreau81f5d942013-12-09 20:51:51 +010010008 - The words "down", "failed", or "stopped", optionally followed by a
10009 description string after a sharp ('#'). All of these mark the server's
10010 operating state as DOWN, but since the word itself is reported on the stats
10011 page, the difference allows an administrator to know if the situation was
10012 expected or not : the service may intentionally be stopped, may appear up
10013 but fail some validity tests, or may be seen as down (eg: missing process,
10014 or port not responding).
Simon Hormand60d6912013-11-25 10:46:36 +090010015
Willy Tarreau81f5d942013-12-09 20:51:51 +010010016 - The word "up" sets back the server's operating state as UP if health checks
10017 also report that the service is accessible.
Simon Hormand60d6912013-11-25 10:46:36 +090010018
Willy Tarreau81f5d942013-12-09 20:51:51 +010010019 Parameters which are not advertised by the agent are not changed. For
10020 example, an agent might be designed to monitor CPU usage and only report a
10021 relative weight and never interact with the operating status. Similarly, an
10022 agent could be designed as an end-user interface with 3 radio buttons
10023 allowing an administrator to change only the administrative state. However,
10024 it is important to consider that only the agent may revert its own actions,
10025 so if a server is set to DRAIN mode or to DOWN state using the agent, the
10026 agent must implement the other equivalent actions to bring the service into
10027 operations again.
Simon Hormand60d6912013-11-25 10:46:36 +090010028
Simon Horman2f1f9552013-11-25 10:46:37 +090010029 Failure to connect to the agent is not considered an error as connectivity
10030 is tested by the regular health check which is enabled by the "check"
Willy Tarreau81f5d942013-12-09 20:51:51 +010010031 parameter. Warning though, it is not a good idea to stop an agent after it
10032 reports "down", since only an agent reporting "up" will be able to turn the
10033 server up again. Note that the CLI on the Unix stats socket is also able to
10034 force an agent's result in order to workaround a bogus agent if needed.
Simon Horman2f1f9552013-11-25 10:46:37 +090010035
Willy Tarreau81f5d942013-12-09 20:51:51 +010010036 Requires the "agent-port" parameter to be set. See also the "agent-inter"
10037 parameter.
Simon Hormand60d6912013-11-25 10:46:36 +090010038
10039 Supported in default-server: No
10040
10041agent-inter <delay>
10042 The "agent-inter" parameter sets the interval between two agent checks
10043 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
10044
10045 Just as with every other time-based parameter, it may be entered in any
10046 other explicit unit among { us, ms, s, m, h, d }. The "agent-inter"
10047 parameter also serves as a timeout for agent checks "timeout check" is
10048 not set. In order to reduce "resonance" effects when multiple servers are
10049 hosted on the same hardware, the agent and health checks of all servers
10050 are started with a small time offset between them. It is also possible to
10051 add some random noise in the agent and health checks interval using the
10052 global "spread-checks" keyword. This makes sense for instance when a lot
10053 of backends use the same servers.
10054
10055 See also the "agent-check" and "agent-port" parameters.
10056
10057 Supported in default-server: Yes
10058
10059agent-port <port>
10060 The "agent-port" parameter sets the TCP port used for agent checks.
10061
10062 See also the "agent-check" and "agent-inter" parameters.
10063
10064 Supported in default-server: Yes
10065
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010066backup
10067 When "backup" is present on a server line, the server is only used in load
10068 balancing when all other non-backup servers are unavailable. Requests coming
10069 with a persistence cookie referencing the server will always be served
10070 though. By default, only the first operational backup server is used, unless
10071 the "allbackups" option is set in the backend. See also the "allbackups"
10072 option.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010073
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010074 Supported in default-server: No
10075
Emeric Brunef42d922012-10-11 16:11:36 +020010076ca-file <cafile>
10077 This setting is only available when support for OpenSSL was built in. It
10078 designates a PEM file from which to load CA certificates used to verify
10079 server's certificate.
10080
10081 Supported in default-server: No
10082
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010083check
10084 This option enables health checks on the server. By default, a server is
Patrick Mézardb7aeec62012-01-22 16:01:22 +010010085 always considered available. If "check" is set, the server is available when
10086 accepting periodic TCP connections, to ensure that it is really able to serve
10087 requests. The default address and port to send the tests to are those of the
10088 server, and the default source is the same as the one defined in the
10089 backend. It is possible to change the address using the "addr" parameter, the
10090 port using the "port" parameter, the source address using the "source"
10091 address, and the interval and timers using the "inter", "rise" and "fall"
Simon Hormanafc47ee2013-11-25 10:46:35 +090010092 parameters. The request method is define in the backend using the "httpchk",
10093 "smtpchk", "mysql-check", "pgsql-check" and "ssl-hello-chk" options. Please
10094 refer to those options and parameters for more information.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010095
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010096 Supported in default-server: No
10097
Willy Tarreau6c16adc2012-10-05 00:04:16 +020010098check-send-proxy
10099 This option forces emission of a PROXY protocol line with outgoing health
10100 checks, regardless of whether the server uses send-proxy or not for the
10101 normal traffic. By default, the PROXY protocol is enabled for health checks
10102 if it is already enabled for normal traffic and if no "port" nor "addr"
10103 directive is present. However, if such a directive is present, the
10104 "check-send-proxy" option needs to be used to force the use of the
10105 protocol. See also the "send-proxy" option for more information.
10106
10107 Supported in default-server: No
10108
Willy Tarreau763a95b2012-10-04 23:15:39 +020010109check-ssl
10110 This option forces encryption of all health checks over SSL, regardless of
10111 whether the server uses SSL or not for the normal traffic. This is generally
10112 used when an explicit "port" or "addr" directive is specified and SSL health
10113 checks are not inherited. It is important to understand that this option
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010114 inserts an SSL transport layer below the checks, so that a simple TCP connect
Willy Tarreau763a95b2012-10-04 23:15:39 +020010115 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
10116 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
10117 All SSL settings are common to health checks and traffic (eg: ciphers).
10118 See the "ssl" option for more information.
10119
10120 Supported in default-server: No
10121
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010122ciphers <ciphers>
10123 This option sets the string describing the list of cipher algorithms that is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010124 is negotiated during the SSL/TLS handshake with the server. The format of the
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010125 string is defined in "man 1 ciphers". When SSL is used to communicate with
10126 servers on the local network, it is common to see a weaker set of algorithms
10127 than what is used over the internet. Doing so reduces CPU usage on both the
10128 server and haproxy while still keeping it compatible with deployed software.
10129 Some algorithms such as RC4-SHA1 are reasonably cheap. If no security at all
10130 is needed and just connectivity, using DES can be appropriate.
10131
Willy Tarreau763a95b2012-10-04 23:15:39 +020010132 Supported in default-server: No
10133
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010134cookie <value>
10135 The "cookie" parameter sets the cookie value assigned to the server to
10136 <value>. This value will be checked in incoming requests, and the first
10137 operational server possessing the same value will be selected. In return, in
10138 cookie insertion or rewrite modes, this value will be assigned to the cookie
10139 sent to the client. There is nothing wrong in having several servers sharing
10140 the same cookie value, and it is in fact somewhat common between normal and
10141 backup servers. See also the "cookie" keyword in backend section.
10142
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010143 Supported in default-server: No
10144
Emeric Brunef42d922012-10-11 16:11:36 +020010145crl-file <crlfile>
10146 This setting is only available when support for OpenSSL was built in. It
10147 designates a PEM file from which to load certificate revocation list used
10148 to verify server's certificate.
10149
10150 Supported in default-server: No
10151
Emeric Bruna7aa3092012-10-26 12:58:00 +020010152crt <cert>
10153 This setting is only available when support for OpenSSL was built in.
10154 It designates a PEM file from which to load both a certificate and the
10155 associated private key. This file can be built by concatenating both PEM
10156 files into one. This certificate will be sent if the server send a client
10157 certificate request.
10158
10159 Supported in default-server: No
10160
Willy Tarreau96839092010-03-29 10:02:24 +020010161disabled
10162 The "disabled" keyword starts the server in the "disabled" state. That means
10163 that it is marked down in maintenance mode, and no connection other than the
10164 ones allowed by persist mode will reach it. It is very well suited to setup
10165 new servers, because normal traffic will never reach them, while it is still
10166 possible to test the service by making use of the force-persist mechanism.
10167
10168 Supported in default-server: No
10169
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010170error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +010010171 If health observing is enabled, the "error-limit" parameter specifies the
10172 number of consecutive errors that triggers event selected by the "on-error"
10173 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010174
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010175 Supported in default-server: Yes
10176
10177 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010178
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010179fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010180 The "fall" parameter states that a server will be considered as dead after
10181 <count> consecutive unsuccessful health checks. This value defaults to 3 if
10182 unspecified. See also the "check", "inter" and "rise" parameters.
10183
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010184 Supported in default-server: Yes
10185
Emeric Brun8694b9a2012-10-05 14:39:07 +020010186force-sslv3
10187 This option enforces use of SSLv3 only when SSL is used to communicate with
10188 the server. SSLv3 is generally less expensive than the TLS counterparts for
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010189 high connection rates. This option is also available on global statement
10190 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010191
10192 Supported in default-server: No
10193
10194force-tlsv10
10195 This option enforces use of TLSv1.0 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010196 the server. This option is also available on global statement
10197 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010198
10199 Supported in default-server: No
10200
10201force-tlsv11
10202 This option enforces use of TLSv1.1 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010203 the server. This option is also available on global statement
10204 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010205
10206 Supported in default-server: No
10207
10208force-tlsv12
10209 This option enforces use of TLSv1.2 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010210 the server. This option is also available on global statement
10211 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010212
10213 Supported in default-server: No
10214
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010215id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +020010216 Set a persistent ID for the server. This ID must be positive and unique for
10217 the proxy. An unused ID will automatically be assigned if unset. The first
10218 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010219
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010220 Supported in default-server: No
10221
10222inter <delay>
10223fastinter <delay>
10224downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010225 The "inter" parameter sets the interval between two consecutive health checks
10226 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
10227 It is also possible to use "fastinter" and "downinter" to optimize delays
10228 between checks depending on the server state :
10229
Pieter Baauw44fc9df2015-09-17 21:30:46 +020010230 Server state | Interval used
10231 ----------------------------------------+----------------------------------
10232 UP 100% (non-transitional) | "inter"
10233 ----------------------------------------+----------------------------------
10234 Transitionally UP (going down "fall"), | "fastinter" if set,
10235 Transitionally DOWN (going up "rise"), | "inter" otherwise.
10236 or yet unchecked. |
10237 ----------------------------------------+----------------------------------
10238 DOWN 100% (non-transitional) | "downinter" if set,
10239 | "inter" otherwise.
10240 ----------------------------------------+----------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +010010241
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010242 Just as with every other time-based parameter, they can be entered in any
10243 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
10244 serves as a timeout for health checks sent to servers if "timeout check" is
10245 not set. In order to reduce "resonance" effects when multiple servers are
Simon Hormand60d6912013-11-25 10:46:36 +090010246 hosted on the same hardware, the agent and health checks of all servers
10247 are started with a small time offset between them. It is also possible to
10248 add some random noise in the agent and health checks interval using the
10249 global "spread-checks" keyword. This makes sense for instance when a lot
10250 of backends use the same servers.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010251
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010252 Supported in default-server: Yes
10253
10254maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010255 The "maxconn" parameter specifies the maximal number of concurrent
10256 connections that will be sent to this server. If the number of incoming
10257 concurrent requests goes higher than this value, they will be queued, waiting
10258 for a connection to be released. This parameter is very important as it can
10259 save fragile servers from going down under extreme loads. If a "minconn"
10260 parameter is specified, the limit becomes dynamic. The default value is "0"
10261 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
10262 the backend's "fullconn" keyword.
10263
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010264 Supported in default-server: Yes
10265
10266maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010267 The "maxqueue" parameter specifies the maximal number of connections which
10268 will wait in the queue for this server. If this limit is reached, next
10269 requests will be redispatched to other servers instead of indefinitely
10270 waiting to be served. This will break persistence but may allow people to
10271 quickly re-log in when the server they try to connect to is dying. The
10272 default value is "0" which means the queue is unlimited. See also the
10273 "maxconn" and "minconn" parameters.
10274
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010275 Supported in default-server: Yes
10276
10277minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010278 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
10279 limit following the backend's load. The server will always accept at least
10280 <minconn> connections, never more than <maxconn>, and the limit will be on
10281 the ramp between both values when the backend has less than <fullconn>
10282 concurrent connections. This makes it possible to limit the load on the
10283 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010284 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010285 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010286
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010287 Supported in default-server: Yes
10288
Willy Tarreau2a3fb1c2015-02-05 16:47:07 +010010289no-ssl-reuse
10290 This option disables SSL session reuse when SSL is used to communicate with
10291 the server. It will force the server to perform a full handshake for every
10292 new connection. It's probably only useful for benchmarking, troubleshooting,
10293 and for paranoid users.
10294
10295 Supported in default-server: No
10296
Emeric Brun9b3009b2012-10-05 11:55:06 +020010297no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010298 This option disables support for SSLv3 when SSL is used to communicate with
10299 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emeric Brun8694b9a2012-10-05 14:39:07 +020010300 using any configuration option. See also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010301
Willy Tarreau763a95b2012-10-04 23:15:39 +020010302 Supported in default-server: No
10303
Emeric Brunf9c5c472012-10-11 15:28:34 +020010304no-tls-tickets
10305 This setting is only available when support for OpenSSL was built in. It
10306 disables the stateless session resumption (RFC 5077 TLS Ticket
10307 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010308 session resumption is more expensive in CPU usage for servers. This option
10309 is also available on global statement "ssl-default-server-options".
Emeric Brunf9c5c472012-10-11 15:28:34 +020010310
10311 Supported in default-server: No
10312
Emeric Brun9b3009b2012-10-05 11:55:06 +020010313no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +020010314 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020010315 the server. Note that SSLv2 is disabled in the code and cannot be enabled
10316 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010317 often makes sense to disable it when communicating with local servers. This
10318 option is also available on global statement "ssl-default-server-options".
10319 See also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +020010320
Willy Tarreau763a95b2012-10-04 23:15:39 +020010321 Supported in default-server: No
10322
Emeric Brun9b3009b2012-10-05 11:55:06 +020010323no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +020010324 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020010325 the server. Note that SSLv2 is disabled in the code and cannot be enabled
10326 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010327 often makes sense to disable it when communicating with local servers. This
10328 option is also available on global statement "ssl-default-server-options".
10329 See also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +020010330
Willy Tarreau763a95b2012-10-04 23:15:39 +020010331 Supported in default-server: No
10332
Emeric Brun9b3009b2012-10-05 11:55:06 +020010333no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +020010334 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010335 the server. Note that SSLv2 is disabled in the code and cannot be enabled
10336 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010337 often makes sense to disable it when communicating with local servers. This
10338 option is also available on global statement "ssl-default-server-options".
10339 See also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010340
Willy Tarreau763a95b2012-10-04 23:15:39 +020010341 Supported in default-server: No
10342
Simon Hormanfa461682011-06-25 09:39:49 +090010343non-stick
10344 Never add connections allocated to this sever to a stick-table.
10345 This may be used in conjunction with backup to ensure that
10346 stick-table persistence is disabled for backup servers.
10347
Willy Tarreau763a95b2012-10-04 23:15:39 +020010348 Supported in default-server: No
10349
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010350observe <mode>
10351 This option enables health adjusting based on observing communication with
10352 the server. By default this functionality is disabled and enabling it also
10353 requires to enable health checks. There are two supported modes: "layer4" and
10354 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
10355 significant. In layer7, which is only allowed for http proxies, responses
10356 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +010010357 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010358
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010359 Supported in default-server: No
10360
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010361 See also the "check", "on-error" and "error-limit".
10362
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010363on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010364 Select what should happen when enough consecutive errors are detected.
10365 Currently, four modes are available:
10366 - fastinter: force fastinter
10367 - fail-check: simulate a failed check, also forces fastinter (default)
10368 - sudden-death: simulate a pre-fatal failed health check, one more failed
10369 check will mark a server down, forces fastinter
10370 - mark-down: mark the server immediately down and force fastinter
10371
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010372 Supported in default-server: Yes
10373
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010374 See also the "check", "observe" and "error-limit".
10375
Simon Hormane0d1bfb2011-06-21 14:34:58 +090010376on-marked-down <action>
10377 Modify what occurs when a server is marked down.
10378 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070010379 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
10380 all connections to the server are immediately terminated when the server
10381 goes down. It might be used if the health check detects more complex cases
10382 than a simple connection status, and long timeouts would cause the service
10383 to remain unresponsive for too long a time. For instance, a health check
10384 might detect that a database is stuck and that there's no chance to reuse
10385 existing connections anymore. Connections killed this way are logged with
10386 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +090010387
10388 Actions are disabled by default
10389
10390 Supported in default-server: Yes
10391
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070010392on-marked-up <action>
10393 Modify what occurs when a server is marked up.
10394 Currently one action is available:
10395 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
10396 done only if the server is not in backup state and if it is not disabled
10397 (it must have an effective weight > 0). This can be used sometimes to force
10398 an active server to take all the traffic back after recovery when dealing
10399 with long sessions (eg: LDAP, SQL, ...). Doing this can cause more trouble
10400 than it tries to solve (eg: incomplete transactions), so use this feature
10401 with extreme care. Sessions killed because a server comes up are logged
10402 with an 'U' termination code (for "Up").
10403
10404 Actions are disabled by default
10405
10406 Supported in default-server: Yes
10407
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010408port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010409 Using the "port" parameter, it becomes possible to use a different port to
10410 send health-checks. On some servers, it may be desirable to dedicate a port
10411 to a specific component able to perform complex tests which are more suitable
10412 to health-checks than the application. It is common to run a simple script in
10413 inetd for instance. This parameter is ignored if the "check" parameter is not
10414 set. See also the "addr" parameter.
10415
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010416 Supported in default-server: Yes
10417
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010418redir <prefix>
10419 The "redir" parameter enables the redirection mode for all GET and HEAD
10420 requests addressing this server. This means that instead of having HAProxy
10421 forward the request to the server, it will send an "HTTP 302" response with
10422 the "Location" header composed of this prefix immediately followed by the
10423 requested URI beginning at the leading '/' of the path component. That means
10424 that no trailing slash should be used after <prefix>. All invalid requests
10425 will be rejected, and all non-GET or HEAD requests will be normally served by
10426 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010427 mangling nor cookie insertion is possible in the response. However, cookies in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010428 requests are still analysed, making this solution completely usable to direct
10429 users to a remote location in case of local disaster. Main use consists in
10430 increasing bandwidth for static servers by having the clients directly
10431 connect to them. Note: never use a relative location here, it would cause a
10432 loop between the client and HAProxy!
10433
10434 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
10435
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010436 Supported in default-server: No
10437
10438rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010439 The "rise" parameter states that a server will be considered as operational
10440 after <count> consecutive successful health checks. This value defaults to 2
10441 if unspecified. See also the "check", "inter" and "fall" parameters.
10442
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010443 Supported in default-server: Yes
10444
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010445resolve-prefer <family>
10446 When DNS resolution is enabled for a server and multiple IP addresses from
10447 different families are returned, HAProxy will prefer using an IP address
10448 from the family mentioned in the "resolve-prefer" parameter.
10449 Available families: "ipv4" and "ipv6"
10450
Baptiste Assmannc4aabae2015-08-04 22:43:06 +020010451 Default value: ipv6
10452
10453 Supported in default-server: Yes
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010454
10455 Example: server s1 app1.domain.com:80 resolvers mydns resolve-prefer ipv6
10456
10457resolvers <id>
10458 Points to an existing "resolvers" section to resolve current server's
10459 hostname.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020010460 In order to be operational, DNS resolution requires that health check is
10461 enabled on the server. Actually, health checks triggers the DNS resolution.
10462 You must precise one 'resolvers' parameter on each server line where DNS
10463 resolution is required.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010464
Baptiste Assmann62b75b42015-09-09 01:11:36 +020010465 Supported in default-server: No
10466
10467 Example: server s1 app1.domain.com:80 check resolvers mydns
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010468
10469 See also chapter 5.3
10470
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010010471send-proxy
10472 The "send-proxy" parameter enforces use of the PROXY protocol over any
10473 connection established to this server. The PROXY protocol informs the other
10474 end about the layer 3/4 addresses of the incoming connection, so that it can
10475 know the client's address or the public address it accessed to, whatever the
10476 upper layer protocol. For connections accepted by an "accept-proxy" listener,
10477 the advertised address will be used. Only TCPv4 and TCPv6 address families
10478 are supported. Other families such as Unix sockets, will report an UNKNOWN
10479 family. Servers using this option can fully be chained to another instance of
10480 haproxy listening with an "accept-proxy" setting. This setting must not be
Willy Tarreau6c16adc2012-10-05 00:04:16 +020010481 used if the server isn't aware of the protocol. When health checks are sent
10482 to the server, the PROXY protocol is automatically used when this option is
10483 set, unless there is an explicit "port" or "addr" directive, in which case an
10484 explicit "check-send-proxy" directive would also be needed to use the PROXY
10485 protocol. See also the "accept-proxy" option of the "bind" keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010010486
10487 Supported in default-server: No
10488
David Safb76832014-05-08 23:42:08 -040010489send-proxy-v2
10490 The "send-proxy-v2" parameter enforces use of the PROXY protocol version 2
10491 over any connection established to this server. The PROXY protocol informs
10492 the other end about the layer 3/4 addresses of the incoming connection, so
10493 that it can know the client's address or the public address it accessed to,
10494 whatever the upper layer protocol. This setting must not be used if the
10495 server isn't aware of this version of the protocol. See also the "send-proxy"
10496 option of the "bind" keyword.
10497
10498 Supported in default-server: No
10499
10500send-proxy-v2-ssl
10501 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
10502 2 over any connection established to this server. The PROXY protocol informs
10503 the other end about the layer 3/4 addresses of the incoming connection, so
10504 that it can know the client's address or the public address it accessed to,
10505 whatever the upper layer protocol. In addition, the SSL information extension
10506 of the PROXY protocol is added to the PROXY protocol header. This setting
10507 must not be used if the server isn't aware of this version of the protocol.
10508 See also the "send-proxy-v2" option of the "bind" keyword.
10509
10510 Supported in default-server: No
10511
10512send-proxy-v2-ssl-cn
10513 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
10514 2 over any connection established to this server. The PROXY protocol informs
10515 the other end about the layer 3/4 addresses of the incoming connection, so
10516 that it can know the client's address or the public address it accessed to,
10517 whatever the upper layer protocol. In addition, the SSL information extension
10518 of the PROXY protocol, along along with the Common Name from the subject of
10519 the client certificate (if any), is added to the PROXY protocol header. This
10520 setting must not be used if the server isn't aware of this version of the
10521 protocol. See also the "send-proxy-v2" option of the "bind" keyword.
10522
10523 Supported in default-server: No
10524
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010525slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010526 The "slowstart" parameter for a server accepts a value in milliseconds which
10527 indicates after how long a server which has just come back up will run at
10528 full speed. Just as with every other time-based parameter, it can be entered
10529 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
10530 linearly from 0 to 100% during this time. The limitation applies to two
10531 parameters :
10532
10533 - maxconn: the number of connections accepted by the server will grow from 1
10534 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
10535
10536 - weight: when the backend uses a dynamic weighted algorithm, the weight
10537 grows linearly from 1 to 100%. In this case, the weight is updated at every
10538 health-check. For this reason, it is important that the "inter" parameter
10539 is smaller than the "slowstart", in order to maximize the number of steps.
10540
10541 The slowstart never applies when haproxy starts, otherwise it would cause
10542 trouble to running servers. It only applies when a server has been previously
10543 seen as failed.
10544
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010545 Supported in default-server: Yes
10546
Willy Tarreau732eac42015-07-09 11:40:25 +020010547sni <expression>
10548 The "sni" parameter evaluates the sample fetch expression, converts it to a
10549 string and uses the result as the host name sent in the SNI TLS extension to
10550 the server. A typical use case is to send the SNI received from the client in
10551 a bridged HTTPS scenario, using the "ssl_fc_sni" sample fetch for the
10552 expression, though alternatives such as req.hdr(host) can also make sense.
10553
10554 Supported in default-server: no
10555
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020010556source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +020010557source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020010558source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010559 The "source" parameter sets the source address which will be used when
10560 connecting to the server. It follows the exact same parameters and principle
10561 as the backend "source" keyword, except that it only applies to the server
10562 referencing it. Please consult the "source" keyword for details.
10563
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020010564 Additionally, the "source" statement on a server line allows one to specify a
10565 source port range by indicating the lower and higher bounds delimited by a
10566 dash ('-'). Some operating systems might require a valid IP address when a
10567 source port range is specified. It is permitted to have the same IP/range for
10568 several servers. Doing so makes it possible to bypass the maximum of 64k
10569 total concurrent connections. The limit will then reach 64k connections per
10570 server.
10571
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010572 Supported in default-server: No
10573
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010574ssl
Willy Tarreau44f65392013-06-25 07:56:20 +020010575 This option enables SSL ciphering on outgoing connections to the server. It
10576 is critical to verify server certificates using "verify" when using SSL to
10577 connect to servers, otherwise the communication is prone to trivial man in
10578 the-middle attacks rendering SSL useless. When this option is used, health
10579 checks are automatically sent in SSL too unless there is a "port" or an
10580 "addr" directive indicating the check should be sent to a different location.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010581 See the "check-ssl" option to force SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +020010582
10583 Supported in default-server: No
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010584
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010585track [<proxy>/]<server>
Willy Tarreau32091232014-05-16 13:52:00 +020010586 This option enables ability to set the current state of the server by tracking
10587 another one. It is possible to track a server which itself tracks another
10588 server, provided that at the end of the chain, a server has health checks
10589 enabled. If <proxy> is omitted the current one is used. If disable-on-404 is
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010590 used, it has to be enabled on both proxies.
10591
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010592 Supported in default-server: No
10593
Emeric Brunef42d922012-10-11 16:11:36 +020010594verify [none|required]
10595 This setting is only available when support for OpenSSL was built in. If set
Emeric Brun850efd52014-01-29 12:24:34 +010010596 to 'none', server certificate is not verified. In the other case, The
10597 certificate provided by the server is verified using CAs from 'ca-file'
10598 and optional CRLs from 'crl-file'. If 'ssl_server_verify' is not specified
10599 in global section, this is the default. On verify failure the handshake
Willy Tarreau44f65392013-06-25 07:56:20 +020010600 is aborted. It is critically important to verify server certificates when
10601 using SSL to connect to servers, otherwise the communication is prone to
10602 trivial man-in-the-middle attacks rendering SSL totally useless.
Emeric Brunef42d922012-10-11 16:11:36 +020010603
10604 Supported in default-server: No
10605
Evan Broderbe554312013-06-27 00:05:25 -070010606verifyhost <hostname>
10607 This setting is only available when support for OpenSSL was built in, and
10608 only takes effect if 'verify required' is also specified. When set, the
10609 hostnames in the subject and subjectAlternateNames of the certificate
10610 provided by the server are checked. If none of the hostnames in the
10611 certificate match the specified hostname, the handshake is aborted. The
10612 hostnames in the server-provided certificate may include wildcards.
10613
10614 Supported in default-server: No
10615
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010616weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010617 The "weight" parameter is used to adjust the server's weight relative to
10618 other servers. All servers will receive a load proportional to their weight
10619 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +020010620 load. The default weight is 1, and the maximal value is 256. A value of 0
10621 means the server will not participate in load-balancing but will still accept
10622 persistent connections. If this parameter is used to distribute the load
10623 according to server's capacity, it is recommended to start with values which
10624 can both grow and shrink, for instance between 10 and 100 to leave enough
10625 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010626
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010627 Supported in default-server: Yes
10628
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010629
Cyril Bonté46175dd2015-07-02 22:45:32 +0200106305.3. Server IP address resolution using DNS
10631-------------------------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010632
Baptiste Assmann62b75b42015-09-09 01:11:36 +020010633HAProxy allows using a host name on the server line to retrieve its IP address
10634using name servers. By default, HAProxy resolves the name when parsing the
10635configuration file, at startup and cache the result for the process' life.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010636This is not sufficient in some cases, such as in Amazon where a server's IP
10637can change after a reboot or an ELB Virtual IP can change based on current
10638workload.
10639This chapter describes how HAProxy can be configured to process server's name
10640resolution at run time.
10641Whether run time server name resolution has been enable or not, HAProxy will
10642carry on doing the first resolution when parsing the configuration.
10643
Baptiste Assmann62b75b42015-09-09 01:11:36 +020010644Bear in mind that DNS resolution is triggered by health checks. This makes
10645health checks mandatory to allow DNS resolution.
10646
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010647
Cyril Bonté46175dd2015-07-02 22:45:32 +0200106485.3.1. Global overview
10649----------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010650
10651As we've seen in introduction, name resolution in HAProxy occurs at two
10652different steps of the process life:
10653
10654 1. when starting up, HAProxy parses the server line definition and matches a
10655 host name. It uses libc functions to get the host name resolved. This
10656 resolution relies on /etc/resolv.conf file.
10657
10658 2. at run time, when HAProxy gets prepared to run a health check on a server,
10659 it verifies if the current name resolution is still considered as valid.
10660 If not, it processes a new resolution, in parallel of the health check.
10661
10662A few other events can trigger a name resolution at run time:
10663 - when a server's health check ends up in a connection timeout: this may be
10664 because the server has a new IP address. So we need to trigger a name
10665 resolution to know this new IP.
10666
10667A few things important to notice:
10668 - all the name servers are queried in the mean time. HAProxy will process the
10669 first valid response.
10670
10671 - a resolution is considered as invalid (NX, timeout, refused), when all the
10672 servers return an error.
10673
10674
Cyril Bonté46175dd2015-07-02 22:45:32 +0200106755.3.2. The resolvers section
10676----------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010677
10678This section is dedicated to host information related to name resolution in
10679HAProxy.
10680There can be as many as resolvers section as needed. Each section can contain
10681many name servers.
10682
Baptiste Assmann62b75b42015-09-09 01:11:36 +020010683When multiple name servers are configured in a resolvers section, then HAProxy
10684uses the first valid response. In case of invalid responses, only the last one
10685is treated. Purpose is to give the chance to a slow server to deliver a valid
10686answer after a fast faulty or outdated server.
10687
10688When each server returns a different error type, then only the last error is
10689used by HAProxy to decide what type of behavior to apply.
10690
10691Two types of behavior can be applied:
10692 1. stop DNS resolution
10693 2. replay the DNS query with a new query type
10694 In such case, the following types are applied in this exact order:
10695 1. ANY query type
10696 2. query type corresponding to family pointed by resolve-prefer
10697 server's parameter
10698 3. remaining family type
10699
10700HAProxy stops DNS resolution when the following errors occur:
10701 - invalid DNS response packet
10702 - wrong name in the query section of the response
10703 - NX domain
10704 - Query refused by server
10705 - CNAME not pointing to an IP address
10706
10707HAProxy tries a new query type when the following errors occur:
10708 - no Answer records in the response
10709 - DNS response truncated
10710 - Error in DNS response
10711 - No expected DNS records found in the response
10712 - name server timeout
10713
10714For example, with 2 name servers configured in a resolvers section:
10715 - first response is valid and is applied directly, second response is ignored
10716 - first response is invalid and second one is valid, then second response is
10717 applied;
10718 - first response is a NX domain and second one a truncated response, then
10719 HAProxy replays the query with a new type;
10720 - first response is truncated and second one is a NX Domain, then HAProxy
10721 stops resolution.
10722
10723
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010724resolvers <resolvers id>
10725 Creates a new name server list labelled <resolvers id>
10726
10727A resolvers section accept the following parameters:
10728
10729nameserver <id> <ip>:<port>
10730 DNS server description:
10731 <id> : label of the server, should be unique
10732 <ip> : IP address of the server
10733 <port> : port where the DNS service actually runs
10734
10735hold <status> <period>
10736 Defines <period> during which the last name resolution should be kept based
10737 on last resolution <status>
10738 <status> : last name resolution status. Only "valid" is accepted for now.
10739 <period> : interval between two successive name resolution when the last
10740 answer was in <status>. It follows the HAProxy time format.
10741 <period> is in milliseconds by default.
10742
10743 Default value is 10s for "valid".
10744
10745 Note: since the name resolution is triggered by the health checks, a new
10746 resolution is triggered after <period> modulo the <inter> parameter of
10747 the healch check.
10748
10749resolve_retries <nb>
10750 Defines the number <nb> of queries to send to resolve a server name before
10751 giving up.
10752 Default value: 3
10753
Baptiste Assmann62b75b42015-09-09 01:11:36 +020010754 A retry occurs on name server timeout or when the full sequence of DNS query
10755 type failover is over and we need to start up from the default ANY query
10756 type.
10757
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010758timeout <event> <time>
10759 Defines timeouts related to name resolution
10760 <event> : the event on which the <time> timeout period applies to.
10761 events available are:
10762 - retry: time between two DNS queries, when no response have
10763 been received.
10764 Default value: 1s
10765 <time> : time related to the event. It follows the HAProxy time format.
10766 <time> is expressed in milliseconds.
10767
10768Example of a resolvers section (with default values):
10769
10770 resolvers mydns
10771 nameserver dns1 10.0.0.1:53
10772 nameserver dns2 10.0.0.2:53
10773 resolve_retries 3
10774 timeout retry 1s
10775 hold valid 10s
10776
10777
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200107786. HTTP header manipulation
10779---------------------------
10780
10781In HTTP mode, it is possible to rewrite, add or delete some of the request and
10782response headers based on regular expressions. It is also possible to block a
10783request or a response if a particular header matches a regular expression,
10784which is enough to stop most elementary protocol attacks, and to protect
Willy Tarreau70dffda2014-01-30 03:07:23 +010010785against information leak from the internal network.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010786
Willy Tarreau70dffda2014-01-30 03:07:23 +010010787If HAProxy encounters an "Informational Response" (status code 1xx), it is able
10788to process all rsp* rules which can allow, deny, rewrite or delete a header,
10789but it will refuse to add a header to any such messages as this is not
10790HTTP-compliant. The reason for still processing headers in such responses is to
10791stop and/or fix any possible information leak which may happen, for instance
10792because another downstream equipment would unconditionally add a header, or if
10793a server name appears there. When such messages are seen, normal processing
10794still occurs on the next non-informational messages.
Willy Tarreau816b9792009-09-15 21:25:21 +020010795
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010796This section covers common usage of the following keywords, described in detail
10797in section 4.2 :
10798
10799 - reqadd <string>
10800 - reqallow <search>
10801 - reqiallow <search>
10802 - reqdel <search>
10803 - reqidel <search>
10804 - reqdeny <search>
10805 - reqideny <search>
10806 - reqpass <search>
10807 - reqipass <search>
10808 - reqrep <search> <replace>
10809 - reqirep <search> <replace>
10810 - reqtarpit <search>
10811 - reqitarpit <search>
10812 - rspadd <string>
10813 - rspdel <search>
10814 - rspidel <search>
10815 - rspdeny <search>
10816 - rspideny <search>
10817 - rsprep <search> <replace>
10818 - rspirep <search> <replace>
10819
10820With all these keywords, the same conventions are used. The <search> parameter
10821is a POSIX extended regular expression (regex) which supports grouping through
10822parenthesis (without the backslash). Spaces and other delimiters must be
10823prefixed with a backslash ('\') to avoid confusion with a field delimiter.
10824Other characters may be prefixed with a backslash to change their meaning :
10825
10826 \t for a tab
10827 \r for a carriage return (CR)
10828 \n for a new line (LF)
10829 \ to mark a space and differentiate it from a delimiter
10830 \# to mark a sharp and differentiate it from a comment
10831 \\ to use a backslash in a regex
10832 \\\\ to use a backslash in the text (*2 for regex, *2 for haproxy)
10833 \xXX to write the ASCII hex code XX as in the C language
10834
10835The <replace> parameter contains the string to be used to replace the largest
10836portion of text matching the regex. It can make use of the special characters
10837above, and can reference a substring which is delimited by parenthesis in the
10838regex, by writing a backslash ('\') immediately followed by one digit from 0 to
108399 indicating the group position (0 designating the entire line). This practice
10840is very common to users of the "sed" program.
10841
10842The <string> parameter represents the string which will systematically be added
10843after the last header line. It can also use special character sequences above.
10844
10845Notes related to these keywords :
10846---------------------------------
10847 - these keywords are not always convenient to allow/deny based on header
10848 contents. It is strongly recommended to use ACLs with the "block" keyword
10849 instead, resulting in far more flexible and manageable rules.
10850
10851 - lines are always considered as a whole. It is not possible to reference
10852 a header name only or a value only. This is important because of the way
10853 headers are written (notably the number of spaces after the colon).
10854
10855 - the first line is always considered as a header, which makes it possible to
10856 rewrite or filter HTTP requests URIs or response codes, but in turn makes
10857 it harder to distinguish between headers and request line. The regex prefix
10858 ^[^\ \t]*[\ \t] matches any HTTP method followed by a space, and the prefix
10859 ^[^ \t:]*: matches any header name followed by a colon.
10860
10861 - for performances reasons, the number of characters added to a request or to
10862 a response is limited at build time to values between 1 and 4 kB. This
10863 should normally be far more than enough for most usages. If it is too short
10864 on occasional usages, it is possible to gain some space by removing some
10865 useless headers before adding new ones.
10866
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010867 - keywords beginning with "reqi" and "rspi" are the same as their counterpart
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010868 without the 'i' letter except that they ignore case when matching patterns.
10869
10870 - when a request passes through a frontend then a backend, all req* rules
10871 from the frontend will be evaluated, then all req* rules from the backend
10872 will be evaluated. The reverse path is applied to responses.
10873
10874 - req* statements are applied after "block" statements, so that "block" is
10875 always the first one, but before "use_backend" in order to permit rewriting
Willy Tarreaud72758d2010-01-12 10:42:19 +010010876 before switching.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010877
10878
Willy Tarreau74ca5042013-06-11 23:12:07 +0200108797. Using ACLs and fetching samples
10880----------------------------------
10881
10882Haproxy is capable of extracting data from request or response streams, from
10883client or server information, from tables, environmental information etc...
10884The action of extracting such data is called fetching a sample. Once retrieved,
10885these samples may be used for various purposes such as a key to a stick-table,
10886but most common usages consist in matching them against predefined constant
10887data called patterns.
10888
10889
108907.1. ACL basics
10891---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010892
10893The use of Access Control Lists (ACL) provides a flexible solution to perform
10894content switching and generally to take decisions based on content extracted
10895from the request, the response or any environmental status. The principle is
10896simple :
10897
Willy Tarreau74ca5042013-06-11 23:12:07 +020010898 - extract a data sample from a stream, table or the environment
Willy Tarreaue6b11e42013-11-26 19:02:32 +010010899 - optionally apply some format conversion to the extracted sample
Willy Tarreau74ca5042013-06-11 23:12:07 +020010900 - apply one or multiple pattern matching methods on this sample
10901 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010902
Willy Tarreau74ca5042013-06-11 23:12:07 +020010903The actions generally consist in blocking a request, selecting a backend, or
10904adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010905
10906In order to define a test, the "acl" keyword is used. The syntax is :
10907
Willy Tarreau74ca5042013-06-11 23:12:07 +020010908 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010909
10910This creates a new ACL <aclname> or completes an existing one with new tests.
10911Those tests apply to the portion of request/response specified in <criterion>
10912and may be adjusted with optional flags [flags]. Some criteria also support
Willy Tarreaue6b11e42013-11-26 19:02:32 +010010913an operator which may be specified before the set of values. Optionally some
10914conversion operators may be applied to the sample, and they will be specified
10915as a comma-delimited list of keywords just after the first keyword. The values
10916are of the type supported by the criterion, and are separated by spaces.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010917
10918ACL names must be formed from upper and lower case letters, digits, '-' (dash),
10919'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
10920which means that "my_acl" and "My_Acl" are two different ACLs.
10921
10922There is no enforced limit to the number of ACLs. The unused ones do not affect
10923performance, they just consume a small amount of memory.
10924
Willy Tarreau74ca5042013-06-11 23:12:07 +020010925The criterion generally is the name of a sample fetch method, or one of its ACL
10926specific declinations. The default test method is implied by the output type of
10927this sample fetch method. The ACL declinations can describe alternate matching
Willy Tarreaue6b11e42013-11-26 19:02:32 +010010928methods of a same sample fetch method. The sample fetch methods are the only
10929ones supporting a conversion.
Willy Tarreau74ca5042013-06-11 23:12:07 +020010930
10931Sample fetch methods return data which can be of the following types :
10932 - boolean
10933 - integer (signed or unsigned)
10934 - IPv4 or IPv6 address
10935 - string
10936 - data block
10937
Willy Tarreaue6b11e42013-11-26 19:02:32 +010010938Converters transform any of these data into any of these. For example, some
10939converters might convert a string to a lower-case string while other ones
10940would turn a string to an IPv4 address, or apply a netmask to an IP address.
10941The resulting sample is of the type of the last converter applied to the list,
10942which defaults to the type of the sample fetch method.
10943
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020010944Each sample or converter returns data of a specific type, specified with its
10945keyword in this documentation. When an ACL is declared using a standard sample
10946fetch method, certain types automatically involved a default matching method
10947which are summarized in the table below :
10948
10949 +---------------------+-----------------+
10950 | Sample or converter | Default |
10951 | output type | matching method |
10952 +---------------------+-----------------+
10953 | boolean | bool |
10954 +---------------------+-----------------+
10955 | integer | int |
10956 +---------------------+-----------------+
10957 | ip | ip |
10958 +---------------------+-----------------+
10959 | string | str |
10960 +---------------------+-----------------+
10961 | binary | none, use "-m" |
10962 +---------------------+-----------------+
10963
10964Note that in order to match a binary samples, it is mandatory to specify a
10965matching method, see below.
10966
Willy Tarreau74ca5042013-06-11 23:12:07 +020010967The ACL engine can match these types against patterns of the following types :
10968 - boolean
10969 - integer or integer range
10970 - IP address / network
10971 - string (exact, substring, suffix, prefix, subdir, domain)
10972 - regular expression
10973 - hex block
10974
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010975The following ACL flags are currently supported :
10976
Willy Tarreau2b5285d2010-05-09 23:45:24 +020010977 -i : ignore case during matching of all subsequent patterns.
10978 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +020010979 -m : use a specific pattern matching method
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010010980 -n : forbid the DNS resolutions
Thierry FOURNIER9860c412014-01-29 14:23:29 +010010981 -M : load the file pointed by -f like a map file.
Thierry FOURNIER3534d882014-01-20 17:01:44 +010010982 -u : force the unique id of the ACL
Willy Tarreau6a06a402007-07-15 20:15:28 +020010983 -- : force end of flags. Useful when a string looks like one of the flags.
10984
Willy Tarreau74ca5042013-06-11 23:12:07 +020010985The "-f" flag is followed by the name of a file from which all lines will be
10986read as individual values. It is even possible to pass multiple "-f" arguments
10987if the patterns are to be loaded from multiple files. Empty lines as well as
10988lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
10989will be stripped. If it is absolutely necessary to insert a valid pattern
10990beginning with a sharp, just prefix it with a space so that it is not taken for
10991a comment. Depending on the data type and match method, haproxy may load the
10992lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
10993exact string matching. In this case, duplicates will automatically be removed.
10994
Thierry FOURNIER9860c412014-01-29 14:23:29 +010010995The "-M" flag allows an ACL to use a map file. If this flag is set, the file is
10996parsed as two column file. The first column contains the patterns used by the
10997ACL, and the second column contain the samples. The sample can be used later by
10998a map. This can be useful in some rare cases where an ACL would just be used to
10999check for the existence of a pattern in a map before a mapping is applied.
11000
Thierry FOURNIER3534d882014-01-20 17:01:44 +010011001The "-u" flag forces the unique id of the ACL. This unique id is used with the
11002socket interface to identify ACL and dynamically change its values. Note that a
11003file is always identified by its name even if an id is set.
11004
Willy Tarreau74ca5042013-06-11 23:12:07 +020011005Also, note that the "-i" flag applies to subsequent entries and not to entries
11006loaded from files preceding it. For instance :
11007
11008 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
11009
11010In this example, each line of "exact-ua.lst" will be exactly matched against
11011the "user-agent" header of the request. Then each line of "generic-ua" will be
11012case-insensitively matched. Then the word "test" will be insensitively matched
11013as well.
11014
11015The "-m" flag is used to select a specific pattern matching method on the input
11016sample. All ACL-specific criteria imply a pattern matching method and generally
11017do not need this flag. However, this flag is useful with generic sample fetch
11018methods to describe how they're going to be matched against the patterns. This
11019is required for sample fetches which return data type for which there is no
11020obvious matching method (eg: string or binary). When "-m" is specified and
11021followed by a pattern matching method name, this method is used instead of the
11022default one for the criterion. This makes it possible to match contents in ways
11023that were not initially planned, or with sample fetch methods which return a
11024string. The matching method also affects the way the patterns are parsed.
11025
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010011026The "-n" flag forbids the dns resolutions. It is used with the load of ip files.
11027By default, if the parser cannot parse ip address it considers that the parsed
11028string is maybe a domain name and try dns resolution. The flag "-n" disable this
11029resolution. It is useful for detecting malformed ip lists. Note that if the DNS
11030server is not reachable, the haproxy configuration parsing may last many minutes
11031waiting fir the timeout. During this time no error messages are displayed. The
11032flag "-n" disable this behavior. Note also that during the runtime, this
11033function is disabled for the dynamic acl modifications.
11034
Willy Tarreau74ca5042013-06-11 23:12:07 +020011035There are some restrictions however. Not all methods can be used with all
11036sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
11037be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +020011038
11039 - "found" : only check if the requested sample could be found in the stream,
11040 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +020011041 to pass any pattern to avoid confusion. This matching method is
11042 particularly useful to detect presence of certain contents such
11043 as headers, cookies, etc... even if they are empty and without
11044 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011045
11046 - "bool" : check the value as a boolean. It can only be applied to fetches
11047 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011048 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011049
11050 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +020011051 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011052
11053 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +020011054 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011055
11056 - "bin" : match the contents against an hexadecimal string representing a
11057 binary sequence. This may be used with binary or string samples.
11058
11059 - "len" : match the sample's length as an integer. This may be used with
11060 binary or string samples.
11061
Willy Tarreau74ca5042013-06-11 23:12:07 +020011062 - "str" : exact match : match the contents against a string. This may be
11063 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011064
Willy Tarreau74ca5042013-06-11 23:12:07 +020011065 - "sub" : substring match : check that the contents contain at least one of
11066 the provided string patterns. This may be used with binary or
11067 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011068
Willy Tarreau74ca5042013-06-11 23:12:07 +020011069 - "reg" : regex match : match the contents against a list of regular
11070 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011071
Willy Tarreau74ca5042013-06-11 23:12:07 +020011072 - "beg" : prefix match : check that the contents begin like the provided
11073 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011074
Willy Tarreau74ca5042013-06-11 23:12:07 +020011075 - "end" : suffix match : check that the contents end like the provided
11076 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011077
Willy Tarreau74ca5042013-06-11 23:12:07 +020011078 - "dir" : subdir match : check that a slash-delimited portion of the
11079 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011080 This may be used with binary or string samples.
11081
Willy Tarreau74ca5042013-06-11 23:12:07 +020011082 - "dom" : domain match : check that a dot-delimited portion of the contents
11083 exactly match one of the provided string patterns. This may be
11084 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011085
11086For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
11087request, it is possible to do :
11088
11089 acl jsess_present cook(JSESSIONID) -m found
11090
11091In order to apply a regular expression on the 500 first bytes of data in the
11092buffer, one would use the following acl :
11093
11094 acl script_tag payload(0,500) -m reg -i <script>
11095
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011096On systems where the regex library is much slower when using "-i", it is
11097possible to convert the sample to lowercase before matching, like this :
11098
11099 acl script_tag payload(0,500),lower -m reg <script>
11100
Willy Tarreau74ca5042013-06-11 23:12:07 +020011101All ACL-specific criteria imply a default matching method. Most often, these
11102criteria are composed by concatenating the name of the original sample fetch
11103method and the matching method. For example, "hdr_beg" applies the "beg" match
11104to samples retrieved using the "hdr" fetch method. Since all ACL-specific
11105criteria rely on a sample fetch method, it is always possible instead to use
11106the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011107
Willy Tarreau74ca5042013-06-11 23:12:07 +020011108If an alternate match is specified using "-m" on an ACL-specific criterion,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011109the matching method is simply applied to the underlying sample fetch method.
11110For example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011111
Willy Tarreau74ca5042013-06-11 23:12:07 +020011112 acl short_form hdr_beg(host) www.
11113 acl alternate1 hdr_beg(host) -m beg www.
11114 acl alternate2 hdr_dom(host) -m beg www.
11115 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011116
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011117
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011118The table below summarizes the compatibility matrix between sample or converter
11119types and the pattern types to fetch against. It indicates for each compatible
11120combination the name of the matching method to be used, surrounded with angle
11121brackets ">" and "<" when the method is the default one and will work by
11122default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +010011123
Willy Tarreau74ca5042013-06-11 23:12:07 +020011124 +-------------------------------------------------+
11125 | Input sample type |
11126 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011127 | pattern type | boolean | integer | ip | string | binary |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011128 +----------------------+---------+---------+---------+---------+---------+
11129 | none (presence only) | found | found | found | found | found |
11130 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011131 | none (boolean value) |> bool <| bool | | bool | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011132 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011133 | integer (value) | int |> int <| int | int | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011134 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011135 | integer (length) | len | len | len | len | len |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011136 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011137 | IP address | | |> ip <| ip | ip |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011138 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011139 | exact string | str | str | str |> str <| str |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011140 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011141 | prefix | beg | beg | beg | beg | beg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011142 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011143 | suffix | end | end | end | end | end |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011144 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011145 | substring | sub | sub | sub | sub | sub |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011146 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011147 | subdir | dir | dir | dir | dir | dir |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011148 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011149 | domain | dom | dom | dom | dom | dom |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011150 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011151 | regex | reg | reg | reg | reg | reg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011152 +----------------------+---------+---------+---------+---------+---------+
11153 | hex block | | | | bin | bin |
11154 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +020011155
11156
Willy Tarreau74ca5042013-06-11 23:12:07 +0200111577.1.1. Matching booleans
11158------------------------
11159
11160In order to match a boolean, no value is needed and all values are ignored.
11161Boolean matching is used by default for all fetch methods of type "boolean".
11162When boolean matching is used, the fetched value is returned as-is, which means
11163that a boolean "true" will always match and a boolean "false" will never match.
11164
11165Boolean matching may also be enforced using "-m bool" on fetch methods which
11166return an integer value. Then, integer value 0 is converted to the boolean
11167"false" and all other values are converted to "true".
11168
Willy Tarreau6a06a402007-07-15 20:15:28 +020011169
Willy Tarreau74ca5042013-06-11 23:12:07 +0200111707.1.2. Matching integers
11171------------------------
11172
11173Integer matching applies by default to integer fetch methods. It can also be
11174enforced on boolean fetches using "-m int". In this case, "false" is converted
11175to the integer 0, and "true" is converted to the integer 1.
11176
11177Integer matching also supports integer ranges and operators. Note that integer
11178matching only applies to positive values. A range is a value expressed with a
11179lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011180
11181For instance, "1024:65535" is a valid range to represent a range of
11182unprivileged ports, and "1024:" would also work. "0:1023" is a valid
11183representation of privileged ports, and ":1023" would also work.
11184
Willy Tarreau62644772008-07-16 18:36:06 +020011185As a special case, some ACL functions support decimal numbers which are in fact
11186two integers separated by a dot. This is used with some version checks for
11187instance. All integer properties apply to those decimal numbers, including
11188ranges and operators.
11189
Willy Tarreau6a06a402007-07-15 20:15:28 +020011190For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +010011191operators with ranges does not make much sense and is strongly discouraged.
11192Similarly, it does not make much sense to perform order comparisons with a set
11193of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011194
Willy Tarreau0ba27502007-12-24 16:55:16 +010011195Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +020011196
11197 eq : true if the tested value equals at least one value
11198 ge : true if the tested value is greater than or equal to at least one value
11199 gt : true if the tested value is greater than at least one value
11200 le : true if the tested value is less than or equal to at least one value
11201 lt : true if the tested value is less than at least one value
11202
Willy Tarreau0ba27502007-12-24 16:55:16 +010011203For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +020011204
11205 acl negative-length hdr_val(content-length) lt 0
11206
Willy Tarreau62644772008-07-16 18:36:06 +020011207This one matches SSL versions between 3.0 and 3.1 (inclusive) :
11208
11209 acl sslv3 req_ssl_ver 3:3.1
11210
Willy Tarreau6a06a402007-07-15 20:15:28 +020011211
Willy Tarreau74ca5042013-06-11 23:12:07 +0200112127.1.3. Matching strings
11213-----------------------
11214
11215String matching applies to string or binary fetch methods, and exists in 6
11216different forms :
11217
11218 - exact match (-m str) : the extracted string must exactly match the
11219 patterns ;
11220
11221 - substring match (-m sub) : the patterns are looked up inside the
11222 extracted string, and the ACL matches if any of them is found inside ;
11223
11224 - prefix match (-m beg) : the patterns are compared with the beginning of
11225 the extracted string, and the ACL matches if any of them matches.
11226
11227 - suffix match (-m end) : the patterns are compared with the end of the
11228 extracted string, and the ACL matches if any of them matches.
11229
11230 - subdir match (-m sub) : the patterns are looked up inside the extracted
11231 string, delimited with slashes ("/"), and the ACL matches if any of them
11232 matches.
11233
11234 - domain match (-m dom) : the patterns are looked up inside the extracted
11235 string, delimited with dots ("."), and the ACL matches if any of them
11236 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011237
11238String matching applies to verbatim strings as they are passed, with the
11239exception of the backslash ("\") which makes it possible to escape some
11240characters such as the space. If the "-i" flag is passed before the first
11241string, then the matching will be performed ignoring the case. In order
11242to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +010011243before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020011244
11245
Willy Tarreau74ca5042013-06-11 23:12:07 +0200112467.1.4. Matching regular expressions (regexes)
11247---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020011248
11249Just like with string matching, regex matching applies to verbatim strings as
11250they are passed, with the exception of the backslash ("\") which makes it
11251possible to escape some characters such as the space. If the "-i" flag is
11252passed before the first regex, then the matching will be performed ignoring
11253the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +010011254the "--" flag before the first string. Same principle applies of course to
11255match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020011256
11257
Willy Tarreau74ca5042013-06-11 23:12:07 +0200112587.1.5. Matching arbitrary data blocks
11259-------------------------------------
11260
11261It is possible to match some extracted samples against a binary block which may
11262not safely be represented as a string. For this, the patterns must be passed as
11263a series of hexadecimal digits in an even number, when the match method is set
11264to binary. Each sequence of two digits will represent a byte. The hexadecimal
11265digits may be used upper or lower case.
11266
11267Example :
11268 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
11269 acl hello payload(0,6) -m bin 48656c6c6f0a
11270
11271
112727.1.6. Matching IPv4 and IPv6 addresses
11273---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020011274
11275IPv4 addresses values can be specified either as plain addresses or with a
11276netmask appended, in which case the IPv4 address matches whenever it is
11277within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010011278host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +010011279difficult to read and debug configurations. If hostnames are used, you should
11280at least ensure that they are present in /etc/hosts so that the configuration
11281does not depend on any random DNS match at the moment the configuration is
11282parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011283
Willy Tarreauceb4ac92012-04-28 00:41:46 +020011284IPv6 may be entered in their usual form, with or without a netmask appended.
11285Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
11286trouble with randomly resolved IP addresses, host names are never allowed in
11287IPv6 patterns.
11288
11289HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
11290following situations :
11291 - tested address is IPv4, pattern address is IPv4, the match applies
11292 in IPv4 using the supplied mask if any.
11293 - tested address is IPv6, pattern address is IPv6, the match applies
11294 in IPv6 using the supplied mask if any.
11295 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
11296 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
11297 ::IPV4 or ::ffff:IPV4, otherwise it fails.
11298 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
11299 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
11300 applied in IPv6 using the supplied IPv6 mask.
11301
Willy Tarreau74ca5042013-06-11 23:12:07 +020011302
113037.2. Using ACLs to form conditions
11304----------------------------------
11305
11306Some actions are only performed upon a valid condition. A condition is a
11307combination of ACLs with operators. 3 operators are supported :
11308
11309 - AND (implicit)
11310 - OR (explicit with the "or" keyword or the "||" operator)
11311 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +020011312
Willy Tarreau74ca5042013-06-11 23:12:07 +020011313A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +020011314
Willy Tarreau74ca5042013-06-11 23:12:07 +020011315 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +020011316
Willy Tarreau74ca5042013-06-11 23:12:07 +020011317Such conditions are generally used after an "if" or "unless" statement,
11318indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +020011319
Willy Tarreau74ca5042013-06-11 23:12:07 +020011320For instance, to block HTTP requests to the "*" URL with methods other than
11321"OPTIONS", as well as POST requests without content-length, and GET or HEAD
11322requests with a content-length greater than 0, and finally every request which
11323is not either GET/HEAD/POST/OPTIONS !
11324
11325 acl missing_cl hdr_cnt(Content-length) eq 0
11326 block if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
11327 block if METH_GET HTTP_CONTENT
11328 block unless METH_GET or METH_POST or METH_OPTIONS
11329
11330To select a different backend for requests to static contents on the "www" site
11331and to every request on the "img", "video", "download" and "ftp" hosts :
11332
11333 acl url_static path_beg /static /images /img /css
11334 acl url_static path_end .gif .png .jpg .css .js
11335 acl host_www hdr_beg(host) -i www
11336 acl host_static hdr_beg(host) -i img. video. download. ftp.
11337
11338 # now use backend "static" for all static-only hosts, and for static urls
11339 # of host "www". Use backend "www" for the rest.
11340 use_backend static if host_static or host_www url_static
11341 use_backend www if host_www
11342
11343It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
11344expressions that are built on the fly without needing to be declared. They must
11345be enclosed between braces, with a space before and after each brace (because
11346the braces must be seen as independent words). Example :
11347
11348 The following rule :
11349
11350 acl missing_cl hdr_cnt(Content-length) eq 0
11351 block if METH_POST missing_cl
11352
11353 Can also be written that way :
11354
11355 block if METH_POST { hdr_cnt(Content-length) eq 0 }
11356
11357It is generally not recommended to use this construct because it's a lot easier
11358to leave errors in the configuration when written that way. However, for very
11359simple rules matching only one source IP address for instance, it can make more
11360sense to use them than to declare ACLs with random names. Another example of
11361good use is the following :
11362
11363 With named ACLs :
11364
11365 acl site_dead nbsrv(dynamic) lt 2
11366 acl site_dead nbsrv(static) lt 2
11367 monitor fail if site_dead
11368
11369 With anonymous ACLs :
11370
11371 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
11372
11373See section 4.2 for detailed help on the "block" and "use_backend" keywords.
11374
11375
113767.3. Fetching samples
11377---------------------
11378
11379Historically, sample fetch methods were only used to retrieve data to match
11380against patterns using ACLs. With the arrival of stick-tables, a new class of
11381sample fetch methods was created, most often sharing the same syntax as their
11382ACL counterpart. These sample fetch methods are also known as "fetches". As
11383of now, ACLs and fetches have converged. All ACL fetch methods have been made
11384available as fetch methods, and ACLs may use any sample fetch method as well.
11385
11386This section details all available sample fetch methods and their output type.
11387Some sample fetch methods have deprecated aliases that are used to maintain
11388compatibility with existing configurations. They are then explicitly marked as
11389deprecated and should not be used in new setups.
11390
11391The ACL derivatives are also indicated when available, with their respective
11392matching methods. These ones all have a well defined default pattern matching
11393method, so it is never necessary (though allowed) to pass the "-m" option to
11394indicate how the sample will be matched using ACLs.
11395
11396As indicated in the sample type versus matching compatibility matrix above,
11397when using a generic sample fetch method in an ACL, the "-m" option is
11398mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
11399the same keyword exists as an ACL keyword and as a standard fetch method, the
11400ACL engine will automatically pick the ACL-only one by default.
11401
11402Some of these keywords support one or multiple mandatory arguments, and one or
11403multiple optional arguments. These arguments are strongly typed and are checked
11404when the configuration is parsed so that there is no risk of running with an
11405incorrect argument (eg: an unresolved backend name). Fetch function arguments
11406are passed between parenthesis and are delimited by commas. When an argument
11407is optional, it will be indicated below between square brackets ('[ ]'). When
11408all arguments are optional, the parenthesis may be omitted.
11409
11410Thus, the syntax of a standard sample fetch method is one of the following :
11411 - name
11412 - name(arg1)
11413 - name(arg1,arg2)
11414
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011415
114167.3.1. Converters
11417-----------------
11418
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011419Sample fetch methods may be combined with transformations to be applied on top
11420of the fetched sample (also called "converters"). These combinations form what
11421is called "sample expressions" and the result is a "sample". Initially this
11422was only supported by "stick on" and "stick store-request" directives but this
11423has now be extended to all places where samples may be used (acls, log-format,
11424unique-id-format, add-header, ...).
11425
11426These transformations are enumerated as a series of specific keywords after the
11427sample fetch method. These keywords may equally be appended immediately after
11428the fetch keyword's argument, delimited by a comma. These keywords can also
11429support some arguments (eg: a netmask) which must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +010011430
Willy Tarreau97707872015-01-27 15:12:13 +010011431A certain category of converters are bitwise and arithmetic operators which
11432support performing basic operations on integers. Some bitwise operations are
11433supported (and, or, xor, cpl) and some arithmetic operations are supported
11434(add, sub, mul, div, mod, neg). Some comparators are provided (odd, even, not,
11435bool) which make it possible to report a match without having to write an ACL.
11436
Willy Tarreau74ca5042013-06-11 23:12:07 +020011437The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +010011438
Willy Tarreau97707872015-01-27 15:12:13 +010011439add(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011440 Adds <value> to the input value of type signed integer, and returns the
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011441 result as a signed integer. <value> can be a numeric value or a variable
11442 name. The name of the variable starts by an indication about its scope. The
11443 allowed scopes are:
11444 "sess" : the variable is shared with all the session,
11445 "txn" : the variable is shared with all the transaction (request and
11446 response),
11447 "req" : the variable is shared only during the request processing,
11448 "res" : the variable is shared only during the response processing.
11449 This prefix is followed by a name. The separator is a '.'. The name may only
11450 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011451
11452and(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011453 Performs a bitwise "AND" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011454 integer, and returns the result as an signed integer. <value> can be a
11455 numeric value or a variable name. The name of the variable starts by an
11456 indication about its scope. The allowed scopes are:
11457 "sess" : the variable is shared with all the session,
11458 "txn" : the variable is shared with all the transaction (request and
11459 response),
11460 "req" : the variable is shared only during the request processing,
11461 "res" : the variable is shared only during the response processing.
11462 This prefix is followed by a name. The separator is a '.'. The name may only
11463 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011464
Emeric Brun53d1a982014-04-30 18:21:37 +020011465base64
11466 Converts a binary input sample to a base64 string. It is used to log or
11467 transfer binary content in a way that can be reliably transferred (eg:
11468 an SSL ID can be copied in a header).
11469
Willy Tarreau97707872015-01-27 15:12:13 +010011470bool
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011471 Returns a boolean TRUE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010011472 non-null, otherwise returns FALSE. Used in conjunction with and(), it can be
11473 used to report true/false for bit testing on input values (eg: verify the
11474 presence of a flag).
11475
Emeric Brun54c4ac82014-11-03 15:32:43 +010011476bytes(<offset>[,<length>])
11477 Extracts some bytes from an input binary sample. The result is a binary
11478 sample starting at an offset (in bytes) of the original sample and
11479 optionnaly truncated at the given length.
11480
Willy Tarreau97707872015-01-27 15:12:13 +010011481cpl
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011482 Takes the input value of type signed integer, applies a ones-complement
11483 (flips all bits) and returns the result as an signed integer.
Willy Tarreau97707872015-01-27 15:12:13 +010011484
Willy Tarreau80599772015-01-20 19:35:24 +010011485crc32([<avalanche>])
11486 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32
11487 hash function. Optionally, it is possible to apply a full avalanche hash
11488 function to the output if the optional <avalanche> argument equals 1. This
11489 converter uses the same functions as used by the various hash-based load
11490 balancing algorithms, so it will provide exactly the same results. It is
11491 provided for compatibility with other software which want a CRC32 to be
11492 computed on some input keys, so it follows the most common implementation as
11493 found in Ethernet, Gzip, PNG, etc... It is slower than the other algorithms
11494 but may provide a better or at least less predictable distribution. It must
11495 not be used for security purposes as a 32-bit hash is trivial to break. See
11496 also "djb2", "sdbm", "wt6" and the "hash-type" directive.
11497
David Carlier29b3ca32015-09-25 14:09:21 +010011498da-csv-conv(<prop>[,<prop>*])
David Carlier4542b102015-06-01 13:54:29 +020011499 Asks the DeviceAtlas converter to identify the User Agent string passed on
11500 input, and to emit a string made of the concatenation of the properties
11501 enumerated in argument, delimited by the separator defined by the global
11502 keyword "deviceatlas-property-separator", or by default the pipe character
11503 ('|'). There's a limit of 5 different properties imposed by the haproxy
11504 configuration language.
11505
11506 Example:
11507 frontend www
Cyril Bonté307ee1e2015-09-28 23:16:06 +020011508 bind *:8881
11509 default_backend servers
11510 http-request set-header X-DeviceAtlas-Data %[req.fhdr(User-Agent),da-csv(primaryHardwareType,osName,osVersion,browserName,browserVersion)]
David Carlier4542b102015-06-01 13:54:29 +020011511
Thierry FOURNIER9687c772015-05-07 15:46:29 +020011512debug
11513 This converter is used as debug tool. It dumps on screen the content and the
11514 type of the input sample. The sample is returned as is on its output. This
11515 converter only exists when haproxy was built with debugging enabled.
11516
Willy Tarreau97707872015-01-27 15:12:13 +010011517div(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011518 Divides the input value of type signed integer by <value>, and returns the
11519 result as an signed integer. If <value> is null, the largest unsigned
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011520 integer is returned (typically 2^63-1). <value> can be a numeric value or a
11521 variable name. The name of the variable starts by an indication about it
11522 scope. The scope allowed are:
11523 "sess" : the variable is shared with all the session,
11524 "txn" : the variable is shared with all the transaction (request and
11525 response),
11526 "req" : the variable is shared only during the request processing,
11527 "res" : the variable is shared only during the response processing.
11528 This prefix is followed by a name. The separator is a '.'. The name may only
11529 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011530
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020011531djb2([<avalanche>])
11532 Hashes a binary input sample into an unsigned 32-bit quantity using the DJB2
11533 hash function. Optionally, it is possible to apply a full avalanche hash
11534 function to the output if the optional <avalanche> argument equals 1. This
11535 converter uses the same functions as used by the various hash-based load
11536 balancing algorithms, so it will provide exactly the same results. It is
11537 mostly intended for debugging, but can be used as a stick-table entry to
11538 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010011539 32-bit hash is trivial to break. See also "crc32", "sdbm", "wt6" and the
11540 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020011541
Willy Tarreau97707872015-01-27 15:12:13 +010011542even
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011543 Returns a boolean TRUE if the input value of type signed integer is even
Willy Tarreau97707872015-01-27 15:12:13 +010011544 otherwise returns FALSE. It is functionally equivalent to "not,and(1),bool".
11545
Emeric Brunf399b0d2014-11-03 17:07:03 +010011546field(<index>,<delimiters>)
11547 Extracts the substring at the given index considering given delimiters from
11548 an input string. Indexes start at 1 and delimiters are a string formatted
11549 list of chars.
11550
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011551hex
11552 Converts a binary input sample to an hex string containing two hex digits per
11553 input byte. It is used to log or transfer hex dumps of some binary input data
11554 in a way that can be reliably transferred (eg: an SSL ID can be copied in a
11555 header).
Thierry FOURNIER2f49d6d2014-03-12 15:01:52 +010011556
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011557http_date([<offset>])
11558 Converts an integer supposed to contain a date since epoch to a string
11559 representing this date in a format suitable for use in HTTP header fields. If
11560 an offset value is specified, then it is a number of seconds that is added to
11561 the date before the conversion is operated. This is particularly useful to
11562 emit Date header fields, Expires values in responses when combined with a
11563 positive offset, or Last-Modified values when the offset is negative.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011564
Willy Tarreaud9f316a2014-07-10 14:03:38 +020011565in_table(<table>)
11566 Uses the string representation of the input sample to perform a look up in
11567 the specified table. If the key is not found in the table, a boolean false
11568 is returned. Otherwise a boolean true is returned. This can be used to verify
11569 the presence of a certain key in a table tracking some elements (eg: whether
11570 or not a source IP address or an Authorization header was already seen).
11571
Willy Tarreauffcb2e42014-07-10 16:29:08 +020011572ipmask(<mask>)
11573 Apply a mask to an IPv4 address, and use the result for lookups and storage.
11574 This can be used to make all hosts within a certain mask to share the same
11575 table entries and as such use the same server. The mask can be passed in
11576 dotted form (eg: 255.255.255.0) or in CIDR form (eg: 24).
11577
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020011578json([<input-code>])
11579 Escapes the input string and produces an ASCII ouput string ready to use as a
11580 JSON string. The converter tries to decode the input string according to the
11581 <input-code> parameter. It can be "ascii", "utf8", "utf8s", "utf8"" or
11582 "utf8ps". The "ascii" decoder never fails. The "utf8" decoder detects 3 types
11583 of errors:
11584 - bad UTF-8 sequence (lone continuation byte, bad number of continuation
11585 bytes, ...)
11586 - invalid range (the decoded value is within a UTF-8 prohibited range),
11587 - code overlong (the value is encoded with more bytes than necessary).
11588
11589 The UTF-8 JSON encoding can produce a "too long value" error when the UTF-8
11590 character is greater than 0xffff because the JSON string escape specification
11591 only authorizes 4 hex digits for the value encoding. The UTF-8 decoder exists
11592 in 4 variants designated by a combination of two suffix letters : "p" for
11593 "permissive" and "s" for "silently ignore". The behaviors of the decoders
11594 are :
11595 - "ascii" : never fails ;
11596 - "utf8" : fails on any detected errors ;
11597 - "utf8s" : never fails, but removes characters corresponding to errors ;
11598 - "utf8p" : accepts and fixes the overlong errors, but fails on any other
11599 error ;
11600 - "utf8ps" : never fails, accepts and fixes the overlong errors, but removes
11601 characters corresponding to the other errors.
11602
11603 This converter is particularly useful for building properly escaped JSON for
11604 logging to servers which consume JSON-formated traffic logs.
11605
11606 Example:
11607 capture request header user-agent len 150
11608 capture request header Host len 15
11609 log-format {"ip":"%[src]","user-agent":"%[capture.req.hdr(1),json]"}
11610
11611 Input request from client 127.0.0.1:
11612 GET / HTTP/1.0
11613 User-Agent: Very "Ugly" UA 1/2
11614
11615 Output log:
11616 {"ip":"127.0.0.1","user-agent":"Very \"Ugly\" UA 1\/2"}
11617
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011618language(<value>[,<default>])
11619 Returns the value with the highest q-factor from a list as extracted from the
11620 "accept-language" header using "req.fhdr". Values with no q-factor have a
11621 q-factor of 1. Values with a q-factor of 0 are dropped. Only values which
11622 belong to the list of semi-colon delimited <values> will be considered. The
11623 argument <value> syntax is "lang[;lang[;lang[;...]]]". If no value matches the
11624 given list and a default value is provided, it is returned. Note that language
11625 names may have a variant after a dash ('-'). If this variant is present in the
11626 list, it will be matched, but if it is not, only the base language is checked.
11627 The match is case-sensitive, and the output string is always one of those
11628 provided in arguments. The ordering of arguments is meaningless, only the
11629 ordering of the values in the request counts, as the first value among
11630 multiple sharing the same q-factor is used.
Thierry FOURNIERad903512014-04-11 17:51:01 +020011631
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011632 Example :
Thierry FOURNIERad903512014-04-11 17:51:01 +020011633
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011634 # this configuration switches to the backend matching a
11635 # given language based on the request :
Thierry FOURNIERad903512014-04-11 17:51:01 +020011636
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011637 acl es req.fhdr(accept-language),language(es;fr;en) -m str es
11638 acl fr req.fhdr(accept-language),language(es;fr;en) -m str fr
11639 acl en req.fhdr(accept-language),language(es;fr;en) -m str en
11640 use_backend spanish if es
11641 use_backend french if fr
11642 use_backend english if en
11643 default_backend choose_your_language
Thierry FOURNIERad903512014-04-11 17:51:01 +020011644
Willy Tarreauffcb2e42014-07-10 16:29:08 +020011645lower
11646 Convert a string sample to lower case. This can only be placed after a string
11647 sample fetch function or after a transformation keyword returning a string
11648 type. The result is of type string.
11649
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020011650ltime(<format>[,<offset>])
11651 Converts an integer supposed to contain a date since epoch to a string
11652 representing this date in local time using a format defined by the <format>
11653 string using strftime(3). The purpose is to allow any date format to be used
11654 in logs. An optional <offset> in seconds may be applied to the input date
11655 (positive or negative). See the strftime() man page for the format supported
11656 by your operating system. See also the utime converter.
11657
11658 Example :
11659
11660 # Emit two colons, one with the local time and another with ip:port
11661 # Eg: 20140710162350 127.0.0.1:57325
11662 log-format %[date,ltime(%Y%m%d%H%M%S)]\ %ci:%cp
11663
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011664map(<map_file>[,<default_value>])
11665map_<match_type>(<map_file>[,<default_value>])
11666map_<match_type>_<output_type>(<map_file>[,<default_value>])
11667 Search the input value from <map_file> using the <match_type> matching method,
11668 and return the associated value converted to the type <output_type>. If the
11669 input value cannot be found in the <map_file>, the converter returns the
11670 <default_value>. If the <default_value> is not set, the converter fails and
11671 acts as if no input value could be fetched. If the <match_type> is not set, it
11672 defaults to "str". Likewise, if the <output_type> is not set, it defaults to
11673 "str". For convenience, the "map" keyword is an alias for "map_str" and maps a
11674 string to another string.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010011675
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011676 It is important to avoid overlapping between the keys : IP addresses and
11677 strings are stored in trees, so the first of the finest match will be used.
11678 Other keys are stored in lists, so the first matching occurrence will be used.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010011679
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011680 The following array contains the list of all map functions avalaible sorted by
11681 input type, match type and output type.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010011682
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011683 input type | match method | output type str | output type int | output type ip
11684 -----------+--------------+-----------------+-----------------+---------------
11685 str | str | map_str | map_str_int | map_str_ip
11686 -----------+--------------+-----------------+-----------------+---------------
Willy Tarreau787a4c02014-05-10 07:55:30 +020011687 str | beg | map_beg | map_beg_int | map_end_ip
11688 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011689 str | sub | map_sub | map_sub_int | map_sub_ip
11690 -----------+--------------+-----------------+-----------------+---------------
11691 str | dir | map_dir | map_dir_int | map_dir_ip
11692 -----------+--------------+-----------------+-----------------+---------------
11693 str | dom | map_dom | map_dom_int | map_dom_ip
11694 -----------+--------------+-----------------+-----------------+---------------
11695 str | end | map_end | map_end_int | map_end_ip
11696 -----------+--------------+-----------------+-----------------+---------------
11697 str | reg | map_reg | map_reg_int | map_reg_ip
11698 -----------+--------------+-----------------+-----------------+---------------
11699 int | int | map_int | map_int_int | map_int_ip
11700 -----------+--------------+-----------------+-----------------+---------------
11701 ip | ip | map_ip | map_ip_int | map_ip_ip
11702 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010011703
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011704 The file contains one key + value per line. Lines which start with '#' are
11705 ignored, just like empty lines. Leading tabs and spaces are stripped. The key
11706 is then the first "word" (series of non-space/tabs characters), and the value
11707 is what follows this series of space/tab till the end of the line excluding
11708 trailing spaces/tabs.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010011709
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011710 Example :
11711
11712 # this is a comment and is ignored
11713 2.22.246.0/23 United Kingdom \n
11714 <-><-----------><--><------------><---->
11715 | | | | `- trailing spaces ignored
11716 | | | `---------- value
11717 | | `-------------------- middle spaces ignored
11718 | `---------------------------- key
11719 `------------------------------------ leading spaces ignored
11720
Willy Tarreau97707872015-01-27 15:12:13 +010011721mod(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011722 Divides the input value of type signed integer by <value>, and returns the
11723 remainder as an signed integer. If <value> is null, then zero is returned.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011724 <value> can be a numeric value or a variable name. The name of the variable
11725 starts by an indication about its scope. The allowed scopes are:
11726 "sess" : the variable is shared with all the session,
11727 "txn" : the variable is shared with all the transaction (request and
11728 response),
11729 "req" : the variable is shared only during the request processing,
11730 "res" : the variable is shared only during the response processing.
11731 This prefix is followed by a name. The separator is a '.'. The name may only
11732 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011733
11734mul(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011735 Multiplies the input value of type signed integer by <value>, and returns
Thierry FOURNIER00c005c2015-07-08 01:10:21 +020011736 the product as an signed integer. In case of overflow, the largest possible
11737 value for the sign is returned so that the operation doesn't wrap around.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011738 <value> can be a numeric value or a variable name. The name of the variable
11739 starts by an indication about its scope. The allowed scopes are:
11740 "sess" : the variable is shared with all the session,
11741 "txn" : the variable is shared with all the transaction (request and
11742 response),
11743 "req" : the variable is shared only during the request processing,
11744 "res" : the variable is shared only during the response processing.
11745 This prefix is followed by a name. The separator is a '.'. The name may only
11746 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011747
11748neg
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011749 Takes the input value of type signed integer, computes the opposite value,
11750 and returns the remainder as an signed integer. 0 is identity. This operator
11751 is provided for reversed subtracts : in order to subtract the input from a
11752 constant, simply perform a "neg,add(value)".
Willy Tarreau97707872015-01-27 15:12:13 +010011753
11754not
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011755 Returns a boolean FALSE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010011756 non-null, otherwise returns TRUE. Used in conjunction with and(), it can be
11757 used to report true/false for bit testing on input values (eg: verify the
11758 absence of a flag).
11759
11760odd
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011761 Returns a boolean TRUE if the input value of type signed integer is odd
Willy Tarreau97707872015-01-27 15:12:13 +010011762 otherwise returns FALSE. It is functionally equivalent to "and(1),bool".
11763
11764or(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011765 Performs a bitwise "OR" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011766 integer, and returns the result as an signed integer. <value> can be a
11767 numeric value or a variable name. The name of the variable starts by an
11768 indication about its scope. The allowed scopes are:
11769 "sess" : the variable is shared with all the session,
11770 "txn" : the variable is shared with all the transaction (request and
11771 response),
11772 "req" : the variable is shared only during the request processing,
11773 "res" : the variable is shared only during the response processing.
11774 This prefix is followed by a name. The separator is a '.'. The name may only
11775 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011776
Willy Tarreauc4dc3502015-01-23 20:39:28 +010011777regsub(<regex>,<subst>[,<flags>])
Willy Tarreau7eda8492015-01-20 19:47:06 +010011778 Applies a regex-based substitution to the input string. It does the same
11779 operation as the well-known "sed" utility with "s/<regex>/<subst>/". By
11780 default it will replace in the input string the first occurrence of the
11781 largest part matching the regular expression <regex> with the substitution
11782 string <subst>. It is possible to replace all occurrences instead by adding
11783 the flag "g" in the third argument <flags>. It is also possible to make the
11784 regex case insensitive by adding the flag "i" in <flags>. Since <flags> is a
11785 string, it is made up from the concatenation of all desired flags. Thus if
11786 both "i" and "g" are desired, using "gi" or "ig" will have the same effect.
11787 It is important to note that due to the current limitations of the
11788 configuration parser, some characters such as closing parenthesis or comma
11789 are not possible to use in the arguments. The first use of this converter is
11790 to replace certain characters or sequence of characters with other ones.
11791
11792 Example :
11793
11794 # de-duplicate "/" in header "x-path".
11795 # input: x-path: /////a///b/c/xzxyz/
11796 # output: x-path: /a/b/c/xzxyz/
11797 http-request set-header x-path %[hdr(x-path),regsub(/+,/,g)]
11798
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020011799capture-req(<id>)
11800 Capture the string entry in the request slot <id> and returns the entry as
11801 is. If the slot doesn't exist, the capture fails silently.
11802
11803 See also: "declare capture", "http-request capture",
11804 "http-response capture", "req.hdr.capture" and
11805 "res.hdr.capture" (sample fetches).
11806
11807capture-res(<id>)
11808 Capture the string entry in the response slot <id> and returns the entry as
11809 is. If the slot doesn't exist, the capture fails silently.
11810
11811 See also: "declare capture", "http-request capture",
11812 "http-response capture", "req.hdr.capture" and
11813 "res.hdr.capture" (sample fetches).
11814
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020011815sdbm([<avalanche>])
11816 Hashes a binary input sample into an unsigned 32-bit quantity using the SDBM
11817 hash function. Optionally, it is possible to apply a full avalanche hash
11818 function to the output if the optional <avalanche> argument equals 1. This
11819 converter uses the same functions as used by the various hash-based load
11820 balancing algorithms, so it will provide exactly the same results. It is
11821 mostly intended for debugging, but can be used as a stick-table entry to
11822 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010011823 32-bit hash is trivial to break. See also "crc32", "djb2", "wt6" and the
11824 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020011825
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020011826set-var(<var name>)
11827 Sets a variable with the input content and return the content on the output as
11828 is. The variable keep the value and the associated input type. The name of the
11829 variable starts by an indication about it scope. The scope allowed are:
11830 "sess" : the variable is shared with all the session,
11831 "txn" : the variable is shared with all the transaction (request and
11832 response),
11833 "req" : the variable is shared only during the request processing,
11834 "res" : the variable is shared only during the response processing.
11835 This prefix is followed by a name. The separator is a '.'. The name may only
11836 contain characters 'a-z', 'A-Z', '0-9' and '_'.
11837
Willy Tarreau97707872015-01-27 15:12:13 +010011838sub(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011839 Subtracts <value> from the input value of type signed integer, and returns
11840 the result as an signed integer. Note: in order to subtract the input from
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011841 a constant, simply perform a "neg,add(value)". <value> can be a numeric value
11842 or a variable name. The name of the variable starts by an indication about its
11843 scope. The allowed scopes are:
11844 "sess" : the variable is shared with all the session,
11845 "txn" : the variable is shared with all the transaction (request and
11846 response),
11847 "req" : the variable is shared only during the request processing,
11848 "res" : the variable is shared only during the response processing.
11849 This prefix is followed by a name. The separator is a '.'. The name may only
11850 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011851
Willy Tarreaud9f316a2014-07-10 14:03:38 +020011852table_bytes_in_rate(<table>)
11853 Uses the string representation of the input sample to perform a look up in
11854 the specified table. If the key is not found in the table, integer value zero
11855 is returned. Otherwise the converter returns the average client-to-server
11856 bytes rate associated with the input sample in the designated table, measured
11857 in amount of bytes over the period configured in the table. See also the
11858 sc_bytes_in_rate sample fetch keyword.
11859
11860
11861table_bytes_out_rate(<table>)
11862 Uses the string representation of the input sample to perform a look up in
11863 the specified table. If the key is not found in the table, integer value zero
11864 is returned. Otherwise the converter returns the average server-to-client
11865 bytes rate associated with the input sample in the designated table, measured
11866 in amount of bytes over the period configured in the table. See also the
11867 sc_bytes_out_rate sample fetch keyword.
11868
11869table_conn_cnt(<table>)
11870 Uses the string representation of the input sample to perform a look up in
11871 the specified table. If the key is not found in the table, integer value zero
11872 is returned. Otherwise the converter returns the cumulated amount of incoming
11873 connections associated with the input sample in the designated table. See
11874 also the sc_conn_cnt sample fetch keyword.
11875
11876table_conn_cur(<table>)
11877 Uses the string representation of the input sample to perform a look up in
11878 the specified table. If the key is not found in the table, integer value zero
11879 is returned. Otherwise the converter returns the current amount of concurrent
11880 tracked connections associated with the input sample in the designated table.
11881 See also the sc_conn_cur sample fetch keyword.
11882
11883table_conn_rate(<table>)
11884 Uses the string representation of the input sample to perform a look up in
11885 the specified table. If the key is not found in the table, integer value zero
11886 is returned. Otherwise the converter returns the average incoming connection
11887 rate associated with the input sample in the designated table. See also the
11888 sc_conn_rate sample fetch keyword.
11889
Thierry FOURNIER236657b2015-08-19 08:25:14 +020011890table_gpt0(<table>)
11891 Uses the string representation of the input sample to perform a look up in
11892 the specified table. If the key is not found in the table, boolean value zero
11893 is returned. Otherwise the converter returns the current value of the first
11894 general purpose tag associated with the input sample in the designated table.
11895 See also the sc_get_gpt0 sample fetch keyword.
11896
Willy Tarreaud9f316a2014-07-10 14:03:38 +020011897table_gpc0(<table>)
11898 Uses the string representation of the input sample to perform a look up in
11899 the specified table. If the key is not found in the table, integer value zero
11900 is returned. Otherwise the converter returns the current value of the first
11901 general purpose counter associated with the input sample in the designated
11902 table. See also the sc_get_gpc0 sample fetch keyword.
11903
11904table_gpc0_rate(<table>)
11905 Uses the string representation of the input sample to perform a look up in
11906 the specified table. If the key is not found in the table, integer value zero
11907 is returned. Otherwise the converter returns the frequency which the gpc0
11908 counter was incremented over the configured period in the table, associated
11909 with the input sample in the designated table. See also the sc_get_gpc0_rate
11910 sample fetch keyword.
11911
11912table_http_err_cnt(<table>)
11913 Uses the string representation of the input sample to perform a look up in
11914 the specified table. If the key is not found in the table, integer value zero
11915 is returned. Otherwise the converter returns the cumulated amount of HTTP
11916 errors associated with the input sample in the designated table. See also the
11917 sc_http_err_cnt sample fetch keyword.
11918
11919table_http_err_rate(<table>)
11920 Uses the string representation of the input sample to perform a look up in
11921 the specified table. If the key is not found in the table, integer value zero
11922 is returned. Otherwise the average rate of HTTP errors associated with the
11923 input sample in the designated table, measured in amount of errors over the
11924 period configured in the table. See also the sc_http_err_rate sample fetch
11925 keyword.
11926
11927table_http_req_cnt(<table>)
11928 Uses the string representation of the input sample to perform a look up in
11929 the specified table. If the key is not found in the table, integer value zero
11930 is returned. Otherwise the converter returns the cumulated amount of HTTP
11931 requests associated with the input sample in the designated table. See also
11932 the sc_http_req_cnt sample fetch keyword.
11933
11934table_http_req_rate(<table>)
11935 Uses the string representation of the input sample to perform a look up in
11936 the specified table. If the key is not found in the table, integer value zero
11937 is returned. Otherwise the average rate of HTTP requests associated with the
11938 input sample in the designated table, measured in amount of requests over the
11939 period configured in the table. See also the sc_http_req_rate sample fetch
11940 keyword.
11941
11942table_kbytes_in(<table>)
11943 Uses the string representation of the input sample to perform a look up in
11944 the specified table. If the key is not found in the table, integer value zero
11945 is returned. Otherwise the converter returns the cumulated amount of client-
11946 to-server data associated with the input sample in the designated table,
11947 measured in kilobytes. The test is currently performed on 32-bit integers,
11948 which limits values to 4 terabytes. See also the sc_kbytes_in sample fetch
11949 keyword.
11950
11951table_kbytes_out(<table>)
11952 Uses the string representation of the input sample to perform a look up in
11953 the specified table. If the key is not found in the table, integer value zero
11954 is returned. Otherwise the converter returns the cumulated amount of server-
11955 to-client data associated with the input sample in the designated table,
11956 measured in kilobytes. The test is currently performed on 32-bit integers,
11957 which limits values to 4 terabytes. See also the sc_kbytes_out sample fetch
11958 keyword.
11959
11960table_server_id(<table>)
11961 Uses the string representation of the input sample to perform a look up in
11962 the specified table. If the key is not found in the table, integer value zero
11963 is returned. Otherwise the converter returns the server ID associated with
11964 the input sample in the designated table. A server ID is associated to a
11965 sample by a "stick" rule when a connection to a server succeeds. A server ID
11966 zero means that no server is associated with this key.
11967
11968table_sess_cnt(<table>)
11969 Uses the string representation of the input sample to perform a look up in
11970 the specified table. If the key is not found in the table, integer value zero
11971 is returned. Otherwise the converter returns the cumulated amount of incoming
11972 sessions associated with the input sample in the designated table. Note that
11973 a session here refers to an incoming connection being accepted by the
11974 "tcp-request connection" rulesets. See also the sc_sess_cnt sample fetch
11975 keyword.
11976
11977table_sess_rate(<table>)
11978 Uses the string representation of the input sample to perform a look up in
11979 the specified table. If the key is not found in the table, integer value zero
11980 is returned. Otherwise the converter returns the average incoming session
11981 rate associated with the input sample in the designated table. Note that a
11982 session here refers to an incoming connection being accepted by the
11983 "tcp-request connection" rulesets. See also the sc_sess_rate sample fetch
11984 keyword.
11985
11986table_trackers(<table>)
11987 Uses the string representation of the input sample to perform a look up in
11988 the specified table. If the key is not found in the table, integer value zero
11989 is returned. Otherwise the converter returns the current amount of concurrent
11990 connections tracking the same key as the input sample in the designated
11991 table. It differs from table_conn_cur in that it does not rely on any stored
11992 information but on the table's reference count (the "use" value which is
11993 returned by "show table" on the CLI). This may sometimes be more suited for
11994 layer7 tracking. It can be used to tell a server how many concurrent
11995 connections there are from a given address for example. See also the
11996 sc_trackers sample fetch keyword.
11997
Willy Tarreauffcb2e42014-07-10 16:29:08 +020011998upper
11999 Convert a string sample to upper case. This can only be placed after a string
12000 sample fetch function or after a transformation keyword returning a string
12001 type. The result is of type string.
12002
Thierry FOURNIER82ff3c92015-05-07 15:46:20 +020012003url_dec
12004 Takes an url-encoded string provided as input and returns the decoded
12005 version as output. The input and the output are of type string.
12006
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020012007utime(<format>[,<offset>])
12008 Converts an integer supposed to contain a date since epoch to a string
12009 representing this date in UTC time using a format defined by the <format>
12010 string using strftime(3). The purpose is to allow any date format to be used
12011 in logs. An optional <offset> in seconds may be applied to the input date
12012 (positive or negative). See the strftime() man page for the format supported
12013 by your operating system. See also the ltime converter.
12014
12015 Example :
12016
12017 # Emit two colons, one with the UTC time and another with ip:port
12018 # Eg: 20140710162350 127.0.0.1:57325
12019 log-format %[date,utime(%Y%m%d%H%M%S)]\ %ci:%cp
12020
Emeric Brunc9a0f6d2014-11-25 14:09:01 +010012021word(<index>,<delimiters>)
12022 Extracts the nth word considering given delimiters from an input string.
12023 Indexes start at 1 and delimiters are a string formatted list of chars.
12024
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012025wt6([<avalanche>])
12026 Hashes a binary input sample into an unsigned 32-bit quantity using the WT6
12027 hash function. Optionally, it is possible to apply a full avalanche hash
12028 function to the output if the optional <avalanche> argument equals 1. This
12029 converter uses the same functions as used by the various hash-based load
12030 balancing algorithms, so it will provide exactly the same results. It is
12031 mostly intended for debugging, but can be used as a stick-table entry to
12032 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010012033 32-bit hash is trivial to break. See also "crc32", "djb2", "sdbm", and the
12034 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012035
Willy Tarreau97707872015-01-27 15:12:13 +010012036xor(<value>)
12037 Performs a bitwise "XOR" (exclusive OR) between <value> and the input value
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012038 of type signed integer, and returns the result as an signed integer.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012039 <value> can be a numeric value or a variable name. The name of the variable
12040 starts by an indication about its scope. The allowed scopes are:
12041 "sess" : the variable is shared with all the session,
12042 "txn" : the variable is shared with all the transaction (request and
12043 response),
12044 "req" : the variable is shared only during the request processing,
12045 "res" : the variable is shared only during the response processing.
12046 This prefix is followed by a name. The separator is a '.'. The name may only
12047 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012048
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012049
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200120507.3.2. Fetching samples from internal states
Willy Tarreau74ca5042013-06-11 23:12:07 +020012051--------------------------------------------
12052
12053A first set of sample fetch methods applies to internal information which does
12054not even relate to any client information. These ones are sometimes used with
12055"monitor-fail" directives to report an internal status to external watchers.
12056The sample fetch methods described in this section are usable anywhere.
12057
12058always_false : boolean
12059 Always returns the boolean "false" value. It may be used with ACLs as a
12060 temporary replacement for another one when adjusting configurations.
12061
12062always_true : boolean
12063 Always returns the boolean "true" value. It may be used with ACLs as a
12064 temporary replacement for another one when adjusting configurations.
12065
12066avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010012067 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +020012068 divided by the number of active servers. The current backend is used if no
12069 backend is specified. This is very similar to "queue" except that the size of
12070 the farm is considered, in order to give a more accurate measurement of the
12071 time it may take for a new connection to be processed. The main usage is with
12072 ACL to return a sorry page to new users when it becomes certain they will get
12073 a degraded service, or to pass to the backend servers in a header so that
12074 they decide to work in degraded mode or to disable some functions to speed up
12075 the processing a bit. Note that in the event there would not be any active
12076 server anymore, twice the number of queued connections would be considered as
12077 the measured value. This is a fair estimate, as we expect one server to get
12078 back soon anyway, but we still prefer to send new traffic to another backend
12079 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
12080 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +010012081
Willy Tarreau74ca5042013-06-11 23:12:07 +020012082be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +020012083 Applies to the number of currently established connections on the backend,
12084 possibly including the connection being evaluated. If no backend name is
12085 specified, the current one is used. But it is also possible to check another
12086 backend. It can be used to use a specific farm when the nominal one is full.
12087 See also the "fe_conn", "queue" and "be_sess_rate" criteria.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012088
Willy Tarreau74ca5042013-06-11 23:12:07 +020012089be_sess_rate([<backend>]) : integer
12090 Returns an integer value corresponding to the sessions creation rate on the
12091 backend, in number of new sessions per second. This is used with ACLs to
12092 switch to an alternate backend when an expensive or fragile one reaches too
12093 high a session rate, or to limit abuse of service (eg. prevent sucking of an
12094 online dictionary). It can also be useful to add this element to logs using a
12095 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012096
12097 Example :
12098 # Redirect to an error page if the dictionary is requested too often
12099 backend dynamic
12100 mode http
12101 acl being_scanned be_sess_rate gt 100
12102 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +010012103
Thierry FOURNIERcc103292015-06-06 19:30:17 +020012104bin(<hexa>) : bin
12105 Returns a binary chain. The input is the hexadecimal representation
12106 of the string.
12107
12108bool(<bool>) : bool
12109 Returns a boolean value. <bool> can be 'true', 'false', '1' or '0'.
12110 'false' and '0' are the same. 'true' and '1' are the same.
12111
Willy Tarreau74ca5042013-06-11 23:12:07 +020012112connslots([<backend>]) : integer
12113 Returns an integer value corresponding to the number of connection slots
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012114 still available in the backend, by totaling the maximum amount of
Willy Tarreau74ca5042013-06-11 23:12:07 +020012115 connections on all servers and the maximum queue size. This is probably only
12116 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -050012117
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012118 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +020012119 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012120 usage; see "use_backend" keyword) can be redirected to a different backend.
12121
Willy Tarreau55165fe2009-05-10 12:02:55 +020012122 'connslots' = number of available server connection slots, + number of
12123 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012124
Willy Tarreaua36af912009-10-10 12:02:45 +020012125 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +020012126 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +020012127 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +020012128 you want to be able to differentiate between different backends, and their
12129 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +020012130 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +020012131 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012132
Willy Tarreau55165fe2009-05-10 12:02:55 +020012133 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
12134 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +020012135 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +020012136 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012137
Willy Tarreau6236d3a2013-07-25 14:28:25 +020012138date([<offset>]) : integer
12139 Returns the current date as the epoch (number of seconds since 01/01/1970).
12140 If an offset value is specified, then it is a number of seconds that is added
12141 to the current date before returning the value. This is particularly useful
12142 to compute relative dates, as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +020012143 It is useful combined with the http_date converter.
12144
12145 Example :
12146
12147 # set an expires header to now+1 hour in every response
12148 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +020012149
Willy Tarreau595ec542013-06-12 21:34:28 +020012150env(<name>) : string
12151 Returns a string containing the value of environment variable <name>. As a
12152 reminder, environment variables are per-process and are sampled when the
12153 process starts. This can be useful to pass some information to a next hop
12154 server, or with ACLs to take specific action when the process is started a
12155 certain way.
12156
12157 Examples :
12158 # Pass the Via header to next hop with the local hostname in it
12159 http-request add-header Via 1.1\ %[env(HOSTNAME)]
12160
12161 # reject cookie-less requests when the STOP environment variable is set
12162 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
12163
Willy Tarreau74ca5042013-06-11 23:12:07 +020012164fe_conn([<frontend>]) : integer
12165 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +010012166 possibly including the connection being evaluated. If no frontend name is
12167 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020012168 frontend. It can be used to return a sorry page before hard-blocking, or to
12169 use a specific backend to drain new requests when the farm is considered
12170 full. This is mostly used with ACLs but can also be used to pass some
12171 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
12172 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +020012173
Willy Tarreau74ca5042013-06-11 23:12:07 +020012174fe_sess_rate([<frontend>]) : integer
12175 Returns an integer value corresponding to the sessions creation rate on the
12176 frontend, in number of new sessions per second. This is used with ACLs to
12177 limit the incoming session rate to an acceptable range in order to prevent
12178 abuse of service at the earliest moment, for example when combined with other
12179 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
12180 down below the limit. It can also be useful to add this element to logs using
12181 a log-format directive. See also the "rate-limit sessions" directive for use
12182 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010012183
12184 Example :
12185 # This frontend limits incoming mails to 10/s with a max of 100
12186 # concurrent connections. We accept any connection below 10/s, and
12187 # force excess clients to wait for 100 ms. Since clients are limited to
12188 # 100 max, there cannot be more than 10 incoming mails per second.
12189 frontend mail
12190 bind :25
12191 mode tcp
12192 maxconn 100
12193 acl too_fast fe_sess_rate ge 10
12194 tcp-request inspect-delay 100ms
12195 tcp-request content accept if ! too_fast
12196 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +010012197
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012198int(<integer>) : signed integer
12199 Returns a signed integer.
12200
Thierry FOURNIERcc103292015-06-06 19:30:17 +020012201ipv4(<ipv4>) : ipv4
12202 Returns an ipv4.
12203
12204ipv6(<ipv6>) : ipv6
12205 Returns an ipv6.
12206
12207meth(<method>) : method
12208 Returns a method.
12209
Willy Tarreau0f30d262014-11-24 16:02:05 +010012210nbproc : integer
12211 Returns an integer value corresponding to the number of processes that were
12212 started (it equals the global "nbproc" setting). This is useful for logging
12213 and debugging purposes.
12214
Willy Tarreau74ca5042013-06-11 23:12:07 +020012215nbsrv([<backend>]) : integer
12216 Returns an integer value corresponding to the number of usable servers of
12217 either the current backend or the named backend. This is mostly used with
12218 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +010012219 switch to an alternate backend when the number of servers is too low to
12220 to handle some load. It is useful to report a failure when combined with
12221 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +010012222
Willy Tarreau0f30d262014-11-24 16:02:05 +010012223proc : integer
12224 Returns an integer value corresponding to the position of the process calling
12225 the function, between 1 and global.nbproc. This is useful for logging and
12226 debugging purposes.
12227
Willy Tarreau74ca5042013-06-11 23:12:07 +020012228queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010012229 Returns the total number of queued connections of the designated backend,
12230 including all the connections in server queues. If no backend name is
12231 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020012232 one. This is useful with ACLs or to pass statistics to backend servers. This
12233 can be used to take actions when queuing goes above a known level, generally
12234 indicating a surge of traffic or a massive slowdown on the servers. One
12235 possible action could be to reject new users but still accept old ones. See
12236 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
12237
Willy Tarreau84310e22014-02-14 11:59:04 +010012238rand([<range>]) : integer
12239 Returns a random integer value within a range of <range> possible values,
12240 starting at zero. If the range is not specified, it defaults to 2^32, which
12241 gives numbers between 0 and 4294967295. It can be useful to pass some values
12242 needed to take some routing decisions for example, or just for debugging
12243 purposes. This random must not be used for security purposes.
12244
Willy Tarreau74ca5042013-06-11 23:12:07 +020012245srv_conn([<backend>/]<server>) : integer
12246 Returns an integer value corresponding to the number of currently established
12247 connections on the designated server, possibly including the connection being
12248 evaluated. If <backend> is omitted, then the server is looked up in the
12249 current backend. It can be used to use a specific farm when one server is
12250 full, or to inform the server about our view of the number of active
12251 connections with it. See also the "fe_conn", "be_conn" and "queue" fetch
12252 methods.
12253
12254srv_is_up([<backend>/]<server>) : boolean
12255 Returns true when the designated server is UP, and false when it is either
12256 DOWN or in maintenance mode. If <backend> is omitted, then the server is
12257 looked up in the current backend. It is mainly used to take action based on
12258 an external status reported via a health check (eg: a geographical site's
12259 availability). Another possible use which is more of a hack consists in
12260 using dummy servers as boolean variables that can be enabled or disabled from
12261 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
12262
12263srv_sess_rate([<backend>/]<server>) : integer
12264 Returns an integer corresponding to the sessions creation rate on the
12265 designated server, in number of new sessions per second. If <backend> is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012266 omitted, then the server is looked up in the current backend. This is mostly
Willy Tarreau74ca5042013-06-11 23:12:07 +020012267 used with ACLs but can make sense with logs too. This is used to switch to an
12268 alternate backend when an expensive or fragile one reaches too high a session
12269 rate, or to limit abuse of service (eg. prevent latent requests from
12270 overloading servers).
12271
12272 Example :
12273 # Redirect to a separate back
12274 acl srv1_full srv_sess_rate(be1/srv1) gt 50
12275 acl srv2_full srv_sess_rate(be1/srv2) gt 50
12276 use_backend be2 if srv1_full or srv2_full
12277
Willy Tarreau0f30d262014-11-24 16:02:05 +010012278stopping : boolean
12279 Returns TRUE if the process calling the function is currently stopping. This
12280 can be useful for logging, or for relaxing certain checks or helping close
12281 certain connections upon graceful shutdown.
12282
Thierry FOURNIERcc103292015-06-06 19:30:17 +020012283str(<string>) : string
12284 Returns a string.
12285
Willy Tarreau74ca5042013-06-11 23:12:07 +020012286table_avl([<table>]) : integer
12287 Returns the total number of available entries in the current proxy's
12288 stick-table or in the designated stick-table. See also table_cnt.
12289
12290table_cnt([<table>]) : integer
12291 Returns the total number of entries currently in use in the current proxy's
12292 stick-table or in the designated stick-table. See also src_conn_cnt and
12293 table_avl for other entry counting methods.
12294
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012295var(<var-name>) : undefined
12296 Returns a variable with the stored type. If the variable is not set, the
12297 sample fetch fails. The name of the variable starts by an indication about its
12298 scope. The scope allowed are:
12299 "sess" : the variable is shared with all the session,
12300 "txn" : the variable is shared with all the transaction (request and
12301 response),
12302 "req" : the variable is shared only during the request processing,
12303 "res" : the variable is shared only during the response processing.
12304 This prefix is followed by a name. The separator is a '.'. The name may only
12305 contain characters 'a-z', 'A-Z', '0-9' and '_'.
12306
Willy Tarreau74ca5042013-06-11 23:12:07 +020012307
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200123087.3.3. Fetching samples at Layer 4
Willy Tarreau74ca5042013-06-11 23:12:07 +020012309----------------------------------
12310
12311The layer 4 usually describes just the transport layer which in haproxy is
12312closest to the connection, where no content is yet made available. The fetch
12313methods described here are usable as low as the "tcp-request connection" rule
12314sets unless they require some future information. Those generally include
12315TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012316the incoming connection. For retrieving a value from a sticky counters, the
12317counter number can be explicitly set as 0, 1, or 2 using the pre-defined
12318"sc0_", "sc1_", or "sc2_" prefix, or it can be specified as the first integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012319argument when using the "sc_" prefix. An optional table may be specified with
12320the "sc*" form, in which case the currently tracked key will be looked up into
12321this alternate table instead of the table currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012322
12323be_id : integer
12324 Returns an integer containing the current backend's id. It can be used in
12325 frontends with responses to check which backend processed the request.
12326
12327dst : ip
12328 This is the destination IPv4 address of the connection on the client side,
12329 which is the address the client connected to. It can be useful when running
12330 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
12331 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
12332 RFC 4291.
12333
12334dst_conn : integer
12335 Returns an integer value corresponding to the number of currently established
12336 connections on the same socket including the one being evaluated. It is
12337 normally used with ACLs but can as well be used to pass the information to
12338 servers in an HTTP header or in logs. It can be used to either return a sorry
12339 page before hard-blocking, or to use a specific backend to drain new requests
12340 when the socket is considered saturated. This offers the ability to assign
12341 different limits to different listening ports or addresses. See also the
12342 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012343
Willy Tarreau74ca5042013-06-11 23:12:07 +020012344dst_port : integer
12345 Returns an integer value corresponding to the destination TCP port of the
12346 connection on the client side, which is the port the client connected to.
12347 This might be used when running in transparent mode, when assigning dynamic
12348 ports to some clients for a whole application session, to stick all users to
12349 a same server, or to pass the destination port information to a server using
12350 an HTTP header.
12351
12352fe_id : integer
12353 Returns an integer containing the current frontend's id. It can be used in
12354 backends to check from which backend it was called, or to stick all users
12355 coming via a same frontend to the same server.
12356
Cyril Bonté62ba8702014-04-22 23:52:25 +020012357sc_bytes_in_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012358sc0_bytes_in_rate([<table>]) : integer
12359sc1_bytes_in_rate([<table>]) : integer
12360sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012361 Returns the average client-to-server bytes rate from the currently tracked
12362 counters, measured in amount of bytes over the period configured in the
12363 table. See also src_bytes_in_rate.
12364
Cyril Bonté62ba8702014-04-22 23:52:25 +020012365sc_bytes_out_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012366sc0_bytes_out_rate([<table>]) : integer
12367sc1_bytes_out_rate([<table>]) : integer
12368sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012369 Returns the average server-to-client bytes rate from the currently tracked
12370 counters, measured in amount of bytes over the period configured in the
12371 table. See also src_bytes_out_rate.
12372
Cyril Bonté62ba8702014-04-22 23:52:25 +020012373sc_clr_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012374sc0_clr_gpc0([<table>]) : integer
12375sc1_clr_gpc0([<table>]) : integer
12376sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +020012377 Clears the first General Purpose Counter associated to the currently tracked
12378 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +010012379 stored value is zero, so first invocation will always return zero. This is
12380 typically used as a second ACL in an expression in order to mark a connection
12381 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020012382
12383 # block if 5 consecutive requests continue to come faster than 10 sess
12384 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012385 acl abuse sc0_http_req_rate gt 10
12386 acl kill sc0_inc_gpc0 gt 5
12387 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020012388 tcp-request connection accept if !abuse save
12389 tcp-request connection reject if abuse kill
12390
Cyril Bonté62ba8702014-04-22 23:52:25 +020012391sc_conn_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012392sc0_conn_cnt([<table>]) : integer
12393sc1_conn_cnt([<table>]) : integer
12394sc2_conn_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012395 Returns the cumulated number of incoming connections from currently tracked
12396 counters. See also src_conn_cnt.
12397
Cyril Bonté62ba8702014-04-22 23:52:25 +020012398sc_conn_cur(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012399sc0_conn_cur([<table>]) : integer
12400sc1_conn_cur([<table>]) : integer
12401sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012402 Returns the current amount of concurrent connections tracking the same
12403 tracked counters. This number is automatically incremented when tracking
12404 begins and decremented when tracking stops. See also src_conn_cur.
12405
Cyril Bonté62ba8702014-04-22 23:52:25 +020012406sc_conn_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012407sc0_conn_rate([<table>]) : integer
12408sc1_conn_rate([<table>]) : integer
12409sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012410 Returns the average connection rate from the currently tracked counters,
12411 measured in amount of connections over the period configured in the table.
12412 See also src_conn_rate.
12413
Cyril Bonté62ba8702014-04-22 23:52:25 +020012414sc_get_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012415sc0_get_gpc0([<table>]) : integer
12416sc1_get_gpc0([<table>]) : integer
12417sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012418 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012419 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020012420
Thierry FOURNIER236657b2015-08-19 08:25:14 +020012421sc_get_gpt0(<ctr>[,<table>]) : integer
12422sc0_get_gpt0([<table>]) : integer
12423sc1_get_gpt0([<table>]) : integer
12424sc2_get_gpt0([<table>]) : integer
12425 Returns the value of the first General Purpose Tag associated to the
12426 currently tracked counters. See also src_get_gpt0.
12427
Cyril Bonté62ba8702014-04-22 23:52:25 +020012428sc_gpc0_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012429sc0_gpc0_rate([<table>]) : integer
12430sc1_gpc0_rate([<table>]) : integer
12431sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020012432 Returns the average increment rate of the first General Purpose Counter
12433 associated to the currently tracked counters. It reports the frequency
12434 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012435 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
12436 that the "gpc0_rate" counter must be stored in the stick-table for a value to
12437 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +020012438
Cyril Bonté62ba8702014-04-22 23:52:25 +020012439sc_http_err_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012440sc0_http_err_cnt([<table>]) : integer
12441sc1_http_err_cnt([<table>]) : integer
12442sc2_http_err_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012443 Returns the cumulated number of HTTP errors from the currently tracked
12444 counters. This includes the both request errors and 4xx error responses.
12445 See also src_http_err_cnt.
12446
Cyril Bonté62ba8702014-04-22 23:52:25 +020012447sc_http_err_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012448sc0_http_err_rate([<table>]) : integer
12449sc1_http_err_rate([<table>]) : integer
12450sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012451 Returns the average rate of HTTP errors from the currently tracked counters,
12452 measured in amount of errors over the period configured in the table. This
12453 includes the both request errors and 4xx error responses. See also
12454 src_http_err_rate.
12455
Cyril Bonté62ba8702014-04-22 23:52:25 +020012456sc_http_req_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012457sc0_http_req_cnt([<table>]) : integer
12458sc1_http_req_cnt([<table>]) : integer
12459sc2_http_req_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012460 Returns the cumulated number of HTTP requests from the currently tracked
12461 counters. This includes every started request, valid or not. See also
12462 src_http_req_cnt.
12463
Cyril Bonté62ba8702014-04-22 23:52:25 +020012464sc_http_req_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012465sc0_http_req_rate([<table>]) : integer
12466sc1_http_req_rate([<table>]) : integer
12467sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012468 Returns the average rate of HTTP requests from the currently tracked
12469 counters, measured in amount of requests over the period configured in
12470 the table. This includes every started request, valid or not. See also
12471 src_http_req_rate.
12472
Cyril Bonté62ba8702014-04-22 23:52:25 +020012473sc_inc_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012474sc0_inc_gpc0([<table>]) : integer
12475sc1_inc_gpc0([<table>]) : integer
12476sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012477 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +010012478 tracked counters, and returns its new value. Before the first invocation,
12479 the stored value is zero, so first invocation will increase it to 1 and will
12480 return 1. This is typically used as a second ACL in an expression in order
12481 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +020012482
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012483 acl abuse sc0_http_req_rate gt 10
12484 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020012485 tcp-request connection reject if abuse kill
12486
Cyril Bonté62ba8702014-04-22 23:52:25 +020012487sc_kbytes_in(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012488sc0_kbytes_in([<table>]) : integer
12489sc1_kbytes_in([<table>]) : integer
12490sc2_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020012491 Returns the total amount of client-to-server data from the currently tracked
12492 counters, measured in kilobytes. The test is currently performed on 32-bit
12493 integers, which limits values to 4 terabytes. See also src_kbytes_in.
Willy Tarreaue9656522010-08-17 15:40:09 +020012494
Cyril Bonté62ba8702014-04-22 23:52:25 +020012495sc_kbytes_out(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012496sc0_kbytes_out([<table>]) : integer
12497sc1_kbytes_out([<table>]) : integer
12498sc2_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020012499 Returns the total amount of server-to-client data from the currently tracked
12500 counters, measured in kilobytes. The test is currently performed on 32-bit
12501 integers, which limits values to 4 terabytes. See also src_kbytes_out.
Willy Tarreaue9656522010-08-17 15:40:09 +020012502
Cyril Bonté62ba8702014-04-22 23:52:25 +020012503sc_sess_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012504sc0_sess_cnt([<table>]) : integer
12505sc1_sess_cnt([<table>]) : integer
12506sc2_sess_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012507 Returns the cumulated number of incoming connections that were transformed
12508 into sessions, which means that they were accepted by a "tcp-request
12509 connection" rule, from the currently tracked counters. A backend may count
12510 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -040012511 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +020012512 with the client. See also src_sess_cnt.
12513
Cyril Bonté62ba8702014-04-22 23:52:25 +020012514sc_sess_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012515sc0_sess_rate([<table>]) : integer
12516sc1_sess_rate([<table>]) : integer
12517sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012518 Returns the average session rate from the currently tracked counters,
12519 measured in amount of sessions over the period configured in the table. A
12520 session is a connection that got past the early "tcp-request connection"
12521 rules. A backend may count more sessions than connections because each
12522 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -040012523 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +020012524
Cyril Bonté62ba8702014-04-22 23:52:25 +020012525sc_tracked(<ctr>[,<table>]) : boolean
Willy Tarreau0f791d42013-07-23 19:56:43 +020012526sc0_tracked([<table>]) : boolean
12527sc1_tracked([<table>]) : boolean
12528sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +020012529 Returns true if the designated session counter is currently being tracked by
12530 the current session. This can be useful when deciding whether or not we want
12531 to set some values in a header passed to the server.
12532
Cyril Bonté62ba8702014-04-22 23:52:25 +020012533sc_trackers(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012534sc0_trackers([<table>]) : integer
12535sc1_trackers([<table>]) : integer
12536sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +010012537 Returns the current amount of concurrent connections tracking the same
12538 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012539 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +010012540 that it does not rely on any stored information but on the table's reference
12541 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020012542 may sometimes be more suited for layer7 tracking. It can be used to tell a
12543 server how many concurrent connections there are from a given address for
12544 example.
Willy Tarreau2406db42012-12-09 12:16:43 +010012545
Willy Tarreau74ca5042013-06-11 23:12:07 +020012546so_id : integer
12547 Returns an integer containing the current listening socket's id. It is useful
12548 in frontends involving many "bind" lines, or to stick all users coming via a
12549 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012550
Willy Tarreau74ca5042013-06-11 23:12:07 +020012551src : ip
12552 This is the source IPv4 address of the client of the session. It is of type
12553 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
12554 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
12555 TCP-level source address which is used, and not the address of a client
12556 behind a proxy. However if the "accept-proxy" bind directive is used, it can
12557 be the address of a client behind another PROXY-protocol compatible component
12558 for all rule sets except "tcp-request connection" which sees the real address.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012559
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012560 Example:
12561 # add an HTTP header in requests with the originating address' country
12562 http-request set-header X-Country %[src,map_ip(geoip.lst)]
12563
Willy Tarreau74ca5042013-06-11 23:12:07 +020012564src_bytes_in_rate([<table>]) : integer
12565 Returns the average bytes rate from the incoming connection's source address
12566 in the current proxy's stick-table or in the designated stick-table, measured
12567 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012568 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012569
Willy Tarreau74ca5042013-06-11 23:12:07 +020012570src_bytes_out_rate([<table>]) : integer
12571 Returns the average bytes rate to the incoming connection's source address in
12572 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +020012573 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012574 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012575
Willy Tarreau74ca5042013-06-11 23:12:07 +020012576src_clr_gpc0([<table>]) : integer
12577 Clears the first General Purpose Counter associated to the incoming
12578 connection's source address in the current proxy's stick-table or in the
12579 designated stick-table, and returns its previous value. If the address is not
12580 found, an entry is created and 0 is returned. This is typically used as a
12581 second ACL in an expression in order to mark a connection when a first ACL
12582 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020012583
12584 # block if 5 consecutive requests continue to come faster than 10 sess
12585 # per second, and reset the counter as soon as the traffic slows down.
12586 acl abuse src_http_req_rate gt 10
12587 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +010012588 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020012589 tcp-request connection accept if !abuse save
12590 tcp-request connection reject if abuse kill
12591
Willy Tarreau74ca5042013-06-11 23:12:07 +020012592src_conn_cnt([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020012593 Returns the cumulated number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +020012594 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020012595 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012596 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012597
Willy Tarreau74ca5042013-06-11 23:12:07 +020012598src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020012599 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +020012600 current incoming connection's source address in the current proxy's
12601 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012602 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012603
Willy Tarreau74ca5042013-06-11 23:12:07 +020012604src_conn_rate([<table>]) : integer
12605 Returns the average connection rate from the incoming connection's source
12606 address in the current proxy's stick-table or in the designated stick-table,
12607 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012608 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012609
Willy Tarreau74ca5042013-06-11 23:12:07 +020012610src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020012611 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +020012612 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020012613 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012614 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012615
Thierry FOURNIER236657b2015-08-19 08:25:14 +020012616src_get_gpt0([<table>]) : integer
12617 Returns the value of the first General Purpose Tag associated to the
12618 incoming connection's source address in the current proxy's stick-table or in
12619 the designated stick-table. If the address is not found, zero is returned.
12620 See also sc/sc0/sc1/sc2_get_gpt0.
12621
Willy Tarreau74ca5042013-06-11 23:12:07 +020012622src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020012623 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +020012624 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +020012625 stick-table or in the designated stick-table. It reports the frequency
12626 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012627 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
12628 that the "gpc0_rate" counter must be stored in the stick-table for a value to
12629 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020012630
Willy Tarreau74ca5042013-06-11 23:12:07 +020012631src_http_err_cnt([<table>]) : integer
12632 Returns the cumulated number of HTTP errors from the incoming connection's
12633 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +020012634 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012635 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +020012636 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012637
Willy Tarreau74ca5042013-06-11 23:12:07 +020012638src_http_err_rate([<table>]) : integer
12639 Returns the average rate of HTTP errors from the incoming connection's source
12640 address in the current proxy's stick-table or in the designated stick-table,
12641 measured in amount of errors over the period configured in the table. This
12642 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012643 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012644
Willy Tarreau74ca5042013-06-11 23:12:07 +020012645src_http_req_cnt([<table>]) : integer
12646 Returns the cumulated number of HTTP requests from the incoming connection's
12647 source address in the current proxy's stick-table or in the designated stick-
12648 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012649 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012650
Willy Tarreau74ca5042013-06-11 23:12:07 +020012651src_http_req_rate([<table>]) : integer
12652 Returns the average rate of HTTP requests from the incoming connection's
12653 source address in the current proxy's stick-table or in the designated stick-
12654 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +020012655 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012656 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012657
Willy Tarreau74ca5042013-06-11 23:12:07 +020012658src_inc_gpc0([<table>]) : integer
12659 Increments the first General Purpose Counter associated to the incoming
12660 connection's source address in the current proxy's stick-table or in the
12661 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012662 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012663 This is typically used as a second ACL in an expression in order to mark a
12664 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +020012665
12666 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +010012667 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020012668 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +020012669
Willy Tarreau74ca5042013-06-11 23:12:07 +020012670src_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020012671 Returns the total amount of data received from the incoming connection's
12672 source address in the current proxy's stick-table or in the designated
12673 stick-table, measured in kilobytes. If the address is not found, zero is
12674 returned. The test is currently performed on 32-bit integers, which limits
12675 values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012676
Willy Tarreau74ca5042013-06-11 23:12:07 +020012677src_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020012678 Returns the total amount of data sent to the incoming connection's source
12679 address in the current proxy's stick-table or in the designated stick-table,
12680 measured in kilobytes. If the address is not found, zero is returned. The
12681 test is currently performed on 32-bit integers, which limits values to 4
12682 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020012683
Willy Tarreau74ca5042013-06-11 23:12:07 +020012684src_port : integer
12685 Returns an integer value corresponding to the TCP source port of the
12686 connection on the client side, which is the port the client connected from.
12687 Usage of this function is very limited as modern protocols do not care much
12688 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010012689
Willy Tarreau74ca5042013-06-11 23:12:07 +020012690src_sess_cnt([<table>]) : integer
12691 Returns the cumulated number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +020012692 connection's source IPv4 address in the current proxy's stick-table or in the
12693 designated stick-table, that were transformed into sessions, which means that
12694 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012695 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012696
Willy Tarreau74ca5042013-06-11 23:12:07 +020012697src_sess_rate([<table>]) : integer
12698 Returns the average session rate from the incoming connection's source
12699 address in the current proxy's stick-table or in the designated stick-table,
12700 measured in amount of sessions over the period configured in the table. A
12701 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012702 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012703
Willy Tarreau74ca5042013-06-11 23:12:07 +020012704src_updt_conn_cnt([<table>]) : integer
12705 Creates or updates the entry associated to the incoming connection's source
12706 address in the current proxy's stick-table or in the designated stick-table.
12707 This table must be configured to store the "conn_cnt" data type, otherwise
12708 the match will be ignored. The current count is incremented by one, and the
12709 expiration timer refreshed. The updated count is returned, so this match
12710 can't return zero. This was used to reject service abusers based on their
12711 source address. Note: it is recommended to use the more complete "track-sc*"
12712 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020012713
12714 Example :
12715 # This frontend limits incoming SSH connections to 3 per 10 second for
12716 # each source address, and rejects excess connections until a 10 second
12717 # silence is observed. At most 20 addresses are tracked.
12718 listen ssh
12719 bind :22
12720 mode tcp
12721 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +020012722 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +020012723 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +020012724 server local 127.0.0.1:22
12725
Willy Tarreau74ca5042013-06-11 23:12:07 +020012726srv_id : integer
12727 Returns an integer containing the server's id when processing the response.
12728 While it's almost only used with ACLs, it may be used for logging or
12729 debugging.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +020012730
Hervé COMMOWICK35ed8012010-12-15 14:04:51 +010012731
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200127327.3.4. Fetching samples at Layer 5
Willy Tarreau74ca5042013-06-11 23:12:07 +020012733----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +020012734
Willy Tarreau74ca5042013-06-11 23:12:07 +020012735The layer 5 usually describes just the session layer which in haproxy is
12736closest to the session once all the connection handshakes are finished, but
12737when no content is yet made available. The fetch methods described here are
12738usable as low as the "tcp-request content" rule sets unless they require some
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012739future information. Those generally include the results of SSL negotiations.
Willy Tarreauc735a072011-03-29 00:57:02 +020012740
Emeric Brun645ae792014-04-30 14:21:06 +020012741ssl_bc : boolean
12742 Returns true when the back connection was made via an SSL/TLS transport
12743 layer and is locally deciphered. This means the outgoing connection was made
12744 other a server with the "ssl" option.
12745
12746ssl_bc_alg_keysize : integer
12747 Returns the symmetric cipher key size supported in bits when the outgoing
12748 connection was made over an SSL/TLS transport layer.
12749
12750ssl_bc_cipher : string
12751 Returns the name of the used cipher when the outgoing connection was made
12752 over an SSL/TLS transport layer.
12753
12754ssl_bc_protocol : string
12755 Returns the name of the used protocol when the outgoing connection was made
12756 over an SSL/TLS transport layer.
12757
Emeric Brunb73a9b02014-04-30 18:49:19 +020012758ssl_bc_unique_id : binary
Emeric Brun645ae792014-04-30 14:21:06 +020012759 When the outgoing connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020012760 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
12761 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
Emeric Brun645ae792014-04-30 14:21:06 +020012762
12763ssl_bc_session_id : binary
12764 Returns the SSL ID of the back connection when the outgoing connection was
12765 made over an SSL/TLS transport layer. It is useful to log if we want to know
12766 if session was reused or not.
12767
12768ssl_bc_use_keysize : integer
12769 Returns the symmetric cipher key size used in bits when the outgoing
12770 connection was made over an SSL/TLS transport layer.
12771
Willy Tarreau74ca5042013-06-11 23:12:07 +020012772ssl_c_ca_err : integer
12773 When the incoming connection was made over an SSL/TLS transport layer,
12774 returns the ID of the first error detected during verification of the client
12775 certificate at depth > 0, or 0 if no error was encountered during this
12776 verification process. Please refer to your SSL library's documentation to
12777 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +020012778
Willy Tarreau74ca5042013-06-11 23:12:07 +020012779ssl_c_ca_err_depth : integer
12780 When the incoming connection was made over an SSL/TLS transport layer,
12781 returns the depth in the CA chain of the first error detected during the
12782 verification of the client certificate. If no error is encountered, 0 is
12783 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +010012784
Emeric Brun43e79582014-10-29 19:03:26 +010012785ssl_c_der : binary
12786 Returns the DER formatted certificate presented by the client when the
12787 incoming connection was made over an SSL/TLS transport layer. When used for
12788 an ACL, the value(s) to match against can be passed in hexadecimal form.
12789
Willy Tarreau74ca5042013-06-11 23:12:07 +020012790ssl_c_err : integer
12791 When the incoming connection was made over an SSL/TLS transport layer,
12792 returns the ID of the first error detected during verification at depth 0, or
12793 0 if no error was encountered during this verification process. Please refer
12794 to your SSL library's documentation to find the exhaustive list of error
12795 codes.
Willy Tarreau62644772008-07-16 18:36:06 +020012796
Willy Tarreau74ca5042013-06-11 23:12:07 +020012797ssl_c_i_dn([<entry>[,<occ>]]) : string
12798 When the incoming connection was made over an SSL/TLS transport layer,
12799 returns the full distinguished name of the issuer of the certificate
12800 presented by the client when no <entry> is specified, or the value of the
12801 first given entry found from the beginning of the DN. If a positive/negative
12802 occurrence number is specified as the optional second argument, it returns
12803 the value of the nth given entry value from the beginning/end of the DN.
12804 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
12805 "ssl_c_i_dn(CN)" retrieves the common name.
Willy Tarreau62644772008-07-16 18:36:06 +020012806
Willy Tarreau74ca5042013-06-11 23:12:07 +020012807ssl_c_key_alg : string
12808 Returns the name of the algorithm used to generate the key of the certificate
12809 presented by the client when the incoming connection was made over an SSL/TLS
12810 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +020012811
Willy Tarreau74ca5042013-06-11 23:12:07 +020012812ssl_c_notafter : string
12813 Returns the end date presented by the client as a formatted string
12814 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
12815 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +020012816
Willy Tarreau74ca5042013-06-11 23:12:07 +020012817ssl_c_notbefore : string
12818 Returns the start date presented by the client as a formatted string
12819 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
12820 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +010012821
Willy Tarreau74ca5042013-06-11 23:12:07 +020012822ssl_c_s_dn([<entry>[,<occ>]]) : string
12823 When the incoming connection was made over an SSL/TLS transport layer,
12824 returns the full distinguished name of the subject of the certificate
12825 presented by the client when no <entry> is specified, or the value of the
12826 first given entry found from the beginning of the DN. If a positive/negative
12827 occurrence number is specified as the optional second argument, it returns
12828 the value of the nth given entry value from the beginning/end of the DN.
12829 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
12830 "ssl_c_s_dn(CN)" retrieves the common name.
Willy Tarreaub6672b52011-12-12 17:23:41 +010012831
Willy Tarreau74ca5042013-06-11 23:12:07 +020012832ssl_c_serial : binary
12833 Returns the serial of the certificate presented by the client when the
12834 incoming connection was made over an SSL/TLS transport layer. When used for
12835 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +020012836
Willy Tarreau74ca5042013-06-11 23:12:07 +020012837ssl_c_sha1 : binary
12838 Returns the SHA-1 fingerprint of the certificate presented by the client when
12839 the incoming connection was made over an SSL/TLS transport layer. This can be
12840 used to stick a client to a server, or to pass this information to a server.
Willy Tarreau2d0caa32014-07-02 19:01:22 +020012841 Note that the output is binary, so if you want to pass that signature to the
12842 server, you need to encode it in hex or base64, such as in the example below:
12843
12844 http-request set-header X-SSL-Client-SHA1 %[ssl_c_sha1,hex]
Emeric Brun2525b6b2012-10-18 15:59:43 +020012845
Willy Tarreau74ca5042013-06-11 23:12:07 +020012846ssl_c_sig_alg : string
12847 Returns the name of the algorithm used to sign the certificate presented by
12848 the client when the incoming connection was made over an SSL/TLS transport
12849 layer.
Emeric Brun87855892012-10-17 17:39:35 +020012850
Willy Tarreau74ca5042013-06-11 23:12:07 +020012851ssl_c_used : boolean
12852 Returns true if current SSL session uses a client certificate even if current
12853 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +020012854
Willy Tarreau74ca5042013-06-11 23:12:07 +020012855ssl_c_verify : integer
12856 Returns the verify result error ID when the incoming connection was made over
12857 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
12858 refer to your SSL library's documentation for an exhaustive list of error
12859 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +020012860
Willy Tarreau74ca5042013-06-11 23:12:07 +020012861ssl_c_version : integer
12862 Returns the version of the certificate presented by the client when the
12863 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +020012864
Emeric Brun43e79582014-10-29 19:03:26 +010012865ssl_f_der : binary
12866 Returns the DER formatted certificate presented by the frontend when the
12867 incoming connection was made over an SSL/TLS transport layer. When used for
12868 an ACL, the value(s) to match against can be passed in hexadecimal form.
12869
Willy Tarreau74ca5042013-06-11 23:12:07 +020012870ssl_f_i_dn([<entry>[,<occ>]]) : string
12871 When the incoming connection was made over an SSL/TLS transport layer,
12872 returns the full distinguished name of the issuer of the certificate
12873 presented by the frontend when no <entry> is specified, or the value of the
12874 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +020012875 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +020012876 the value of the nth given entry value from the beginning/end of the DN.
12877 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
12878 "ssl_f_i_dn(CN)" retrieves the common name.
Emeric Brun87855892012-10-17 17:39:35 +020012879
Willy Tarreau74ca5042013-06-11 23:12:07 +020012880ssl_f_key_alg : string
12881 Returns the name of the algorithm used to generate the key of the certificate
12882 presented by the frontend when the incoming connection was made over an
12883 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020012884
Willy Tarreau74ca5042013-06-11 23:12:07 +020012885ssl_f_notafter : string
12886 Returns the end date presented by the frontend as a formatted string
12887 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
12888 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +020012889
Willy Tarreau74ca5042013-06-11 23:12:07 +020012890ssl_f_notbefore : string
12891 Returns the start date presented by the frontend as a formatted string
12892 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
12893 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +020012894
Willy Tarreau74ca5042013-06-11 23:12:07 +020012895ssl_f_s_dn([<entry>[,<occ>]]) : string
12896 When the incoming connection was made over an SSL/TLS transport layer,
12897 returns the full distinguished name of the subject of the certificate
12898 presented by the frontend when no <entry> is specified, or the value of the
12899 first given entry found from the beginning of the DN. If a positive/negative
12900 occurrence number is specified as the optional second argument, it returns
12901 the value of the nth given entry value from the beginning/end of the DN.
12902 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
12903 "ssl_f_s_dn(CN)" retrieves the common name.
Emeric Brunce5ad802012-10-22 14:11:22 +020012904
Willy Tarreau74ca5042013-06-11 23:12:07 +020012905ssl_f_serial : binary
12906 Returns the serial of the certificate presented by the frontend when the
12907 incoming connection was made over an SSL/TLS transport layer. When used for
12908 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +020012909
Emeric Brun55f4fa82014-04-30 17:11:25 +020012910ssl_f_sha1 : binary
12911 Returns the SHA-1 fingerprint of the certificate presented by the frontend
12912 when the incoming connection was made over an SSL/TLS transport layer. This
12913 can be used to know which certificate was chosen using SNI.
12914
Willy Tarreau74ca5042013-06-11 23:12:07 +020012915ssl_f_sig_alg : string
12916 Returns the name of the algorithm used to sign the certificate presented by
12917 the frontend when the incoming connection was made over an SSL/TLS transport
12918 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020012919
Willy Tarreau74ca5042013-06-11 23:12:07 +020012920ssl_f_version : integer
12921 Returns the version of the certificate presented by the frontend when the
12922 incoming connection was made over an SSL/TLS transport layer.
12923
12924ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020012925 Returns true when the front connection was made via an SSL/TLS transport
12926 layer and is locally deciphered. This means it has matched a socket declared
12927 with a "bind" line having the "ssl" option.
12928
Willy Tarreau74ca5042013-06-11 23:12:07 +020012929 Example :
12930 # This passes "X-Proto: https" to servers when client connects over SSL
12931 listen http-https
12932 bind :80
12933 bind :443 ssl crt /etc/haproxy.pem
12934 http-request add-header X-Proto https if { ssl_fc }
12935
12936ssl_fc_alg_keysize : integer
12937 Returns the symmetric cipher key size supported in bits when the incoming
12938 connection was made over an SSL/TLS transport layer.
12939
12940ssl_fc_alpn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012941 This extracts the Application Layer Protocol Negotiation field from an
Willy Tarreau74ca5042013-06-11 23:12:07 +020012942 incoming connection made via a TLS transport layer and locally deciphered by
12943 haproxy. The result is a string containing the protocol name advertised by
12944 the client. The SSL library must have been built with support for TLS
12945 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
12946 not advertised unless the "alpn" keyword on the "bind" line specifies a
12947 protocol list. Also, nothing forces the client to pick a protocol from this
12948 list, any other one may be requested. The TLS ALPN extension is meant to
12949 replace the TLS NPN extension. See also "ssl_fc_npn".
12950
Willy Tarreau74ca5042013-06-11 23:12:07 +020012951ssl_fc_cipher : string
12952 Returns the name of the used cipher when the incoming connection was made
12953 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +020012954
Willy Tarreau74ca5042013-06-11 23:12:07 +020012955ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020012956 Returns true if a client certificate is present in an incoming connection over
12957 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +010012958 Note: on SSL session resumption with Session ID or TLS ticket, client
12959 certificate is not present in the current connection but may be retrieved
12960 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
12961 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +020012962
Willy Tarreau74ca5042013-06-11 23:12:07 +020012963ssl_fc_has_sni : boolean
12964 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020012965 in an incoming connection was made over an SSL/TLS transport layer. Returns
12966 true when the incoming connection presents a TLS SNI field. This requires
12967 that the SSL library is build with support for TLS extensions enabled (check
12968 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +020012969
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020012970ssl_fc_is_resumed: boolean
12971 Returns true if the SSL/TLS session has been resumed through the use of
12972 SSL session cache or TLS tickets.
12973
Willy Tarreau74ca5042013-06-11 23:12:07 +020012974ssl_fc_npn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012975 This extracts the Next Protocol Negotiation field from an incoming connection
Willy Tarreau74ca5042013-06-11 23:12:07 +020012976 made via a TLS transport layer and locally deciphered by haproxy. The result
12977 is a string containing the protocol name advertised by the client. The SSL
12978 library must have been built with support for TLS extensions enabled (check
12979 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
12980 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
12981 forces the client to pick a protocol from this list, any other one may be
12982 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +020012983
Willy Tarreau74ca5042013-06-11 23:12:07 +020012984ssl_fc_protocol : string
12985 Returns the name of the used protocol when the incoming connection was made
12986 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +020012987
Emeric Brunb73a9b02014-04-30 18:49:19 +020012988ssl_fc_unique_id : binary
David Sc1ad52e2014-04-08 18:48:47 -040012989 When the incoming connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020012990 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
12991 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
David Sc1ad52e2014-04-08 18:48:47 -040012992
Willy Tarreau74ca5042013-06-11 23:12:07 +020012993ssl_fc_session_id : binary
12994 Returns the SSL ID of the front connection when the incoming connection was
12995 made over an SSL/TLS transport layer. It is useful to stick a given client to
12996 a server. It is important to note that some browsers refresh their session ID
12997 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +020012998
Willy Tarreau74ca5042013-06-11 23:12:07 +020012999ssl_fc_sni : string
13000 This extracts the Server Name Indication TLS extension (SNI) field from an
13001 incoming connection made via an SSL/TLS transport layer and locally
13002 deciphered by haproxy. The result (when present) typically is a string
13003 matching the HTTPS host name (253 chars or less). The SSL library must have
13004 been built with support for TLS extensions enabled (check haproxy -vv).
13005
13006 This fetch is different from "req_ssl_sni" above in that it applies to the
13007 connection being deciphered by haproxy and not to SSL contents being blindly
13008 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +020013009 requires that the SSL library is build with support for TLS extensions
13010 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +020013011
Willy Tarreau74ca5042013-06-11 23:12:07 +020013012 ACL derivatives :
Willy Tarreau74ca5042013-06-11 23:12:07 +020013013 ssl_fc_sni_end : suffix match
13014 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +020013015
Willy Tarreau74ca5042013-06-11 23:12:07 +020013016ssl_fc_use_keysize : integer
13017 Returns the symmetric cipher key size used in bits when the incoming
13018 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020013019
Willy Tarreaub6fb4202008-07-20 11:18:28 +020013020
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200130217.3.5. Fetching samples from buffer contents (Layer 6)
Willy Tarreau74ca5042013-06-11 23:12:07 +020013022------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +020013023
Willy Tarreau74ca5042013-06-11 23:12:07 +020013024Fetching samples from buffer contents is a bit different from the previous
13025sample fetches above because the sampled data are ephemeral. These data can
13026only be used when they're available and will be lost when they're forwarded.
13027For this reason, samples fetched from buffer contents during a request cannot
13028be used in a response for example. Even while the data are being fetched, they
13029can change. Sometimes it is necessary to set some delays or combine multiple
13030sample fetch methods to ensure that the expected data are complete and usable,
13031for example through TCP request content inspection. Please see the "tcp-request
13032content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +020013033
Willy Tarreau74ca5042013-06-11 23:12:07 +020013034payload(<offset>,<length>) : binary (deprecated)
13035 This is an alias for "req.payload" when used in the context of a request (eg:
13036 "stick on", "stick match"), and for "res.payload" when used in the context of
13037 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010013038
Willy Tarreau74ca5042013-06-11 23:12:07 +020013039payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
13040 This is an alias for "req.payload_lv" when used in the context of a request
13041 (eg: "stick on", "stick match"), and for "res.payload_lv" when used in the
13042 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010013043
Willy Tarreau74ca5042013-06-11 23:12:07 +020013044req.len : integer
13045req_len : integer (deprecated)
13046 Returns an integer value corresponding to the number of bytes present in the
13047 request buffer. This is mostly used in ACL. It is important to understand
13048 that this test does not return false as long as the buffer is changing. This
13049 means that a check with equality to zero will almost always immediately match
13050 at the beginning of the session, while a test for more data will wait for
13051 that data to come in and return false only when haproxy is certain that no
13052 more data will come in. This test was designed to be used with TCP request
13053 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013054
Willy Tarreau74ca5042013-06-11 23:12:07 +020013055req.payload(<offset>,<length>) : binary
13056 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020013057 in the request buffer. As a special case, if the <length> argument is zero,
13058 the the whole buffer from <offset> to the end is extracted. This can be used
13059 with ACLs in order to check for the presence of some content in a buffer at
13060 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013061
Willy Tarreau74ca5042013-06-11 23:12:07 +020013062 ACL alternatives :
13063 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013064
Willy Tarreau74ca5042013-06-11 23:12:07 +020013065req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
13066 This extracts a binary block whose size is specified at <offset1> for <length>
13067 bytes, and which starts at <offset2> if specified or just after the length in
13068 the request buffer. The <offset2> parameter also supports relative offsets if
13069 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013070
Willy Tarreau74ca5042013-06-11 23:12:07 +020013071 ACL alternatives :
13072 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013073
Willy Tarreau74ca5042013-06-11 23:12:07 +020013074 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013075
Willy Tarreau74ca5042013-06-11 23:12:07 +020013076req.proto_http : boolean
13077req_proto_http : boolean (deprecated)
13078 Returns true when data in the request buffer look like HTTP and correctly
13079 parses as such. It is the same parser as the common HTTP request parser which
13080 is used so there should be no surprises. The test does not match until the
13081 request is complete, failed or timed out. This test may be used to report the
13082 protocol in TCP logs, but the biggest use is to block TCP request analysis
13083 until a complete HTTP request is present in the buffer, for example to track
13084 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013085
Willy Tarreau74ca5042013-06-11 23:12:07 +020013086 Example:
13087 # track request counts per "base" (concatenation of Host+URL)
13088 tcp-request inspect-delay 10s
13089 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020013090 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013091
Willy Tarreau74ca5042013-06-11 23:12:07 +020013092req.rdp_cookie([<name>]) : string
13093rdp_cookie([<name>]) : string (deprecated)
13094 When the request buffer looks like the RDP protocol, extracts the RDP cookie
13095 <name>, or any cookie if unspecified. The parser only checks for the first
13096 cookie, as illustrated in the RDP protocol specification. The cookie name is
13097 case insensitive. Generally the "MSTS" cookie name will be used, as it can
13098 contain the user name of the client connecting to the server if properly
13099 configured on the client. The "MSTSHASH" cookie is often used as well for
13100 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013101
Willy Tarreau74ca5042013-06-11 23:12:07 +020013102 This differs from "balance rdp-cookie" in that any balancing algorithm may be
13103 used and thus the distribution of clients to backend servers is not linked to
13104 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
13105 such as "balance roundrobin" or "balance leastconn" will lead to a more even
13106 distribution of clients to backend servers than the hash used by "balance
13107 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013108
Willy Tarreau74ca5042013-06-11 23:12:07 +020013109 ACL derivatives :
13110 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013111
Willy Tarreau74ca5042013-06-11 23:12:07 +020013112 Example :
13113 listen tse-farm
13114 bind 0.0.0.0:3389
13115 # wait up to 5s for an RDP cookie in the request
13116 tcp-request inspect-delay 5s
13117 tcp-request content accept if RDP_COOKIE
13118 # apply RDP cookie persistence
13119 persist rdp-cookie
13120 # Persist based on the mstshash cookie
13121 # This is only useful makes sense if
13122 # balance rdp-cookie is not used
13123 stick-table type string size 204800
13124 stick on req.rdp_cookie(mstshash)
13125 server srv1 1.1.1.1:3389
13126 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013127
Willy Tarreau74ca5042013-06-11 23:12:07 +020013128 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
13129 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013130
Willy Tarreau74ca5042013-06-11 23:12:07 +020013131req.rdp_cookie_cnt([name]) : integer
13132rdp_cookie_cnt([name]) : integer (deprecated)
13133 Tries to parse the request buffer as RDP protocol, then returns an integer
13134 corresponding to the number of RDP cookies found. If an optional cookie name
13135 is passed, only cookies matching this name are considered. This is mostly
13136 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013137
Willy Tarreau74ca5042013-06-11 23:12:07 +020013138 ACL derivatives :
13139 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013140
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020013141req.ssl_ec_ext : boolean
13142 Returns a boolean identifying if client sent the Supported Elliptic Curves
13143 Extension as defined in RFC4492, section 5.1. within the SSL ClientHello
Cyril Bonté307ee1e2015-09-28 23:16:06 +020013144 message. This can be used to present ECC compatible clients with EC
13145 certificate and to use RSA for all others, on the same IP address. Note that
13146 this only applies to raw contents found in the request buffer and not to
13147 contents deciphered via an SSL data layer, so this will not work with "bind"
13148 lines having the "ssl" option.
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020013149
Willy Tarreau74ca5042013-06-11 23:12:07 +020013150req.ssl_hello_type : integer
13151req_ssl_hello_type : integer (deprecated)
13152 Returns an integer value containing the type of the SSL hello message found
13153 in the request buffer if the buffer contains data that parse as a complete
13154 SSL (v3 or superior) client hello message. Note that this only applies to raw
13155 contents found in the request buffer and not to contents deciphered via an
13156 SSL data layer, so this will not work with "bind" lines having the "ssl"
13157 option. This is mostly used in ACL to detect presence of an SSL hello message
13158 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013159
Willy Tarreau74ca5042013-06-11 23:12:07 +020013160req.ssl_sni : string
13161req_ssl_sni : string (deprecated)
13162 Returns a string containing the value of the Server Name TLS extension sent
13163 by a client in a TLS stream passing through the request buffer if the buffer
13164 contains data that parse as a complete SSL (v3 or superior) client hello
13165 message. Note that this only applies to raw contents found in the request
13166 buffer and not to contents deciphered via an SSL data layer, so this will not
13167 work with "bind" lines having the "ssl" option. SNI normally contains the
13168 name of the host the client tries to connect to (for recent browsers). SNI is
13169 useful for allowing or denying access to certain hosts when SSL/TLS is used
13170 by the client. This test was designed to be used with TCP request content
13171 inspection. If content switching is needed, it is recommended to first wait
13172 for a complete client hello (type 1), like in the example below. See also
13173 "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013174
Willy Tarreau74ca5042013-06-11 23:12:07 +020013175 ACL derivatives :
13176 req_ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013177
Willy Tarreau74ca5042013-06-11 23:12:07 +020013178 Examples :
13179 # Wait for a client hello for at most 5 seconds
13180 tcp-request inspect-delay 5s
13181 tcp-request content accept if { req_ssl_hello_type 1 }
13182 use_backend bk_allow if { req_ssl_sni -f allowed_sites }
13183 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013184
Pradeep Jindalbb2acf52015-09-29 10:12:57 +053013185req.ssl_st_ext : integer
13186 Returns 0 if the client didn't send a SessionTicket TLS Extension (RFC5077)
13187 Returns 1 if the client sent SessionTicket TLS Extension
13188 Returns 2 if the client also sent non-zero length TLS SessionTicket
13189 Note that this only applies to raw contents found in the request buffer and
13190 not to contents deciphered via an SSL data layer, so this will not work with
13191 "bind" lines having the "ssl" option. This can for example be used to detect
13192 whether the client sent a SessionTicket or not and stick it accordingly, if
13193 no SessionTicket then stick on SessionID or don't stick as there's no server
13194 side state is there when SessionTickets are in use.
13195
Willy Tarreau74ca5042013-06-11 23:12:07 +020013196req.ssl_ver : integer
13197req_ssl_ver : integer (deprecated)
13198 Returns an integer value containing the version of the SSL/TLS protocol of a
13199 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
13200 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
13201 composed of the major version multiplied by 65536, added to the minor
13202 version. Note that this only applies to raw contents found in the request
13203 buffer and not to contents deciphered via an SSL data layer, so this will not
13204 work with "bind" lines having the "ssl" option. The ACL version of the test
13205 matches against a decimal notation in the form MAJOR.MINOR (eg: 3.1). This
13206 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013207
Willy Tarreau74ca5042013-06-11 23:12:07 +020013208 ACL derivatives :
13209 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +010013210
Willy Tarreau47e8eba2013-09-11 23:28:46 +020013211res.len : integer
13212 Returns an integer value corresponding to the number of bytes present in the
13213 response buffer. This is mostly used in ACL. It is important to understand
13214 that this test does not return false as long as the buffer is changing. This
13215 means that a check with equality to zero will almost always immediately match
13216 at the beginning of the session, while a test for more data will wait for
13217 that data to come in and return false only when haproxy is certain that no
13218 more data will come in. This test was designed to be used with TCP response
13219 content inspection.
13220
Willy Tarreau74ca5042013-06-11 23:12:07 +020013221res.payload(<offset>,<length>) : binary
13222 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020013223 in the response buffer. As a special case, if the <length> argument is zero,
13224 the the whole buffer from <offset> to the end is extracted. This can be used
13225 with ACLs in order to check for the presence of some content in a buffer at
13226 any location.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013227
Willy Tarreau74ca5042013-06-11 23:12:07 +020013228res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
13229 This extracts a binary block whose size is specified at <offset1> for <length>
13230 bytes, and which starts at <offset2> if specified or just after the length in
13231 the response buffer. The <offset2> parameter also supports relative offsets
13232 if prepended with a '+' or '-' sign.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013233
Willy Tarreau74ca5042013-06-11 23:12:07 +020013234 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013235
Willy Tarreau971f7b62015-09-29 14:06:59 +020013236res.ssl_hello_type : integer
13237rep_ssl_hello_type : integer (deprecated)
13238 Returns an integer value containing the type of the SSL hello message found
13239 in the response buffer if the buffer contains data that parses as a complete
13240 SSL (v3 or superior) hello message. Note that this only applies to raw
13241 contents found in the response buffer and not to contents deciphered via an
13242 SSL data layer, so this will not work with "server" lines having the "ssl"
13243 option. This is mostly used in ACL to detect presence of an SSL hello message
13244 that is supposed to contain an SSL session ID usable for stickiness.
13245
Willy Tarreau74ca5042013-06-11 23:12:07 +020013246wait_end : boolean
13247 This fetch either returns true when the inspection period is over, or does
13248 not fetch. It is only used in ACLs, in conjunction with content analysis to
13249 avoid returning a wrong verdict early. It may also be used to delay some
13250 actions, such as a delayed reject for some special addresses. Since it either
13251 stops the rules evaluation or immediately returns true, it is recommended to
13252 use this acl as the last one in a rule. Please note that the default ACL
13253 "WAIT_END" is always usable without prior declaration. This test was designed
13254 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013255
Willy Tarreau74ca5042013-06-11 23:12:07 +020013256 Examples :
13257 # delay every incoming request by 2 seconds
13258 tcp-request inspect-delay 2s
13259 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +010013260
Willy Tarreau74ca5042013-06-11 23:12:07 +020013261 # don't immediately tell bad guys they are rejected
13262 tcp-request inspect-delay 10s
13263 acl goodguys src 10.0.0.0/24
13264 acl badguys src 10.0.1.0/24
13265 tcp-request content accept if goodguys
13266 tcp-request content reject if badguys WAIT_END
13267 tcp-request content reject
13268
13269
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200132707.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +020013271--------------------------------------
13272
13273It is possible to fetch samples from HTTP contents, requests and responses.
13274This application layer is also called layer 7. It is only possible to fetch the
13275data in this section when a full HTTP request or response has been parsed from
13276its respective request or response buffer. This is always the case with all
13277HTTP specific rules and for sections running with "mode http". When using TCP
13278content inspection, it may be necessary to support an inspection delay in order
13279to let the request or response come in first. These fetches may require a bit
13280more CPU resources than the layer 4 ones, but not much since the request and
13281response are indexed.
13282
13283base : string
13284 This returns the concatenation of the first Host header and the path part of
13285 the request, which starts at the first slash and ends before the question
13286 mark. It can be useful in virtual hosted environments to detect URL abuses as
13287 well as to improve shared caches efficiency. Using this with a limited size
13288 stick table also allows one to collect statistics about most commonly
13289 requested objects by host/path. With ACLs it can allow simple content
13290 switching rules involving the host and the path at the same time, such as
13291 "www.example.com/favicon.ico". See also "path" and "uri".
13292
13293 ACL derivatives :
13294 base : exact string match
13295 base_beg : prefix match
13296 base_dir : subdir match
13297 base_dom : domain match
13298 base_end : suffix match
13299 base_len : length match
13300 base_reg : regex match
13301 base_sub : substring match
13302
13303base32 : integer
13304 This returns a 32-bit hash of the value returned by the "base" fetch method
13305 above. This is useful to track per-URL activity on high traffic sites without
13306 having to store all URLs. Instead a shorter hash is stored, saving a lot of
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013307 memory. The output type is an unsigned integer. The hash function used is
13308 SDBM with full avalanche on the output. Technically, base32 is exactly equal
13309 to "base,sdbm(1)".
Willy Tarreau74ca5042013-06-11 23:12:07 +020013310
13311base32+src : binary
13312 This returns the concatenation of the base32 fetch above and the src fetch
13313 below. The resulting type is of type binary, with a size of 8 or 20 bytes
13314 depending on the source address family. This can be used to track per-IP,
13315 per-URL counters.
13316
William Lallemand65ad6e12014-01-31 15:08:02 +010013317capture.req.hdr(<idx>) : string
13318 This extracts the content of the header captured by the "capture request
13319 header", idx is the position of the capture keyword in the configuration.
13320 The first entry is an index of 0. See also: "capture request header".
13321
13322capture.req.method : string
13323 This extracts the METHOD of an HTTP request. It can be used in both request
13324 and response. Unlike "method", it can be used in both request and response
13325 because it's allocated.
13326
13327capture.req.uri : string
13328 This extracts the request's URI, which starts at the first slash and ends
13329 before the first space in the request (without the host part). Unlike "path"
13330 and "url", it can be used in both request and response because it's
13331 allocated.
13332
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020013333capture.req.ver : string
13334 This extracts the request's HTTP version and returns either "HTTP/1.0" or
13335 "HTTP/1.1". Unlike "req.ver", it can be used in both request, response, and
13336 logs because it relies on a persistent flag.
13337
William Lallemand65ad6e12014-01-31 15:08:02 +010013338capture.res.hdr(<idx>) : string
13339 This extracts the content of the header captured by the "capture response
13340 header", idx is the position of the capture keyword in the configuration.
13341 The first entry is an index of 0.
13342 See also: "capture response header"
13343
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020013344capture.res.ver : string
13345 This extracts the response's HTTP version and returns either "HTTP/1.0" or
13346 "HTTP/1.1". Unlike "res.ver", it can be used in logs because it relies on a
13347 persistent flag.
13348
Willy Tarreaua5910cc2015-05-02 00:46:08 +020013349req.body : binary
13350 This returns the HTTP request's available body as a block of data. It
13351 requires that the request body has been buffered made available using
13352 "option http-buffer-request". In case of chunked-encoded body, currently only
13353 the first chunk is analyzed.
13354
Thierry FOURNIER9826c772015-05-20 15:50:54 +020013355req.body_param([<name>) : string
13356 This fetch assumes that the body of the POST request is url-encoded. The user
13357 can check if the "content-type" contains the value
13358 "application/x-www-form-urlencoded". This extracts the first occurrence of the
13359 parameter <name> in the body, which ends before '&'. The parameter name is
13360 case-sensitive. If no name is given, any parameter will match, and the first
13361 one will be returned. The result is a string corresponding to the value of the
13362 parameter <name> as presented in the request body (no URL decoding is
13363 performed). Note that the ACL version of this fetch iterates over multiple
13364 parameters and will iteratively report all parameters values if no name is
13365 given.
13366
Willy Tarreaua5910cc2015-05-02 00:46:08 +020013367req.body_len : integer
13368 This returns the length of the HTTP request's available body in bytes. It may
13369 be lower than the advertised length if the body is larger than the buffer. It
13370 requires that the request body has been buffered made available using
13371 "option http-buffer-request".
13372
13373req.body_size : integer
13374 This returns the advertised length of the HTTP request's body in bytes. It
13375 will represent the advertised Content-Length header, or the size of the first
13376 chunk in case of chunked encoding. In order to parse the chunks, it requires
13377 that the request body has been buffered made available using
13378 "option http-buffer-request".
13379
Willy Tarreau74ca5042013-06-11 23:12:07 +020013380req.cook([<name>]) : string
13381cook([<name>]) : string (deprecated)
13382 This extracts the last occurrence of the cookie name <name> on a "Cookie"
13383 header line from the request, and returns its value as string. If no name is
13384 specified, the first cookie value is returned. When used with ACLs, all
13385 matching cookies are evaluated. Spaces around the name and the value are
13386 ignored as requested by the Cookie header specification (RFC6265). The cookie
13387 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
13388 well return an empty value if it is present. Use the "found" match to detect
13389 presence. Use the res.cook() variant for response cookies sent by the server.
13390
13391 ACL derivatives :
13392 cook([<name>]) : exact string match
13393 cook_beg([<name>]) : prefix match
13394 cook_dir([<name>]) : subdir match
13395 cook_dom([<name>]) : domain match
13396 cook_end([<name>]) : suffix match
13397 cook_len([<name>]) : length match
13398 cook_reg([<name>]) : regex match
13399 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +010013400
Willy Tarreau74ca5042013-06-11 23:12:07 +020013401req.cook_cnt([<name>]) : integer
13402cook_cnt([<name>]) : integer (deprecated)
13403 Returns an integer value representing the number of occurrences of the cookie
13404 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013405
Willy Tarreau74ca5042013-06-11 23:12:07 +020013406req.cook_val([<name>]) : integer
13407cook_val([<name>]) : integer (deprecated)
13408 This extracts the last occurrence of the cookie name <name> on a "Cookie"
13409 header line from the request, and converts its value to an integer which is
13410 returned. If no name is specified, the first cookie value is returned. When
13411 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +020013412
Willy Tarreau74ca5042013-06-11 23:12:07 +020013413cookie([<name>]) : string (deprecated)
13414 This extracts the last occurrence of the cookie name <name> on a "Cookie"
13415 header line from the request, or a "Set-Cookie" header from the response, and
13416 returns its value as a string. A typical use is to get multiple clients
13417 sharing a same profile use the same server. This can be similar to what
Willy Tarreau294d0f02015-08-10 19:40:12 +020013418 "appsession" did with the "request-learn" statement, but with support for
Willy Tarreau74ca5042013-06-11 23:12:07 +020013419 multi-peer synchronization and state keeping across restarts. If no name is
13420 specified, the first cookie value is returned. This fetch should not be used
13421 anymore and should be replaced by req.cook() or res.cook() instead as it
13422 ambiguously uses the direction based on the context where it is used.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013423
Willy Tarreau74ca5042013-06-11 23:12:07 +020013424hdr([<name>[,<occ>]]) : string
13425 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
13426 used on responses. Please refer to these respective fetches for more details.
13427 In case of doubt about the fetch direction, please use the explicit ones.
13428 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013429 unambiguously apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013430
Willy Tarreau74ca5042013-06-11 23:12:07 +020013431req.fhdr(<name>[,<occ>]) : string
13432 This extracts the last occurrence of header <name> in an HTTP request. When
13433 used from an ACL, all occurrences are iterated over until a match is found.
13434 Optionally, a specific occurrence might be specified as a position number.
13435 Positive values indicate a position from the first occurrence, with 1 being
13436 the first one. Negative values indicate positions relative to the last one,
13437 with -1 being the last one. It differs from req.hdr() in that any commas
13438 present in the value are returned and are not used as delimiters. This is
13439 sometimes useful with headers such as User-Agent.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013440
Willy Tarreau74ca5042013-06-11 23:12:07 +020013441req.fhdr_cnt([<name>]) : integer
13442 Returns an integer value representing the number of occurrences of request
13443 header field name <name>, or the total number of header fields if <name> is
13444 not specified. Contrary to its req.hdr_cnt() cousin, this function returns
13445 the number of full line headers and does not stop on commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013446
Willy Tarreau74ca5042013-06-11 23:12:07 +020013447req.hdr([<name>[,<occ>]]) : string
13448 This extracts the last occurrence of header <name> in an HTTP request. When
13449 used from an ACL, all occurrences are iterated over until a match is found.
13450 Optionally, a specific occurrence might be specified as a position number.
13451 Positive values indicate a position from the first occurrence, with 1 being
13452 the first one. Negative values indicate positions relative to the last one,
13453 with -1 being the last one. A typical use is with the X-Forwarded-For header
13454 once converted to IP, associated with an IP stick-table. The function
13455 considers any comma as a delimiter for distinct values. If full-line headers
13456 are desired instead, use req.fhdr(). Please carefully check RFC2616 to know
13457 how certain headers are supposed to be parsed. Also, some of them are case
13458 insensitive (eg: Connection).
Willy Tarreaud63335a2010-02-26 12:56:52 +010013459
Willy Tarreau74ca5042013-06-11 23:12:07 +020013460 ACL derivatives :
13461 hdr([<name>[,<occ>]]) : exact string match
13462 hdr_beg([<name>[,<occ>]]) : prefix match
13463 hdr_dir([<name>[,<occ>]]) : subdir match
13464 hdr_dom([<name>[,<occ>]]) : domain match
13465 hdr_end([<name>[,<occ>]]) : suffix match
13466 hdr_len([<name>[,<occ>]]) : length match
13467 hdr_reg([<name>[,<occ>]]) : regex match
13468 hdr_sub([<name>[,<occ>]]) : substring match
13469
13470req.hdr_cnt([<name>]) : integer
13471hdr_cnt([<header>]) : integer (deprecated)
13472 Returns an integer value representing the number of occurrences of request
13473 header field name <name>, or the total number of header field values if
13474 <name> is not specified. It is important to remember that one header line may
13475 count as several headers if it has several values. The function considers any
13476 comma as a delimiter for distinct values. If full-line headers are desired
13477 instead, req.fhdr_cnt() should be used instead. With ACLs, it can be used to
13478 detect presence, absence or abuse of a specific header, as well as to block
13479 request smuggling attacks by rejecting requests which contain more than one
13480 of certain headers. See "req.hdr" for more information on header matching.
13481
13482req.hdr_ip([<name>[,<occ>]]) : ip
13483hdr_ip([<name>[,<occ>]]) : ip (deprecated)
13484 This extracts the last occurrence of header <name> in an HTTP request,
13485 converts it to an IPv4 or IPv6 address and returns this address. When used
13486 with ACLs, all occurrences are checked, and if <name> is omitted, every value
13487 of every header is checked. Optionally, a specific occurrence might be
13488 specified as a position number. Positive values indicate a position from the
13489 first occurrence, with 1 being the first one. Negative values indicate
13490 positions relative to the last one, with -1 being the last one. A typical use
13491 is with the X-Forwarded-For and X-Client-IP headers.
13492
13493req.hdr_val([<name>[,<occ>]]) : integer
13494hdr_val([<name>[,<occ>]]) : integer (deprecated)
13495 This extracts the last occurrence of header <name> in an HTTP request, and
13496 converts it to an integer value. When used with ACLs, all occurrences are
13497 checked, and if <name> is omitted, every value of every header is checked.
13498 Optionally, a specific occurrence might be specified as a position number.
13499 Positive values indicate a position from the first occurrence, with 1 being
13500 the first one. Negative values indicate positions relative to the last one,
13501 with -1 being the last one. A typical use is with the X-Forwarded-For header.
13502
13503http_auth(<userlist>) : boolean
13504 Returns a boolean indicating whether the authentication data received from
13505 the client match a username & password stored in the specified userlist. This
13506 fetch function is not really useful outside of ACLs. Currently only http
13507 basic auth is supported.
13508
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010013509http_auth_group(<userlist>) : string
13510 Returns a string corresponding to the user name found in the authentication
13511 data received from the client if both the user name and password are valid
13512 according to the specified userlist. The main purpose is to use it in ACLs
13513 where it is then checked whether the user belongs to any group within a list.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013514 This fetch function is not really useful outside of ACLs. Currently only http
13515 basic auth is supported.
13516
13517 ACL derivatives :
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010013518 http_auth_group(<userlist>) : group ...
13519 Returns true when the user extracted from the request and whose password is
13520 valid according to the specified userlist belongs to at least one of the
13521 groups.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013522
13523http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020013524 Returns true when the request being processed is the first one of the
13525 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020013526 from some requests when a request is not the first one, or to help grouping
13527 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020013528
Willy Tarreau74ca5042013-06-11 23:12:07 +020013529method : integer + string
13530 Returns an integer value corresponding to the method in the HTTP request. For
13531 example, "GET" equals 1 (check sources to establish the matching). Value 9
13532 means "other method" and may be converted to a string extracted from the
13533 stream. This should not be used directly as a sample, this is only meant to
13534 be used from ACLs, which transparently convert methods from patterns to these
13535 integer + string values. Some predefined ACL already check for most common
13536 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013537
Willy Tarreau74ca5042013-06-11 23:12:07 +020013538 ACL derivatives :
13539 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020013540
Willy Tarreau74ca5042013-06-11 23:12:07 +020013541 Example :
13542 # only accept GET and HEAD requests
13543 acl valid_method method GET HEAD
13544 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020013545
Willy Tarreau74ca5042013-06-11 23:12:07 +020013546path : string
13547 This extracts the request's URL path, which starts at the first slash and
13548 ends before the question mark (without the host part). A typical use is with
13549 prefetch-capable caches, and with portals which need to aggregate multiple
13550 information from databases and keep them in caches. Note that with outgoing
13551 caches, it would be wiser to use "url" instead. With ACLs, it's typically
13552 used to match exact file names (eg: "/login.php"), or directory parts using
13553 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013554
Willy Tarreau74ca5042013-06-11 23:12:07 +020013555 ACL derivatives :
13556 path : exact string match
13557 path_beg : prefix match
13558 path_dir : subdir match
13559 path_dom : domain match
13560 path_end : suffix match
13561 path_len : length match
13562 path_reg : regex match
13563 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020013564
Willy Tarreau49ad95c2015-01-19 15:06:26 +010013565query : string
13566 This extracts the request's query string, which starts after the first
13567 question mark. If no question mark is present, this fetch returns nothing. If
13568 a question mark is present but nothing follows, it returns an empty string.
13569 This means it's possible to easily know whether a query string is present
13570 using the "found" matching method. This fetch is the completemnt of "path"
13571 which stops before the question mark.
13572
Willy Tarreaueb27ec72015-02-20 13:55:29 +010013573req.hdr_names([<delim>]) : string
13574 This builds a string made from the concatenation of all header names as they
13575 appear in the request when the rule is evaluated. The default delimiter is
13576 the comma (',') but it may be overridden as an optional argument <delim>. In
13577 this case, only the first character of <delim> is considered.
13578
Willy Tarreau74ca5042013-06-11 23:12:07 +020013579req.ver : string
13580req_ver : string (deprecated)
13581 Returns the version string from the HTTP request, for example "1.1". This can
13582 be useful for logs, but is mostly there for ACL. Some predefined ACL already
13583 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013584
Willy Tarreau74ca5042013-06-11 23:12:07 +020013585 ACL derivatives :
13586 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020013587
Willy Tarreau74ca5042013-06-11 23:12:07 +020013588res.comp : boolean
13589 Returns the boolean "true" value if the response has been compressed by
13590 HAProxy, otherwise returns boolean "false". This may be used to add
13591 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013592
Willy Tarreau74ca5042013-06-11 23:12:07 +020013593res.comp_algo : string
13594 Returns a string containing the name of the algorithm used if the response
13595 was compressed by HAProxy, for example : "deflate". This may be used to add
13596 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013597
Willy Tarreau74ca5042013-06-11 23:12:07 +020013598res.cook([<name>]) : string
13599scook([<name>]) : string (deprecated)
13600 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
13601 header line from the response, and returns its value as string. If no name is
13602 specified, the first cookie value is returned.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020013603
Willy Tarreau74ca5042013-06-11 23:12:07 +020013604 ACL derivatives :
13605 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020013606
Willy Tarreau74ca5042013-06-11 23:12:07 +020013607res.cook_cnt([<name>]) : integer
13608scook_cnt([<name>]) : integer (deprecated)
13609 Returns an integer value representing the number of occurrences of the cookie
13610 <name> in the response, or all cookies if <name> is not specified. This is
13611 mostly useful when combined with ACLs to detect suspicious responses.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013612
Willy Tarreau74ca5042013-06-11 23:12:07 +020013613res.cook_val([<name>]) : integer
13614scook_val([<name>]) : integer (deprecated)
13615 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
13616 header line from the response, and converts its value to an integer which is
13617 returned. If no name is specified, the first cookie value is returned.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013618
Willy Tarreau74ca5042013-06-11 23:12:07 +020013619res.fhdr([<name>[,<occ>]]) : string
13620 This extracts the last occurrence of header <name> in an HTTP response, or of
13621 the last header if no <name> is specified. Optionally, a specific occurrence
13622 might be specified as a position number. Positive values indicate a position
13623 from the first occurrence, with 1 being the first one. Negative values
13624 indicate positions relative to the last one, with -1 being the last one. It
13625 differs from res.hdr() in that any commas present in the value are returned
13626 and are not used as delimiters. If this is not desired, the res.hdr() fetch
13627 should be used instead. This is sometimes useful with headers such as Date or
13628 Expires.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013629
Willy Tarreau74ca5042013-06-11 23:12:07 +020013630res.fhdr_cnt([<name>]) : integer
13631 Returns an integer value representing the number of occurrences of response
13632 header field name <name>, or the total number of header fields if <name> is
13633 not specified. Contrary to its res.hdr_cnt() cousin, this function returns
13634 the number of full line headers and does not stop on commas. If this is not
13635 desired, the res.hdr_cnt() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013636
Willy Tarreau74ca5042013-06-11 23:12:07 +020013637res.hdr([<name>[,<occ>]]) : string
13638shdr([<name>[,<occ>]]) : string (deprecated)
13639 This extracts the last occurrence of header <name> in an HTTP response, or of
13640 the last header if no <name> is specified. Optionally, a specific occurrence
13641 might be specified as a position number. Positive values indicate a position
13642 from the first occurrence, with 1 being the first one. Negative values
13643 indicate positions relative to the last one, with -1 being the last one. This
13644 can be useful to learn some data into a stick-table. The function considers
13645 any comma as a delimiter for distinct values. If this is not desired, the
13646 res.fhdr() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013647
Willy Tarreau74ca5042013-06-11 23:12:07 +020013648 ACL derivatives :
13649 shdr([<name>[,<occ>]]) : exact string match
13650 shdr_beg([<name>[,<occ>]]) : prefix match
13651 shdr_dir([<name>[,<occ>]]) : subdir match
13652 shdr_dom([<name>[,<occ>]]) : domain match
13653 shdr_end([<name>[,<occ>]]) : suffix match
13654 shdr_len([<name>[,<occ>]]) : length match
13655 shdr_reg([<name>[,<occ>]]) : regex match
13656 shdr_sub([<name>[,<occ>]]) : substring match
13657
13658res.hdr_cnt([<name>]) : integer
13659shdr_cnt([<name>]) : integer (deprecated)
13660 Returns an integer value representing the number of occurrences of response
13661 header field name <name>, or the total number of header fields if <name> is
13662 not specified. The function considers any comma as a delimiter for distinct
13663 values. If this is not desired, the res.fhdr_cnt() fetch should be used
13664 instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013665
Willy Tarreau74ca5042013-06-11 23:12:07 +020013666res.hdr_ip([<name>[,<occ>]]) : ip
13667shdr_ip([<name>[,<occ>]]) : ip (deprecated)
13668 This extracts the last occurrence of header <name> in an HTTP response,
13669 convert it to an IPv4 or IPv6 address and returns this address. Optionally, a
13670 specific occurrence might be specified as a position number. Positive values
13671 indicate a position from the first occurrence, with 1 being the first one.
13672 Negative values indicate positions relative to the last one, with -1 being
13673 the last one. This can be useful to learn some data into a stick table.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013674
Willy Tarreaueb27ec72015-02-20 13:55:29 +010013675res.hdr_names([<delim>]) : string
13676 This builds a string made from the concatenation of all header names as they
13677 appear in the response when the rule is evaluated. The default delimiter is
13678 the comma (',') but it may be overridden as an optional argument <delim>. In
13679 this case, only the first character of <delim> is considered.
13680
Willy Tarreau74ca5042013-06-11 23:12:07 +020013681res.hdr_val([<name>[,<occ>]]) : integer
13682shdr_val([<name>[,<occ>]]) : integer (deprecated)
13683 This extracts the last occurrence of header <name> in an HTTP response, and
13684 converts it to an integer value. Optionally, a specific occurrence might be
13685 specified as a position number. Positive values indicate a position from the
13686 first occurrence, with 1 being the first one. Negative values indicate
13687 positions relative to the last one, with -1 being the last one. This can be
13688 useful to learn some data into a stick table.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010013689
Willy Tarreau74ca5042013-06-11 23:12:07 +020013690res.ver : string
13691resp_ver : string (deprecated)
13692 Returns the version string from the HTTP response, for example "1.1". This
13693 can be useful for logs, but is mostly there for ACL.
Willy Tarreau0e698542011-09-16 08:32:32 +020013694
Willy Tarreau74ca5042013-06-11 23:12:07 +020013695 ACL derivatives :
13696 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010013697
Willy Tarreau74ca5042013-06-11 23:12:07 +020013698set-cookie([<name>]) : string (deprecated)
13699 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
13700 header line from the response and uses the corresponding value to match. This
Willy Tarreau294d0f02015-08-10 19:40:12 +020013701 can be comparable to what "appsession" did with default options, but with
Willy Tarreau74ca5042013-06-11 23:12:07 +020013702 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010013703
Willy Tarreau74ca5042013-06-11 23:12:07 +020013704 This fetch function is deprecated and has been superseded by the "res.cook"
13705 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010013706
Willy Tarreau74ca5042013-06-11 23:12:07 +020013707status : integer
13708 Returns an integer containing the HTTP status code in the HTTP response, for
13709 example, 302. It is mostly used within ACLs and integer ranges, for example,
13710 to remove any Location header if the response is not a 3xx.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013711
Willy Tarreau74ca5042013-06-11 23:12:07 +020013712url : string
13713 This extracts the request's URL as presented in the request. A typical use is
13714 with prefetch-capable caches, and with portals which need to aggregate
13715 multiple information from databases and keep them in caches. With ACLs, using
13716 "path" is preferred over using "url", because clients may send a full URL as
13717 is normally done with proxies. The only real use is to match "*" which does
13718 not match in "path", and for which there is already a predefined ACL. See
13719 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013720
Willy Tarreau74ca5042013-06-11 23:12:07 +020013721 ACL derivatives :
13722 url : exact string match
13723 url_beg : prefix match
13724 url_dir : subdir match
13725 url_dom : domain match
13726 url_end : suffix match
13727 url_len : length match
13728 url_reg : regex match
13729 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013730
Willy Tarreau74ca5042013-06-11 23:12:07 +020013731url_ip : ip
13732 This extracts the IP address from the request's URL when the host part is
13733 presented as an IP address. Its use is very limited. For instance, a
13734 monitoring system might use this field as an alternative for the source IP in
13735 order to test what path a given source address would follow, or to force an
13736 entry in a table for a given source address. With ACLs it can be used to
13737 restrict access to certain systems through a proxy, for example when combined
13738 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013739
Willy Tarreau74ca5042013-06-11 23:12:07 +020013740url_port : integer
13741 This extracts the port part from the request's URL. Note that if the port is
13742 not specified in the request, port 80 is assumed. With ACLs it can be used to
13743 restrict access to certain systems through a proxy, for example when combined
13744 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013745
Willy Tarreau1ede1da2015-05-07 16:06:18 +020013746urlp([<name>[,<delim>]]) : string
13747url_param([<name>[,<delim>]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020013748 This extracts the first occurrence of the parameter <name> in the query
13749 string, which begins after either '?' or <delim>, and which ends before '&',
Willy Tarreau1ede1da2015-05-07 16:06:18 +020013750 ';' or <delim>. The parameter name is case-sensitive. If no name is given,
13751 any parameter will match, and the first one will be returned. The result is
13752 a string corresponding to the value of the parameter <name> as presented in
13753 the request (no URL decoding is performed). This can be used for session
Willy Tarreau74ca5042013-06-11 23:12:07 +020013754 stickiness based on a client ID, to extract an application cookie passed as a
13755 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
Willy Tarreau1ede1da2015-05-07 16:06:18 +020013756 this fetch iterates over multiple parameters and will iteratively report all
13757 parameters values if no name is given
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013758
Willy Tarreau74ca5042013-06-11 23:12:07 +020013759 ACL derivatives :
13760 urlp(<name>[,<delim>]) : exact string match
13761 urlp_beg(<name>[,<delim>]) : prefix match
13762 urlp_dir(<name>[,<delim>]) : subdir match
13763 urlp_dom(<name>[,<delim>]) : domain match
13764 urlp_end(<name>[,<delim>]) : suffix match
13765 urlp_len(<name>[,<delim>]) : length match
13766 urlp_reg(<name>[,<delim>]) : regex match
13767 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013768
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013769
Willy Tarreau74ca5042013-06-11 23:12:07 +020013770 Example :
13771 # match http://example.com/foo?PHPSESSIONID=some_id
13772 stick on urlp(PHPSESSIONID)
13773 # match http://example.com/foo;JSESSIONID=some_id
13774 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013775
Willy Tarreau1ede1da2015-05-07 16:06:18 +020013776urlp_val([<name>[,<delim>])] : integer
Willy Tarreau74ca5042013-06-11 23:12:07 +020013777 See "urlp" above. This one extracts the URL parameter <name> in the request
13778 and converts it to an integer value. This can be used for session stickiness
13779 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020013780
Willy Tarreau198a7442008-01-17 12:05:32 +010013781
Willy Tarreau74ca5042013-06-11 23:12:07 +0200137827.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013783---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010013784
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013785Some predefined ACLs are hard-coded so that they do not have to be declared in
13786every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020013787order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010013788
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013789ACL name Equivalent to Usage
13790---------------+-----------------------------+---------------------------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013791FALSE always_false never match
Willy Tarreau2492d5b2009-07-11 00:06:00 +020013792HTTP req_proto_http match if protocol is valid HTTP
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013793HTTP_1.0 req_ver 1.0 match HTTP version 1.0
13794HTTP_1.1 req_ver 1.1 match HTTP version 1.1
Willy Tarreaud63335a2010-02-26 12:56:52 +010013795HTTP_CONTENT hdr_val(content-length) gt 0 match an existing content-length
13796HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
13797HTTP_URL_SLASH url_beg / match URL beginning with "/"
13798HTTP_URL_STAR url * match URL equal to "*"
13799LOCALHOST src 127.0.0.1/8 match connection from local host
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013800METH_CONNECT method CONNECT match HTTP CONNECT method
13801METH_GET method GET HEAD match HTTP GET or HEAD method
13802METH_HEAD method HEAD match HTTP HEAD method
13803METH_OPTIONS method OPTIONS match HTTP OPTIONS method
13804METH_POST method POST match HTTP POST method
13805METH_TRACE method TRACE match HTTP TRACE method
Emeric Brunbede3d02009-06-30 17:54:00 +020013806RDP_COOKIE req_rdp_cookie_cnt gt 0 match presence of an RDP cookie
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013807REQ_CONTENT req_len gt 0 match data in the request buffer
Willy Tarreaud63335a2010-02-26 12:56:52 +010013808TRUE always_true always match
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013809WAIT_END wait_end wait for end of content analysis
13810---------------+-----------------------------+---------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010013811
Willy Tarreaub937b7e2010-01-12 15:27:54 +010013812
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200138138. Logging
13814----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010013815
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013816One of HAProxy's strong points certainly lies is its precise logs. It probably
13817provides the finest level of information available for such a product, which is
13818very important for troubleshooting complex environments. Standard information
13819provided in logs include client ports, TCP/HTTP state timers, precise session
13820state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010013821to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013822headers.
13823
13824In order to improve administrators reactivity, it offers a great transparency
13825about encountered problems, both internal and external, and it is possible to
13826send logs to different sources at the same time with different level filters :
13827
13828 - global process-level logs (system errors, start/stop, etc..)
13829 - per-instance system and internal errors (lack of resource, bugs, ...)
13830 - per-instance external troubles (servers up/down, max connections)
13831 - per-instance activity (client connections), either at the establishment or
13832 at the termination.
Jim Freeman9e8714b2015-05-26 09:16:34 -060013833 - per-request control of log-level, eg:
13834 http-request set-log-level silent if sensitive_request
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013835
13836The ability to distribute different levels of logs to different log servers
13837allow several production teams to interact and to fix their problems as soon
13838as possible. For example, the system team might monitor system-wide errors,
13839while the application team might be monitoring the up/down for their servers in
13840real time, and the security team might analyze the activity logs with one hour
13841delay.
13842
13843
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200138448.1. Log levels
13845---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013846
Simon Hormandf791f52011-05-29 15:01:10 +090013847TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013848source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090013849HTTP request, HTTP return code, number of bytes transmitted, conditions
13850in which the session ended, and even exchanged cookies values. For example
13851track a particular user's problems. All messages may be sent to up to two
13852syslog servers. Check the "log" keyword in section 4.2 for more information
13853about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013854
13855
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200138568.2. Log formats
13857----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013858
William Lallemand48940402012-01-30 16:47:22 +010013859HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090013860and will be detailed in the following sections. A few of them may vary
13861slightly with the configuration, due to indicators specific to certain
13862options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013863
13864 - the default format, which is very basic and very rarely used. It only
13865 provides very basic information about the incoming connection at the moment
13866 it is accepted : source IP:port, destination IP:port, and frontend-name.
13867 This mode will eventually disappear so it will not be described to great
13868 extents.
13869
13870 - the TCP format, which is more advanced. This format is enabled when "option
13871 tcplog" is set on the frontend. HAProxy will then usually wait for the
13872 connection to terminate before logging. This format provides much richer
13873 information, such as timers, connection counts, queue size, etc... This
13874 format is recommended for pure TCP proxies.
13875
13876 - the HTTP format, which is the most advanced for HTTP proxying. This format
13877 is enabled when "option httplog" is set on the frontend. It provides the
13878 same information as the TCP format with some HTTP-specific fields such as
13879 the request, the status code, and captures of headers and cookies. This
13880 format is recommended for HTTP proxies.
13881
Emeric Brun3a058f32009-06-30 18:26:00 +020013882 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
13883 fields arranged in the same order as the CLF format. In this mode, all
13884 timers, captures, flags, etc... appear one per field after the end of the
13885 common fields, in the same order they appear in the standard HTTP format.
13886
William Lallemand48940402012-01-30 16:47:22 +010013887 - the custom log format, allows you to make your own log line.
13888
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013889Next sections will go deeper into details for each of these formats. Format
13890specification will be performed on a "field" basis. Unless stated otherwise, a
13891field is a portion of text delimited by any number of spaces. Since syslog
13892servers are susceptible of inserting fields at the beginning of a line, it is
13893always assumed that the first field is the one containing the process name and
13894identifier.
13895
13896Note : Since log lines may be quite long, the log examples in sections below
13897 might be broken into multiple lines. The example log lines will be
13898 prefixed with 3 closing angle brackets ('>>>') and each time a log is
13899 broken into multiple lines, each non-final line will end with a
13900 backslash ('\') and the next line will start indented by two characters.
13901
13902
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200139038.2.1. Default log format
13904-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013905
13906This format is used when no specific option is set. The log is emitted as soon
13907as the connection is accepted. One should note that this currently is the only
13908format which logs the request's destination IP and ports.
13909
13910 Example :
13911 listen www
13912 mode http
13913 log global
13914 server srv1 127.0.0.1:8000
13915
13916 >>> Feb 6 12:12:09 localhost \
13917 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
13918 (www/HTTP)
13919
13920 Field Format Extract from the example above
13921 1 process_name '[' pid ']:' haproxy[14385]:
13922 2 'Connect from' Connect from
13923 3 source_ip ':' source_port 10.0.1.2:33312
13924 4 'to' to
13925 5 destination_ip ':' destination_port 10.0.3.31:8012
13926 6 '(' frontend_name '/' mode ')' (www/HTTP)
13927
13928Detailed fields description :
13929 - "source_ip" is the IP address of the client which initiated the connection.
13930 - "source_port" is the TCP port of the client which initiated the connection.
13931 - "destination_ip" is the IP address the client connected to.
13932 - "destination_port" is the TCP port the client connected to.
13933 - "frontend_name" is the name of the frontend (or listener) which received
13934 and processed the connection.
13935 - "mode is the mode the frontend is operating (TCP or HTTP).
13936
Willy Tarreauceb24bc2010-11-09 12:46:41 +010013937In case of a UNIX socket, the source and destination addresses are marked as
13938"unix:" and the ports reflect the internal ID of the socket which accepted the
13939connection (the same ID as reported in the stats).
13940
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013941It is advised not to use this deprecated format for newer installations as it
13942will eventually disappear.
13943
13944
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200139458.2.2. TCP log format
13946---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013947
13948The TCP format is used when "option tcplog" is specified in the frontend, and
13949is the recommended format for pure TCP proxies. It provides a lot of precious
13950information for troubleshooting. Since this format includes timers and byte
13951counts, the log is normally emitted at the end of the session. It can be
13952emitted earlier if "option logasap" is specified, which makes sense in most
13953environments with long sessions such as remote terminals. Sessions which match
13954the "monitor" rules are never logged. It is also possible not to emit logs for
13955sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020013956specifying "option dontlognull" in the frontend. Successful connections will
13957not be logged if "option dontlog-normal" is specified in the frontend. A few
13958fields may slightly vary depending on some configuration options, those are
13959marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013960
13961 Example :
13962 frontend fnt
13963 mode tcp
13964 option tcplog
13965 log global
13966 default_backend bck
13967
13968 backend bck
13969 server srv1 127.0.0.1:8000
13970
13971 >>> Feb 6 12:12:56 localhost \
13972 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
13973 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
13974
13975 Field Format Extract from the example above
13976 1 process_name '[' pid ']:' haproxy[14387]:
13977 2 client_ip ':' client_port 10.0.1.2:33313
13978 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
13979 4 frontend_name fnt
13980 5 backend_name '/' server_name bck/srv1
13981 6 Tw '/' Tc '/' Tt* 0/0/5007
13982 7 bytes_read* 212
13983 8 termination_state --
13984 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
13985 10 srv_queue '/' backend_queue 0/0
13986
13987Detailed fields description :
13988 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010013989 connection to haproxy. If the connection was accepted on a UNIX socket
13990 instead, the IP address would be replaced with the word "unix". Note that
13991 when the connection is accepted on a socket configured with "accept-proxy"
13992 and the PROXY protocol is correctly used, then the logs will reflect the
13993 forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013994
13995 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010013996 If the connection was accepted on a UNIX socket instead, the port would be
13997 replaced with the ID of the accepting socket, which is also reported in the
13998 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013999
14000 - "accept_date" is the exact date when the connection was received by haproxy
14001 (which might be very slightly different from the date observed on the
14002 network if there was some queuing in the system's backlog). This is usually
14003 the same date which may appear in any upstream firewall's log.
14004
14005 - "frontend_name" is the name of the frontend (or listener) which received
14006 and processed the connection.
14007
14008 - "backend_name" is the name of the backend (or listener) which was selected
14009 to manage the connection to the server. This will be the same as the
14010 frontend if no switching rule has been applied, which is common for TCP
14011 applications.
14012
14013 - "server_name" is the name of the last server to which the connection was
14014 sent, which might differ from the first one if there were connection errors
14015 and a redispatch occurred. Note that this server belongs to the backend
14016 which processed the request. If the connection was aborted before reaching
14017 a server, "<NOSRV>" is indicated instead of a server name.
14018
14019 - "Tw" is the total time in milliseconds spent waiting in the various queues.
14020 It can be "-1" if the connection was aborted before reaching the queue.
14021 See "Timers" below for more details.
14022
14023 - "Tc" is the total time in milliseconds spent waiting for the connection to
14024 establish to the final server, including retries. It can be "-1" if the
14025 connection was aborted before a connection could be established. See
14026 "Timers" below for more details.
14027
14028 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014029 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014030 "option logasap" was specified, then the time counting stops at the moment
14031 the log is emitted. In this case, a '+' sign is prepended before the value,
14032 indicating that the final one will be larger. See "Timers" below for more
14033 details.
14034
14035 - "bytes_read" is the total number of bytes transmitted from the server to
14036 the client when the log is emitted. If "option logasap" is specified, the
14037 this value will be prefixed with a '+' sign indicating that the final one
14038 may be larger. Please note that this value is a 64-bit counter, so log
14039 analysis tools must be able to handle it without overflowing.
14040
14041 - "termination_state" is the condition the session was in when the session
14042 ended. This indicates the session state, which side caused the end of
14043 session to happen, and for what reason (timeout, error, ...). The normal
14044 flags should be "--", indicating the session was closed by either end with
14045 no data remaining in buffers. See below "Session state at disconnection"
14046 for more details.
14047
14048 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014049 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014050 limits have been reached. For instance, if actconn is close to 512 when
14051 multiple connection errors occur, chances are high that the system limits
14052 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014053 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014054
14055 - "feconn" is the total number of concurrent connections on the frontend when
14056 the session was logged. It is useful to estimate the amount of resource
14057 required to sustain high loads, and to detect when the frontend's "maxconn"
14058 has been reached. Most often when this value increases by huge jumps, it is
14059 because there is congestion on the backend servers, but sometimes it can be
14060 caused by a denial of service attack.
14061
14062 - "beconn" is the total number of concurrent connections handled by the
14063 backend when the session was logged. It includes the total number of
14064 concurrent connections active on servers as well as the number of
14065 connections pending in queues. It is useful to estimate the amount of
14066 additional servers needed to support high loads for a given application.
14067 Most often when this value increases by huge jumps, it is because there is
14068 congestion on the backend servers, but sometimes it can be caused by a
14069 denial of service attack.
14070
14071 - "srv_conn" is the total number of concurrent connections still active on
14072 the server when the session was logged. It can never exceed the server's
14073 configured "maxconn" parameter. If this value is very often close or equal
14074 to the server's "maxconn", it means that traffic regulation is involved a
14075 lot, meaning that either the server's maxconn value is too low, or that
14076 there aren't enough servers to process the load with an optimal response
14077 time. When only one of the server's "srv_conn" is high, it usually means
14078 that this server has some trouble causing the connections to take longer to
14079 be processed than on other servers.
14080
14081 - "retries" is the number of connection retries experienced by this session
14082 when trying to connect to the server. It must normally be zero, unless a
14083 server is being stopped at the same moment the connection was attempted.
14084 Frequent retries generally indicate either a network problem between
14085 haproxy and the server, or a misconfigured system backlog on the server
14086 preventing new connections from being queued. This field may optionally be
14087 prefixed with a '+' sign, indicating that the session has experienced a
14088 redispatch after the maximal retry count has been reached on the initial
14089 server. In this case, the server name appearing in the log is the one the
14090 connection was redispatched to, and not the first one, though both may
14091 sometimes be the same in case of hashing for instance. So as a general rule
14092 of thumb, when a '+' is present in front of the retry count, this count
14093 should not be attributed to the logged server.
14094
14095 - "srv_queue" is the total number of requests which were processed before
14096 this one in the server queue. It is zero when the request has not gone
14097 through the server queue. It makes it possible to estimate the approximate
14098 server's response time by dividing the time spent in queue by the number of
14099 requests in the queue. It is worth noting that if a session experiences a
14100 redispatch and passes through two server queues, their positions will be
14101 cumulated. A request should not pass through both the server queue and the
14102 backend queue unless a redispatch occurs.
14103
14104 - "backend_queue" is the total number of requests which were processed before
14105 this one in the backend's global queue. It is zero when the request has not
14106 gone through the global queue. It makes it possible to estimate the average
14107 queue length, which easily translates into a number of missing servers when
14108 divided by a server's "maxconn" parameter. It is worth noting that if a
14109 session experiences a redispatch, it may pass twice in the backend's queue,
14110 and then both positions will be cumulated. A request should not pass
14111 through both the server queue and the backend queue unless a redispatch
14112 occurs.
14113
14114
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200141158.2.3. HTTP log format
14116----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014117
14118The HTTP format is the most complete and the best suited for HTTP proxies. It
14119is enabled by when "option httplog" is specified in the frontend. It provides
14120the same level of information as the TCP format with additional features which
14121are specific to the HTTP protocol. Just like the TCP format, the log is usually
14122emitted at the end of the session, unless "option logasap" is specified, which
14123generally only makes sense for download sites. A session which matches the
14124"monitor" rules will never logged. It is also possible not to log sessions for
14125which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020014126frontend. Successful connections will not be logged if "option dontlog-normal"
14127is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014128
14129Most fields are shared with the TCP log, some being different. A few fields may
14130slightly vary depending on some configuration options. Those ones are marked
14131with a star ('*') after the field name below.
14132
14133 Example :
14134 frontend http-in
14135 mode http
14136 option httplog
14137 log global
14138 default_backend bck
14139
14140 backend static
14141 server srv1 127.0.0.1:8000
14142
14143 >>> Feb 6 12:14:14 localhost \
14144 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
14145 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 +010014146 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014147
14148 Field Format Extract from the example above
14149 1 process_name '[' pid ']:' haproxy[14389]:
14150 2 client_ip ':' client_port 10.0.1.2:33317
14151 3 '[' accept_date ']' [06/Feb/2009:12:14:14.655]
14152 4 frontend_name http-in
14153 5 backend_name '/' server_name static/srv1
14154 6 Tq '/' Tw '/' Tc '/' Tr '/' Tt* 10/0/30/69/109
14155 7 status_code 200
14156 8 bytes_read* 2750
14157 9 captured_request_cookie -
14158 10 captured_response_cookie -
14159 11 termination_state ----
14160 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
14161 13 srv_queue '/' backend_queue 0/0
14162 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
14163 15 '{' captured_response_headers* '}' {}
14164 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010014165
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014166
14167Detailed fields description :
14168 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014169 connection to haproxy. If the connection was accepted on a UNIX socket
14170 instead, the IP address would be replaced with the word "unix". Note that
14171 when the connection is accepted on a socket configured with "accept-proxy"
14172 and the PROXY protocol is correctly used, then the logs will reflect the
14173 forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014174
14175 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014176 If the connection was accepted on a UNIX socket instead, the port would be
14177 replaced with the ID of the accepting socket, which is also reported in the
14178 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014179
14180 - "accept_date" is the exact date when the TCP connection was received by
14181 haproxy (which might be very slightly different from the date observed on
14182 the network if there was some queuing in the system's backlog). This is
14183 usually the same date which may appear in any upstream firewall's log. This
14184 does not depend on the fact that the client has sent the request or not.
14185
14186 - "frontend_name" is the name of the frontend (or listener) which received
14187 and processed the connection.
14188
14189 - "backend_name" is the name of the backend (or listener) which was selected
14190 to manage the connection to the server. This will be the same as the
14191 frontend if no switching rule has been applied.
14192
14193 - "server_name" is the name of the last server to which the connection was
14194 sent, which might differ from the first one if there were connection errors
14195 and a redispatch occurred. Note that this server belongs to the backend
14196 which processed the request. If the request was aborted before reaching a
14197 server, "<NOSRV>" is indicated instead of a server name. If the request was
14198 intercepted by the stats subsystem, "<STATS>" is indicated instead.
14199
14200 - "Tq" is the total time in milliseconds spent waiting for the client to send
14201 a full HTTP request, not counting data. It can be "-1" if the connection
14202 was aborted before a complete request could be received. It should always
14203 be very small because a request generally fits in one single packet. Large
14204 times here generally indicate network trouble between the client and
14205 haproxy. See "Timers" below for more details.
14206
14207 - "Tw" is the total time in milliseconds spent waiting in the various queues.
14208 It can be "-1" if the connection was aborted before reaching the queue.
14209 See "Timers" below for more details.
14210
14211 - "Tc" is the total time in milliseconds spent waiting for the connection to
14212 establish to the final server, including retries. It can be "-1" if the
14213 request was aborted before a connection could be established. See "Timers"
14214 below for more details.
14215
14216 - "Tr" is the total time in milliseconds spent waiting for the server to send
14217 a full HTTP response, not counting data. It can be "-1" if the request was
14218 aborted before a complete response could be received. It generally matches
14219 the server's processing time for the request, though it may be altered by
14220 the amount of data sent by the client to the server. Large times here on
14221 "GET" requests generally indicate an overloaded server. See "Timers" below
14222 for more details.
14223
14224 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014225 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014226 "option logasap" was specified, then the time counting stops at the moment
14227 the log is emitted. In this case, a '+' sign is prepended before the value,
14228 indicating that the final one will be larger. See "Timers" below for more
14229 details.
14230
14231 - "status_code" is the HTTP status code returned to the client. This status
14232 is generally set by the server, but it might also be set by haproxy when
14233 the server cannot be reached or when its response is blocked by haproxy.
14234
14235 - "bytes_read" is the total number of bytes transmitted to the client when
14236 the log is emitted. This does include HTTP headers. If "option logasap" is
14237 specified, the this value will be prefixed with a '+' sign indicating that
14238 the final one may be larger. Please note that this value is a 64-bit
14239 counter, so log analysis tools must be able to handle it without
14240 overflowing.
14241
14242 - "captured_request_cookie" is an optional "name=value" entry indicating that
14243 the client had this cookie in the request. The cookie name and its maximum
14244 length are defined by the "capture cookie" statement in the frontend
14245 configuration. The field is a single dash ('-') when the option is not
14246 set. Only one cookie may be captured, it is generally used to track session
14247 ID exchanges between a client and a server to detect session crossing
14248 between clients due to application bugs. For more details, please consult
14249 the section "Capturing HTTP headers and cookies" below.
14250
14251 - "captured_response_cookie" is an optional "name=value" entry indicating
14252 that the server has returned a cookie with its response. The cookie name
14253 and its maximum length are defined by the "capture cookie" statement in the
14254 frontend configuration. The field is a single dash ('-') when the option is
14255 not set. Only one cookie may be captured, it is generally used to track
14256 session ID exchanges between a client and a server to detect session
14257 crossing between clients due to application bugs. For more details, please
14258 consult the section "Capturing HTTP headers and cookies" below.
14259
14260 - "termination_state" is the condition the session was in when the session
14261 ended. This indicates the session state, which side caused the end of
14262 session to happen, for what reason (timeout, error, ...), just like in TCP
14263 logs, and information about persistence operations on cookies in the last
14264 two characters. The normal flags should begin with "--", indicating the
14265 session was closed by either end with no data remaining in buffers. See
14266 below "Session state at disconnection" for more details.
14267
14268 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014269 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014270 limits have been reached. For instance, if actconn is close to 512 or 1024
14271 when multiple connection errors occur, chances are high that the system
14272 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014273 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014274 system.
14275
14276 - "feconn" is the total number of concurrent connections on the frontend when
14277 the session was logged. It is useful to estimate the amount of resource
14278 required to sustain high loads, and to detect when the frontend's "maxconn"
14279 has been reached. Most often when this value increases by huge jumps, it is
14280 because there is congestion on the backend servers, but sometimes it can be
14281 caused by a denial of service attack.
14282
14283 - "beconn" is the total number of concurrent connections handled by the
14284 backend when the session was logged. It includes the total number of
14285 concurrent connections active on servers as well as the number of
14286 connections pending in queues. It is useful to estimate the amount of
14287 additional servers needed to support high loads for a given application.
14288 Most often when this value increases by huge jumps, it is because there is
14289 congestion on the backend servers, but sometimes it can be caused by a
14290 denial of service attack.
14291
14292 - "srv_conn" is the total number of concurrent connections still active on
14293 the server when the session was logged. It can never exceed the server's
14294 configured "maxconn" parameter. If this value is very often close or equal
14295 to the server's "maxconn", it means that traffic regulation is involved a
14296 lot, meaning that either the server's maxconn value is too low, or that
14297 there aren't enough servers to process the load with an optimal response
14298 time. When only one of the server's "srv_conn" is high, it usually means
14299 that this server has some trouble causing the requests to take longer to be
14300 processed than on other servers.
14301
14302 - "retries" is the number of connection retries experienced by this session
14303 when trying to connect to the server. It must normally be zero, unless a
14304 server is being stopped at the same moment the connection was attempted.
14305 Frequent retries generally indicate either a network problem between
14306 haproxy and the server, or a misconfigured system backlog on the server
14307 preventing new connections from being queued. This field may optionally be
14308 prefixed with a '+' sign, indicating that the session has experienced a
14309 redispatch after the maximal retry count has been reached on the initial
14310 server. In this case, the server name appearing in the log is the one the
14311 connection was redispatched to, and not the first one, though both may
14312 sometimes be the same in case of hashing for instance. So as a general rule
14313 of thumb, when a '+' is present in front of the retry count, this count
14314 should not be attributed to the logged server.
14315
14316 - "srv_queue" is the total number of requests which were processed before
14317 this one in the server queue. It is zero when the request has not gone
14318 through the server queue. It makes it possible to estimate the approximate
14319 server's response time by dividing the time spent in queue by the number of
14320 requests in the queue. It is worth noting that if a session experiences a
14321 redispatch and passes through two server queues, their positions will be
14322 cumulated. A request should not pass through both the server queue and the
14323 backend queue unless a redispatch occurs.
14324
14325 - "backend_queue" is the total number of requests which were processed before
14326 this one in the backend's global queue. It is zero when the request has not
14327 gone through the global queue. It makes it possible to estimate the average
14328 queue length, which easily translates into a number of missing servers when
14329 divided by a server's "maxconn" parameter. It is worth noting that if a
14330 session experiences a redispatch, it may pass twice in the backend's queue,
14331 and then both positions will be cumulated. A request should not pass
14332 through both the server queue and the backend queue unless a redispatch
14333 occurs.
14334
14335 - "captured_request_headers" is a list of headers captured in the request due
14336 to the presence of the "capture request header" statement in the frontend.
14337 Multiple headers can be captured, they will be delimited by a vertical bar
14338 ('|'). When no capture is enabled, the braces do not appear, causing a
14339 shift of remaining fields. It is important to note that this field may
14340 contain spaces, and that using it requires a smarter log parser than when
14341 it's not used. Please consult the section "Capturing HTTP headers and
14342 cookies" below for more details.
14343
14344 - "captured_response_headers" is a list of headers captured in the response
14345 due to the presence of the "capture response header" statement in the
14346 frontend. Multiple headers can be captured, they will be delimited by a
14347 vertical bar ('|'). When no capture is enabled, the braces do not appear,
14348 causing a shift of remaining fields. It is important to note that this
14349 field may contain spaces, and that using it requires a smarter log parser
14350 than when it's not used. Please consult the section "Capturing HTTP headers
14351 and cookies" below for more details.
14352
14353 - "http_request" is the complete HTTP request line, including the method,
14354 request and HTTP version string. Non-printable characters are encoded (see
14355 below the section "Non-printable characters"). This is always the last
14356 field, and it is always delimited by quotes and is the only one which can
14357 contain quotes. If new fields are added to the log format, they will be
14358 added before this field. This field might be truncated if the request is
14359 huge and does not fit in the standard syslog buffer (1024 characters). This
14360 is the reason why this field must always remain the last one.
14361
14362
Cyril Bontédc4d9032012-04-08 21:57:39 +0200143638.2.4. Custom log format
14364------------------------
William Lallemand48940402012-01-30 16:47:22 +010014365
Willy Tarreau2beef582012-12-20 17:22:52 +010014366The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010014367mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010014368
14369HAproxy understands some log format variables. % precedes log format variables.
14370Variables can take arguments using braces ('{}'), and multiple arguments are
14371separated by commas within the braces. Flags may be added or removed by
14372prefixing them with a '+' or '-' sign.
14373
14374Special variable "%o" may be used to propagate its flags to all other
14375variables on the same format string. This is particularly handy with quoted
14376string formats ("Q").
14377
Willy Tarreauc8368452012-12-21 00:09:23 +010014378If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreaube722a22014-06-13 16:31:59 +020014379as a sample expression rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010014380less common information such as the client's SSL certificate's DN, or to log
14381the key that would be used to store an entry into a stick table.
14382
William Lallemand48940402012-01-30 16:47:22 +010014383Note: spaces must be escaped. A space character is considered as a separator.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014384In order to emit a verbatim '%', it must be preceded by another '%' resulting
Willy Tarreau06d97f92013-12-02 17:45:48 +010014385in '%%'. HAProxy will automatically merge consecutive separators.
William Lallemand48940402012-01-30 16:47:22 +010014386
14387Flags are :
14388 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014389 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
William Lallemand48940402012-01-30 16:47:22 +010014390
14391 Example:
14392
14393 log-format %T\ %t\ Some\ Text
14394 log-format %{+Q}o\ %t\ %s\ %{-Q}r
14395
14396At the moment, the default HTTP format is defined this way :
14397
Willy Tarreau2beef582012-12-20 17:22:52 +010014398 log-format %ci:%cp\ [%t]\ %ft\ %b/%s\ %Tq/%Tw/%Tc/%Tr/%Tt\ %ST\ %B\ %CC\ \
14399 %CS\ %tsc\ %ac/%fc/%bc/%sc/%rc\ %sq/%bq\ %hr\ %hs\ %{+Q}r
William Lallemand48940402012-01-30 16:47:22 +010014400
William Lallemandbddd4fd2012-02-27 11:23:10 +010014401the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010014402
Willy Tarreau2beef582012-12-20 17:22:52 +010014403 log-format %{+Q}o\ %{-Q}ci\ -\ -\ [%T]\ %r\ %ST\ %B\ \"\"\ \"\"\ %cp\ \
Willy Tarreau773d65f2012-10-12 14:56:11 +020014404 %ms\ %ft\ %b\ %s\ \%Tq\ %Tw\ %Tc\ %Tr\ %Tt\ %tsc\ %ac\ %fc\ \
Willy Tarreau2beef582012-12-20 17:22:52 +010014405 %bc\ %sc\ %rc\ %sq\ %bq\ %CC\ %CS\ \%hrl\ %hsl
William Lallemand48940402012-01-30 16:47:22 +010014406
William Lallemandbddd4fd2012-02-27 11:23:10 +010014407and the default TCP format is defined this way :
14408
Willy Tarreau2beef582012-12-20 17:22:52 +010014409 log-format %ci:%cp\ [%t]\ %ft\ %b/%s\ %Tw/%Tc/%Tt\ %B\ %ts\ \
William Lallemandbddd4fd2012-02-27 11:23:10 +010014410 %ac/%fc/%bc/%sc/%rc\ %sq/%bq
14411
William Lallemand48940402012-01-30 16:47:22 +010014412Please refer to the table below for currently defined variables :
14413
William Lallemandbddd4fd2012-02-27 11:23:10 +010014414 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014415 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014416 +---+------+-----------------------------------------------+-------------+
14417 | | %o | special variable, apply flags on all next var | |
14418 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010014419 | | %B | bytes_read (from server to client) | numeric |
14420 | H | %CC | captured_request_cookie | string |
14421 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020014422 | | %H | hostname | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000014423 | H | %HM | HTTP method (ex: POST) | string |
14424 | H | %HP | HTTP request URI without query string (path) | string |
Andrew Hayworthe63ac872015-07-31 16:14:16 +000014425 | H | %HQ | HTTP request URI query string (ex: ?bar=baz) | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000014426 | H | %HU | HTTP request URI (ex: /foo?bar=baz) | string |
14427 | H | %HV | HTTP version (ex: HTTP/1.0) | string |
William Lallemanda73203e2012-03-12 12:48:57 +010014428 | | %ID | unique-id | string |
Willy Tarreau4bf99632014-06-13 12:21:40 +020014429 | | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020014430 | | %T | gmt_date_time | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014431 | | %Tc | Tc | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080014432 | | %Tl | local_date_time | date |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014433 | H | %Tq | Tq | numeric |
14434 | H | %Tr | Tr | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020014435 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014436 | | %Tt | Tt | numeric |
14437 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010014438 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014439 | | %ac | actconn | numeric |
14440 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010014441 | | %bc | beconn (backend concurrent connections) | numeric |
14442 | | %bi | backend_source_ip (connecting address) | IP |
14443 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014444 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010014445 | | %ci | client_ip (accepted address) | IP |
14446 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014447 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010014448 | | %fc | feconn (frontend concurrent connections) | numeric |
14449 | | %fi | frontend_ip (accepting address) | IP |
14450 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020014451 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreau7346acb2014-08-28 15:03:15 +020014452 | | %lc | frontend_log_counter | numeric |
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020014453 | | %hr | captured_request_headers default style | string |
14454 | | %hrl | captured_request_headers CLF style | string list |
14455 | | %hs | captured_response_headers default style | string |
14456 | | %hsl | captured_response_headers CLF style | string list |
Willy Tarreau812c88e2015-08-09 10:56:35 +020014457 | | %ms | accept date milliseconds (left-padded with 0) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020014458 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014459 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014460 | | %rc | retries | numeric |
Willy Tarreau1f0da242014-01-25 11:01:50 +010014461 | | %rt | request_counter (HTTP req or TCP session) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014462 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010014463 | | %sc | srv_conn (server concurrent connections) | numeric |
14464 | | %si | server_IP (target address) | IP |
14465 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014466 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014467 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
14468 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010014469 | | %t | date_time (with millisecond resolution) | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014470 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014471 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014472 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010014473
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014474 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010014475
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010014476
144778.2.5. Error log format
14478-----------------------
14479
14480When an incoming connection fails due to an SSL handshake or an invalid PROXY
14481protocol header, haproxy will log the event using a shorter, fixed line format.
14482By default, logs are emitted at the LOG_INFO level, unless the option
14483"log-separate-errors" is set in the backend, in which case the LOG_ERR level
14484will be used. Connections on which no data are exchanged (eg: probes) are not
14485logged if the "dontlognull" option is set.
14486
14487The format looks like this :
14488
14489 >>> Dec 3 18:27:14 localhost \
14490 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
14491 Connection error during SSL handshake
14492
14493 Field Format Extract from the example above
14494 1 process_name '[' pid ']:' haproxy[6103]:
14495 2 client_ip ':' client_port 127.0.0.1:56059
14496 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
14497 4 frontend_name "/" bind_name ":" frt/f1:
14498 5 message Connection error during SSL handshake
14499
14500These fields just provide minimal information to help debugging connection
14501failures.
14502
14503
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200145048.3. Advanced logging options
14505-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014506
14507Some advanced logging options are often looked for but are not easy to find out
14508just by looking at the various options. Here is an entry point for the few
14509options which can enable better logging. Please refer to the keywords reference
14510for more information about their usage.
14511
14512
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200145138.3.1. Disabling logging of external tests
14514------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014515
14516It is quite common to have some monitoring tools perform health checks on
14517haproxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
14518commercial load-balancer, and sometimes it will simply be a more complete
14519monitoring system such as Nagios. When the tests are very frequent, users often
14520ask how to disable logging for those checks. There are three possibilities :
14521
14522 - if connections come from everywhere and are just TCP probes, it is often
14523 desired to simply disable logging of connections without data exchange, by
14524 setting "option dontlognull" in the frontend. It also disables logging of
14525 port scans, which may or may not be desired.
14526
14527 - if the connection come from a known source network, use "monitor-net" to
14528 declare this network as monitoring only. Any host in this network will then
14529 only be able to perform health checks, and their requests will not be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014530 logged. This is generally appropriate to designate a list of equipment
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014531 such as other load-balancers.
14532
14533 - if the tests are performed on a known URI, use "monitor-uri" to declare
14534 this URI as dedicated to monitoring. Any host sending this request will
14535 only get the result of a health-check, and the request will not be logged.
14536
14537
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200145388.3.2. Logging before waiting for the session to terminate
14539----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014540
14541The problem with logging at end of connection is that you have no clue about
14542what is happening during very long sessions, such as remote terminal sessions
14543or large file downloads. This problem can be worked around by specifying
14544"option logasap" in the frontend. Haproxy will then log as soon as possible,
14545just before data transfer begins. This means that in case of TCP, it will still
14546log the connection status to the server, and in case of HTTP, it will log just
14547after processing the server headers. In this case, the number of bytes reported
14548is the number of header bytes sent to the client. In order to avoid confusion
14549with normal logs, the total time field and the number of bytes are prefixed
14550with a '+' sign which means that real numbers are certainly larger.
14551
14552
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200145538.3.3. Raising log level upon errors
14554------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020014555
14556Sometimes it is more convenient to separate normal traffic from errors logs,
14557for instance in order to ease error monitoring from log files. When the option
14558"log-separate-errors" is used, connections which experience errors, timeouts,
14559retries, redispatches or HTTP status codes 5xx will see their syslog level
14560raised from "info" to "err". This will help a syslog daemon store the log in
14561a separate file. It is very important to keep the errors in the normal traffic
14562file too, so that log ordering is not altered. You should also be careful if
14563you already have configured your syslog daemon to store all logs higher than
14564"notice" in an "admin" file, because the "err" level is higher than "notice".
14565
14566
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200145678.3.4. Disabling logging of successful connections
14568--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020014569
14570Although this may sound strange at first, some large sites have to deal with
14571multiple thousands of logs per second and are experiencing difficulties keeping
14572them intact for a long time or detecting errors within them. If the option
14573"dontlog-normal" is set on the frontend, all normal connections will not be
14574logged. In this regard, a normal connection is defined as one without any
14575error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
14576and a response with a status 5xx is not considered normal and will be logged
14577too. Of course, doing is is really discouraged as it will remove most of the
14578useful information from the logs. Do this only if you have no other
14579alternative.
14580
14581
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200145828.4. Timing events
14583------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014584
14585Timers provide a great help in troubleshooting network problems. All values are
14586reported in milliseconds (ms). These timers should be used in conjunction with
14587the session termination flags. In TCP mode with "option tcplog" set on the
14588frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
14589mode, 5 control points are reported under the form "Tq/Tw/Tc/Tr/Tt" :
14590
14591 - Tq: total time to get the client request (HTTP mode only). It's the time
14592 elapsed between the moment the client connection was accepted and the
14593 moment the proxy received the last HTTP header. The value "-1" indicates
14594 that the end of headers (empty line) has never been seen. This happens when
14595 the client closes prematurely or times out.
14596
14597 - Tw: total time spent in the queues waiting for a connection slot. It
14598 accounts for backend queue as well as the server queues, and depends on the
14599 queue size, and the time needed for the server to complete previous
14600 requests. The value "-1" means that the request was killed before reaching
14601 the queue, which is generally what happens with invalid or denied requests.
14602
14603 - Tc: total time to establish the TCP connection to the server. It's the time
14604 elapsed between the moment the proxy sent the connection request, and the
14605 moment it was acknowledged by the server, or between the TCP SYN packet and
14606 the matching SYN/ACK packet in return. The value "-1" means that the
14607 connection never established.
14608
14609 - Tr: server response time (HTTP mode only). It's the time elapsed between
14610 the moment the TCP connection was established to the server and the moment
14611 the server sent its complete response headers. It purely shows its request
14612 processing time, without the network overhead due to the data transmission.
14613 It is worth noting that when the client has data to send to the server, for
14614 instance during a POST request, the time already runs, and this can distort
14615 apparent response time. For this reason, it's generally wise not to trust
14616 too much this field for POST requests initiated from clients behind an
14617 untrusted network. A value of "-1" here means that the last the response
14618 header (empty line) was never seen, most likely because the server timeout
14619 stroke before the server managed to process the request.
14620
14621 - Tt: total session duration time, between the moment the proxy accepted it
14622 and the moment both ends were closed. The exception is when the "logasap"
14623 option is specified. In this case, it only equals (Tq+Tw+Tc+Tr), and is
14624 prefixed with a '+' sign. From this field, we can deduce "Td", the data
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014625 transmission time, by subtracting other timers when valid :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014626
14627 Td = Tt - (Tq + Tw + Tc + Tr)
14628
14629 Timers with "-1" values have to be excluded from this equation. In TCP
14630 mode, "Tq" and "Tr" have to be excluded too. Note that "Tt" can never be
14631 negative.
14632
14633These timers provide precious indications on trouble causes. Since the TCP
14634protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
14635that timers close to multiples of 3s are nearly always related to lost packets
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010014636due to network problems (wires, negotiation, congestion). Moreover, if "Tt" is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014637close to a timeout value specified in the configuration, it often means that a
14638session has been aborted on timeout.
14639
14640Most common cases :
14641
14642 - If "Tq" is close to 3000, a packet has probably been lost between the
14643 client and the proxy. This is very rare on local networks but might happen
14644 when clients are on far remote networks and send large requests. It may
14645 happen that values larger than usual appear here without any network cause.
14646 Sometimes, during an attack or just after a resource starvation has ended,
14647 haproxy may accept thousands of connections in a few milliseconds. The time
14648 spent accepting these connections will inevitably slightly delay processing
14649 of other connections, and it can happen that request times in the order of
14650 a few tens of milliseconds are measured after a few thousands of new
Patrick Mezard105faca2010-06-12 17:02:46 +020014651 connections have been accepted at once. Setting "option http-server-close"
14652 may display larger request times since "Tq" also measures the time spent
14653 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014654
14655 - If "Tc" is close to 3000, a packet has probably been lost between the
14656 server and the proxy during the server connection phase. This value should
14657 always be very low, such as 1 ms on local networks and less than a few tens
14658 of ms on remote networks.
14659
Willy Tarreau55165fe2009-05-10 12:02:55 +020014660 - If "Tr" is nearly always lower than 3000 except some rare values which seem
14661 to be the average majored by 3000, there are probably some packets lost
14662 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014663
14664 - If "Tt" is large even for small byte counts, it generally is because
14665 neither the client nor the server decides to close the connection, for
14666 instance because both have agreed on a keep-alive connection mode. In order
14667 to solve this issue, it will be needed to specify "option httpclose" on
14668 either the frontend or the backend. If the problem persists, it means that
14669 the server ignores the "close" connection mode and expects the client to
14670 close. Then it will be required to use "option forceclose". Having the
14671 smallest possible 'Tt' is important when connection regulation is used with
14672 the "maxconn" option on the servers, since no new connection will be sent
14673 to the server until another one is released.
14674
14675Other noticeable HTTP log cases ('xx' means any value to be ignored) :
14676
14677 Tq/Tw/Tc/Tr/+Tt The "option logasap" is present on the frontend and the log
14678 was emitted before the data phase. All the timers are valid
14679 except "Tt" which is shorter than reality.
14680
14681 -1/xx/xx/xx/Tt The client was not able to send a complete request in time
14682 or it aborted too early. Check the session termination flags
14683 then "timeout http-request" and "timeout client" settings.
14684
14685 Tq/-1/xx/xx/Tt It was not possible to process the request, maybe because
14686 servers were out of order, because the request was invalid
14687 or forbidden by ACL rules. Check the session termination
14688 flags.
14689
14690 Tq/Tw/-1/xx/Tt The connection could not establish on the server. Either it
14691 actively refused it or it timed out after Tt-(Tq+Tw) ms.
14692 Check the session termination flags, then check the
14693 "timeout connect" setting. Note that the tarpit action might
14694 return similar-looking patterns, with "Tw" equal to the time
14695 the client connection was maintained open.
14696
14697 Tq/Tw/Tc/-1/Tt The server has accepted the connection but did not return
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014698 a complete response in time, or it closed its connection
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014699 unexpectedly after Tt-(Tq+Tw+Tc) ms. Check the session
14700 termination flags, then check the "timeout server" setting.
14701
14702
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200147038.5. Session state at disconnection
14704-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014705
14706TCP and HTTP logs provide a session termination indicator in the
14707"termination_state" field, just before the number of active connections. It is
147082-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
14709each of which has a special meaning :
14710
14711 - On the first character, a code reporting the first event which caused the
14712 session to terminate :
14713
14714 C : the TCP session was unexpectedly aborted by the client.
14715
14716 S : the TCP session was unexpectedly aborted by the server, or the
14717 server explicitly refused it.
14718
14719 P : the session was prematurely aborted by the proxy, because of a
14720 connection limit enforcement, because a DENY filter was matched,
14721 because of a security check which detected and blocked a dangerous
14722 error in server response which might have caused information leak
Willy Tarreau570f2212013-06-10 16:42:09 +020014723 (eg: cacheable cookie).
14724
14725 L : the session was locally processed by haproxy and was not passed to
14726 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014727
14728 R : a resource on the proxy has been exhausted (memory, sockets, source
14729 ports, ...). Usually, this appears during the connection phase, and
14730 system logs should contain a copy of the precise error. If this
14731 happens, it must be considered as a very serious anomaly which
14732 should be fixed as soon as possible by any means.
14733
14734 I : an internal error was identified by the proxy during a self-check.
14735 This should NEVER happen, and you are encouraged to report any log
14736 containing this, because this would almost certainly be a bug. It
14737 would be wise to preventively restart the process after such an
14738 event too, in case it would be caused by memory corruption.
14739
Simon Horman752dc4a2011-06-21 14:34:59 +090014740 D : the session was killed by haproxy because the server was detected
14741 as down and was configured to kill all connections when going down.
14742
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070014743 U : the session was killed by haproxy on this backup server because an
14744 active server was detected as up and was configured to kill all
14745 backup connections when going up.
14746
Willy Tarreaua2a64e92011-09-07 23:01:56 +020014747 K : the session was actively killed by an admin operating on haproxy.
14748
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014749 c : the client-side timeout expired while waiting for the client to
14750 send or receive data.
14751
14752 s : the server-side timeout expired while waiting for the server to
14753 send or receive data.
14754
14755 - : normal session completion, both the client and the server closed
14756 with nothing left in the buffers.
14757
14758 - on the second character, the TCP or HTTP session state when it was closed :
14759
Willy Tarreauf7b30a92010-12-06 22:59:17 +010014760 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014761 (HTTP mode only). Nothing was sent to any server.
14762
14763 Q : the proxy was waiting in the QUEUE for a connection slot. This can
14764 only happen when servers have a 'maxconn' parameter set. It can
14765 also happen in the global queue after a redispatch consecutive to
14766 a failed attempt to connect to a dying server. If no redispatch is
14767 reported, then no connection attempt was made to any server.
14768
14769 C : the proxy was waiting for the CONNECTION to establish on the
14770 server. The server might at most have noticed a connection attempt.
14771
14772 H : the proxy was waiting for complete, valid response HEADERS from the
14773 server (HTTP only).
14774
14775 D : the session was in the DATA phase.
14776
14777 L : the proxy was still transmitting LAST data to the client while the
14778 server had already finished. This one is very rare as it can only
14779 happen when the client dies while receiving the last packets.
14780
14781 T : the request was tarpitted. It has been held open with the client
14782 during the whole "timeout tarpit" duration or until the client
14783 closed, both of which will be reported in the "Tw" timer.
14784
14785 - : normal session completion after end of data transfer.
14786
14787 - the third character tells whether the persistence cookie was provided by
14788 the client (only in HTTP mode) :
14789
14790 N : the client provided NO cookie. This is usually the case for new
14791 visitors, so counting the number of occurrences of this flag in the
14792 logs generally indicate a valid trend for the site frequentation.
14793
14794 I : the client provided an INVALID cookie matching no known server.
14795 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020014796 cookies between HTTP/HTTPS sites, persistence conditionally
14797 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014798
14799 D : the client provided a cookie designating a server which was DOWN,
14800 so either "option persist" was used and the client was sent to
14801 this server, or it was not set and the client was redispatched to
14802 another server.
14803
Willy Tarreau996a92c2010-10-13 19:30:47 +020014804 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014805 server.
14806
Willy Tarreau996a92c2010-10-13 19:30:47 +020014807 E : the client provided a valid cookie, but with a last date which was
14808 older than what is allowed by the "maxidle" cookie parameter, so
14809 the cookie is consider EXPIRED and is ignored. The request will be
14810 redispatched just as if there was no cookie.
14811
14812 O : the client provided a valid cookie, but with a first date which was
14813 older than what is allowed by the "maxlife" cookie parameter, so
14814 the cookie is consider too OLD and is ignored. The request will be
14815 redispatched just as if there was no cookie.
14816
Willy Tarreauc89ccb62012-04-05 21:18:22 +020014817 U : a cookie was present but was not used to select the server because
14818 some other server selection mechanism was used instead (typically a
14819 "use-server" rule).
14820
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014821 - : does not apply (no cookie set in configuration).
14822
14823 - the last character reports what operations were performed on the persistence
14824 cookie returned by the server (only in HTTP mode) :
14825
14826 N : NO cookie was provided by the server, and none was inserted either.
14827
14828 I : no cookie was provided by the server, and the proxy INSERTED one.
14829 Note that in "cookie insert" mode, if the server provides a cookie,
14830 it will still be overwritten and reported as "I" here.
14831
Willy Tarreau996a92c2010-10-13 19:30:47 +020014832 U : the proxy UPDATED the last date in the cookie that was presented by
14833 the client. This can only happen in insert mode with "maxidle". It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014834 happens every time there is activity at a different date than the
Willy Tarreau996a92c2010-10-13 19:30:47 +020014835 date indicated in the cookie. If any other change happens, such as
14836 a redispatch, then the cookie will be marked as inserted instead.
14837
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014838 P : a cookie was PROVIDED by the server and transmitted as-is.
14839
14840 R : the cookie provided by the server was REWRITTEN by the proxy, which
14841 happens in "cookie rewrite" or "cookie prefix" modes.
14842
14843 D : the cookie provided by the server was DELETED by the proxy.
14844
14845 - : does not apply (no cookie set in configuration).
14846
Willy Tarreau996a92c2010-10-13 19:30:47 +020014847The combination of the two first flags gives a lot of information about what
14848was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014849helpful to detect server saturation, network troubles, local system resource
14850starvation, attacks, etc...
14851
14852The most common termination flags combinations are indicated below. They are
14853alphabetically sorted, with the lowercase set just after the upper case for
14854easier finding and understanding.
14855
14856 Flags Reason
14857
14858 -- Normal termination.
14859
14860 CC The client aborted before the connection could be established to the
14861 server. This can happen when haproxy tries to connect to a recently
14862 dead (or unchecked) server, and the client aborts while haproxy is
14863 waiting for the server to respond or for "timeout connect" to expire.
14864
14865 CD The client unexpectedly aborted during data transfer. This can be
14866 caused by a browser crash, by an intermediate equipment between the
14867 client and haproxy which decided to actively break the connection,
14868 by network routing issues between the client and haproxy, or by a
14869 keep-alive session between the server and the client terminated first
14870 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010014871
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014872 cD The client did not send nor acknowledge any data for as long as the
14873 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020014874 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014875
14876 CH The client aborted while waiting for the server to start responding.
14877 It might be the server taking too long to respond or the client
14878 clicking the 'Stop' button too fast.
14879
14880 cH The "timeout client" stroke while waiting for client data during a
14881 POST request. This is sometimes caused by too large TCP MSS values
14882 for PPPoE networks which cannot transport full-sized packets. It can
14883 also happen when client timeout is smaller than server timeout and
14884 the server takes too long to respond.
14885
14886 CQ The client aborted while its session was queued, waiting for a server
14887 with enough empty slots to accept it. It might be that either all the
14888 servers were saturated or that the assigned server was taking too
14889 long a time to respond.
14890
14891 CR The client aborted before sending a full HTTP request. Most likely
14892 the request was typed by hand using a telnet client, and aborted
14893 too early. The HTTP status code is likely a 400 here. Sometimes this
14894 might also be caused by an IDS killing the connection between haproxy
Willy Tarreau0f228a02015-05-01 15:37:53 +020014895 and the client. "option http-ignore-probes" can be used to ignore
14896 connections without any data transfer.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014897
14898 cR The "timeout http-request" stroke before the client sent a full HTTP
14899 request. This is sometimes caused by too large TCP MSS values on the
14900 client side for PPPoE networks which cannot transport full-sized
14901 packets, or by clients sending requests by hand and not typing fast
14902 enough, or forgetting to enter the empty line at the end of the
Willy Tarreau2705a612014-05-23 17:38:34 +020014903 request. The HTTP status code is likely a 408 here. Note: recently,
Willy Tarreau0f228a02015-05-01 15:37:53 +020014904 some browsers started to implement a "pre-connect" feature consisting
14905 in speculatively connecting to some recently visited web sites just
14906 in case the user would like to visit them. This results in many
14907 connections being established to web sites, which end up in 408
14908 Request Timeout if the timeout strikes first, or 400 Bad Request when
14909 the browser decides to close them first. These ones pollute the log
14910 and feed the error counters. Some versions of some browsers have even
14911 been reported to display the error code. It is possible to work
14912 around the undesirable effects of this behaviour by adding "option
14913 http-ignore-probes" in the frontend, resulting in connections with
14914 zero data transfer to be totally ignored. This will definitely hide
14915 the errors of people experiencing connectivity issues though.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014916
14917 CT The client aborted while its session was tarpitted. It is important to
14918 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020014919 wrong tarpit rules have been written. If a lot of them happen, it
14920 might make sense to lower the "timeout tarpit" value to something
14921 closer to the average reported "Tw" timer, in order not to consume
14922 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014923
Willy Tarreau570f2212013-06-10 16:42:09 +020014924 LR The request was intercepted and locally handled by haproxy. Generally
14925 it means that this was a redirect or a stats request.
14926
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010014927 SC The server or an equipment between it and haproxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014928 the TCP connection (the proxy received a TCP RST or an ICMP message
14929 in return). Under some circumstances, it can also be the network
14930 stack telling the proxy that the server is unreachable (eg: no route,
14931 or no ARP response on local network). When this happens in HTTP mode,
14932 the status code is likely a 502 or 503 here.
14933
14934 sC The "timeout connect" stroke before a connection to the server could
14935 complete. When this happens in HTTP mode, the status code is likely a
14936 503 or 504 here.
14937
14938 SD The connection to the server died with an error during the data
14939 transfer. This usually means that haproxy has received an RST from
14940 the server or an ICMP message from an intermediate equipment while
14941 exchanging data with the server. This can be caused by a server crash
14942 or by a network issue on an intermediate equipment.
14943
14944 sD The server did not send nor acknowledge any data for as long as the
14945 "timeout server" setting during the data phase. This is often caused
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010014946 by too short timeouts on L4 equipments before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014947 load-balancers, ...), as well as keep-alive sessions maintained
14948 between the client and the server expiring first on haproxy.
14949
14950 SH The server aborted before sending its full HTTP response headers, or
14951 it crashed while processing the request. Since a server aborting at
14952 this moment is very rare, it would be wise to inspect its logs to
14953 control whether it crashed and why. The logged request may indicate a
14954 small set of faulty requests, demonstrating bugs in the application.
14955 Sometimes this might also be caused by an IDS killing the connection
14956 between haproxy and the server.
14957
14958 sH The "timeout server" stroke before the server could return its
14959 response headers. This is the most common anomaly, indicating too
14960 long transactions, probably caused by server or database saturation.
14961 The immediate workaround consists in increasing the "timeout server"
14962 setting, but it is important to keep in mind that the user experience
14963 will suffer from these long response times. The only long term
14964 solution is to fix the application.
14965
14966 sQ The session spent too much time in queue and has been expired. See
14967 the "timeout queue" and "timeout connect" settings to find out how to
14968 fix this if it happens too often. If it often happens massively in
14969 short periods, it may indicate general problems on the affected
14970 servers due to I/O or database congestion, or saturation caused by
14971 external attacks.
14972
14973 PC The proxy refused to establish a connection to the server because the
14974 process' socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020014975 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014976 so that it does not happen anymore. This status is very rare and
14977 might happen when the global "ulimit-n" parameter is forced by hand.
14978
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010014979 PD The proxy blocked an incorrectly formatted chunked encoded message in
14980 a request or a response, after the server has emitted its headers. In
14981 most cases, this will indicate an invalid message from the server to
Willy Tarreauf3a3e132013-08-31 08:16:26 +020014982 the client. Haproxy supports chunk sizes of up to 2GB - 1 (2147483647
14983 bytes). Any larger size will be considered as an error.
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010014984
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014985 PH The proxy blocked the server's response, because it was invalid,
14986 incomplete, dangerous (cache control), or matched a security filter.
14987 In any case, an HTTP 502 error is sent to the client. One possible
14988 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010014989 containing unauthorized characters. It is also possible but quite
14990 rare, that the proxy blocked a chunked-encoding request from the
14991 client due to an invalid syntax, before the server responded. In this
14992 case, an HTTP 400 error is sent to the client and reported in the
14993 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014994
14995 PR The proxy blocked the client's HTTP request, either because of an
14996 invalid HTTP syntax, in which case it returned an HTTP 400 error to
14997 the client, or because a deny filter matched, in which case it
14998 returned an HTTP 403 error.
14999
15000 PT The proxy blocked the client's request and has tarpitted its
15001 connection before returning it a 500 server error. Nothing was sent
15002 to the server. The connection was maintained open for as long as
15003 reported by the "Tw" timer field.
15004
15005 RC A local resource has been exhausted (memory, sockets, source ports)
15006 preventing the connection to the server from establishing. The error
15007 logs will tell precisely what was missing. This is very rare and can
15008 only be solved by proper system tuning.
15009
Willy Tarreau996a92c2010-10-13 19:30:47 +020015010The combination of the two last flags gives a lot of information about how
15011persistence was handled by the client, the server and by haproxy. This is very
15012important to troubleshoot disconnections, when users complain they have to
15013re-authenticate. The commonly encountered flags are :
15014
15015 -- Persistence cookie is not enabled.
15016
15017 NN No cookie was provided by the client, none was inserted in the
15018 response. For instance, this can be in insert mode with "postonly"
15019 set on a GET request.
15020
15021 II A cookie designating an invalid server was provided by the client,
15022 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015023 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020015024 value can be presented by a client when no other server knows it.
15025
15026 NI No cookie was provided by the client, one was inserted in the
15027 response. This typically happens for first requests from every user
15028 in "insert" mode, which makes it an easy way to count real users.
15029
15030 VN A cookie was provided by the client, none was inserted in the
15031 response. This happens for most responses for which the client has
15032 already got a cookie.
15033
15034 VU A cookie was provided by the client, with a last visit date which is
15035 not completely up-to-date, so an updated cookie was provided in
15036 response. This can also happen if there was no date at all, or if
15037 there was a date but the "maxidle" parameter was not set, so that the
15038 cookie can be switched to unlimited time.
15039
15040 EI A cookie was provided by the client, with a last visit date which is
15041 too old for the "maxidle" parameter, so the cookie was ignored and a
15042 new cookie was inserted in the response.
15043
15044 OI A cookie was provided by the client, with a first visit date which is
15045 too old for the "maxlife" parameter, so the cookie was ignored and a
15046 new cookie was inserted in the response.
15047
15048 DI The server designated by the cookie was down, a new server was
15049 selected and a new cookie was emitted in the response.
15050
15051 VI The server designated by the cookie was not marked dead but could not
15052 be reached. A redispatch happened and selected another one, which was
15053 then advertised in the response.
15054
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015055
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200150568.6. Non-printable characters
15057-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015058
15059In order not to cause trouble to log analysis tools or terminals during log
15060consulting, non-printable characters are not sent as-is into log files, but are
15061converted to the two-digits hexadecimal representation of their ASCII code,
15062prefixed by the character '#'. The only characters that can be logged without
15063being escaped are comprised between 32 and 126 (inclusive). Obviously, the
15064escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
15065is the same for the character '"' which becomes "#22", as well as '{', '|' and
15066'}' when logging headers.
15067
15068Note that the space character (' ') is not encoded in headers, which can cause
15069issues for tools relying on space count to locate fields. A typical header
15070containing spaces is "User-Agent".
15071
15072Last, it has been observed that some syslog daemons such as syslog-ng escape
15073the quote ('"') with a backslash ('\'). The reverse operation can safely be
15074performed since no quote may appear anywhere else in the logs.
15075
15076
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200150778.7. Capturing HTTP cookies
15078---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015079
15080Cookie capture simplifies the tracking a complete user session. This can be
15081achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015082section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015083cookie will simultaneously be checked in the request ("Cookie:" header) and in
15084the response ("Set-Cookie:" header). The respective values will be reported in
15085the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015086locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015087not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
15088user switches to a new session for example, because the server will reassign it
15089a new cookie. It is also possible to detect if a server unexpectedly sets a
15090wrong cookie to a client, leading to session crossing.
15091
15092 Examples :
15093 # capture the first cookie whose name starts with "ASPSESSION"
15094 capture cookie ASPSESSION len 32
15095
15096 # capture the first cookie whose name is exactly "vgnvisitor"
15097 capture cookie vgnvisitor= len 32
15098
15099
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200151008.8. Capturing HTTP headers
15101---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015102
15103Header captures are useful to track unique request identifiers set by an upper
15104proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
15105the response, one can search for information about the response length, how the
15106server asked the cache to behave, or an object location during a redirection.
15107
15108Header captures are performed using the "capture request header" and "capture
15109response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015110section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015111
15112It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010015113time. Non-existent headers are logged as empty strings, and if one header
15114appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015115are grouped within braces '{' and '}' in the same order as they were declared,
15116and delimited with a vertical bar '|' without any space. Response headers
15117follow the same representation, but are displayed after a space following the
15118request headers block. These blocks are displayed just before the HTTP request
15119in the logs.
15120
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020015121As a special case, it is possible to specify an HTTP header capture in a TCP
15122frontend. The purpose is to enable logging of headers which will be parsed in
15123an HTTP backend if the request is then switched to this HTTP backend.
15124
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015125 Example :
15126 # This instance chains to the outgoing proxy
15127 listen proxy-out
15128 mode http
15129 option httplog
15130 option logasap
15131 log global
15132 server cache1 192.168.1.1:3128
15133
15134 # log the name of the virtual server
15135 capture request header Host len 20
15136
15137 # log the amount of data uploaded during a POST
15138 capture request header Content-Length len 10
15139
15140 # log the beginning of the referrer
15141 capture request header Referer len 20
15142
15143 # server name (useful for outgoing proxies only)
15144 capture response header Server len 20
15145
15146 # logging the content-length is useful with "option logasap"
15147 capture response header Content-Length len 10
15148
15149 # log the expected cache behaviour on the response
15150 capture response header Cache-Control len 8
15151
15152 # the Via header will report the next proxy's name
15153 capture response header Via len 20
15154
15155 # log the URL location during a redirection
15156 capture response header Location len 20
15157
15158 >>> Aug 9 20:26:09 localhost \
15159 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
15160 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
15161 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
15162 "GET http://fr.adserver.yahoo.com/"
15163
15164 >>> Aug 9 20:30:46 localhost \
15165 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
15166 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
15167 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010015168 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015169
15170 >>> Aug 9 20:30:46 localhost \
15171 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
15172 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
15173 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
15174 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010015175 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015176
15177
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200151788.9. Examples of logs
15179---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015180
15181These are real-world examples of logs accompanied with an explanation. Some of
15182them have been made up by hand. The syslog part has been removed for better
15183reading. Their sole purpose is to explain how to decipher them.
15184
15185 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
15186 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
15187 "HEAD / HTTP/1.0"
15188
15189 => long request (6.5s) entered by hand through 'telnet'. The server replied
15190 in 147 ms, and the session ended normally ('----')
15191
15192 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
15193 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
15194 0/9 "HEAD / HTTP/1.0"
15195
15196 => Idem, but the request was queued in the global queue behind 9 other
15197 requests, and waited there for 1230 ms.
15198
15199 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
15200 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
15201 "GET /image.iso HTTP/1.0"
15202
15203 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010015204 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015205 14 ms, 243 bytes of headers were sent to the client, and total time from
15206 accept to first data byte is 30 ms.
15207
15208 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
15209 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
15210 "GET /cgi-bin/bug.cgi? HTTP/1.0"
15211
15212 => the proxy blocked a server response either because of an "rspdeny" or
15213 "rspideny" filter, or because the response was improperly formatted and
Willy Tarreau3c92c5f2011-08-28 09:45:47 +020015214 not HTTP-compliant, or because it blocked sensitive information which
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015215 risked being cached. In this case, the response is replaced with a "502
15216 bad gateway". The flags ("PH--") tell us that it was haproxy who decided
15217 to return the 502 and not the server.
15218
15219 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010015220 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 +010015221
15222 => the client never completed its request and aborted itself ("C---") after
15223 8.5s, while the proxy was waiting for the request headers ("-R--").
15224 Nothing was sent to any server.
15225
15226 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
15227 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
15228
15229 => The client never completed its request, which was aborted by the
15230 time-out ("c---") after 50s, while the proxy was waiting for the request
15231 headers ("-R--"). Nothing was sent to any server, but the proxy could
15232 send a 408 return code to the client.
15233
15234 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
15235 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
15236
15237 => This log was produced with "option tcplog". The client timed out after
15238 5 seconds ("c----").
15239
15240 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
15241 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010015242 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015243
15244 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015245 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015246 (config says 'retries 3'), and no redispatch (otherwise we would have
15247 seen "/+3"). Status code 503 was returned to the client. There were 115
15248 connections on this server, 202 connections on this proxy, and 205 on
15249 the global process. It is possible that the server refused the
15250 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010015251
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010015252
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200152539. Statistics and monitoring
15254----------------------------
15255
15256It is possible to query HAProxy about its status. The most commonly used
15257mechanism is the HTTP statistics page. This page also exposes an alternative
15258CSV output format for monitoring tools. The same format is provided on the
15259Unix socket.
15260
15261
152629.1. CSV format
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010015263---------------
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +010015264
Willy Tarreau7f062c42009-03-05 18:43:00 +010015265The statistics may be consulted either from the unix socket or from the HTTP
Willy Tarreaua3310dc2014-06-16 15:43:21 +020015266page. Both means provide a CSV format whose fields follow. The first line
15267begins with a sharp ('#') and has one word per comma-delimited field which
15268represents the title of the column. All other lines starting at the second one
15269use a classical CSV format using a comma as the delimiter, and the double quote
15270('"') as an optional text delimiter, but only if the enclosed text is ambiguous
15271(if it contains a quote or a comma). The double-quote character ('"') in the
15272text is doubled ('""'), which is the format that most tools recognize. Please
15273do not insert any column before these ones in order not to break tools which
15274use hard-coded column positions.
Willy Tarreau7f062c42009-03-05 18:43:00 +010015275
James Westbyebe62d62014-07-08 10:14:57 -040015276In brackets after each field name are the types which may have a value for
15277that field. The types are L (Listeners), F (Frontends), B (Backends), and
15278S (Servers).
15279
15280 0. pxname [LFBS]: proxy name
15281 1. svname [LFBS]: service name (FRONTEND for frontend, BACKEND for backend,
15282 any name for server/listener)
15283 2. qcur [..BS]: current queued requests. For the backend this reports the
15284 number queued without a server assigned.
15285 3. qmax [..BS]: max value of qcur
15286 4. scur [LFBS]: current sessions
15287 5. smax [LFBS]: max sessions
15288 6. slim [LFBS]: configured session limit
15289 7. stot [LFBS]: cumulative number of connections
15290 8. bin [LFBS]: bytes in
15291 9. bout [LFBS]: bytes out
15292 10. dreq [LFB.]: requests denied because of security concerns.
15293 - For tcp this is because of a matched tcp-request content rule.
15294 - For http this is because of a matched http-request or tarpit rule.
15295 11. dresp [LFBS]: responses denied because of security concerns.
15296 - For http this is because of a matched http-request rule, or
15297 "option checkcache".
15298 12. ereq [LF..]: request errors. Some of the possible causes are:
15299 - early termination from the client, before the request has been sent.
15300 - read error from the client
15301 - client timeout
15302 - client closed connection
15303 - various bad requests from the client.
15304 - request was tarpitted.
15305 13. econ [..BS]: number of requests that encountered an error trying to
15306 connect to a backend server. The backend stat is the sum of the stat
15307 for all servers of that backend, plus any connection errors not
15308 associated with a particular server (such as the backend having no
15309 active servers).
15310 14. eresp [..BS]: response errors. srv_abrt will be counted here also.
15311 Some other errors are:
15312 - write error on the client socket (won't be counted for the server stat)
15313 - failure applying filters to the response.
15314 15. wretr [..BS]: number of times a connection to a server was retried.
15315 16. wredis [..BS]: number of times a request was redispatched to another
15316 server. The server value counts the number of times that server was
15317 switched away from.
15318 17. status [LFBS]: status (UP/DOWN/NOLB/MAINT/MAINT(via)...)
Pavlos Parissis1f673c72015-05-02 20:30:44 +020015319 18. weight [..BS]: total weight (backend), server weight (server)
15320 19. act [..BS]: number of active servers (backend), server is active (server)
15321 20. bck [..BS]: number of backup servers (backend), server is backup (server)
James Westbyebe62d62014-07-08 10:14:57 -040015322 21. chkfail [...S]: number of failed checks. (Only counts checks failed when
15323 the server is up.)
15324 22. chkdown [..BS]: number of UP->DOWN transitions. The backend counter counts
15325 transitions to the whole backend being down, rather than the sum of the
15326 counters for each server.
15327 23. lastchg [..BS]: number of seconds since the last UP<->DOWN transition
15328 24. downtime [..BS]: total downtime (in seconds). The value for the backend
15329 is the downtime for the whole backend, not the sum of the server downtime.
15330 25. qlimit [...S]: configured maxqueue for the server, or nothing in the
15331 value is 0 (default, meaning no limit)
15332 26. pid [LFBS]: process id (0 for first instance, 1 for second, ...)
15333 27. iid [LFBS]: unique proxy id
15334 28. sid [L..S]: server id (unique inside a proxy)
15335 29. throttle [...S]: current throttle percentage for the server, when
15336 slowstart is active, or no value if not in slowstart.
15337 30. lbtot [..BS]: total number of times a server was selected, either for new
15338 sessions, or when re-dispatching. The server counter is the number
15339 of times that server was selected.
15340 31. tracked [...S]: id of proxy/server if tracking is enabled.
15341 32. type [LFBS]: (0=frontend, 1=backend, 2=server, 3=socket/listener)
15342 33. rate [.FBS]: number of sessions per second over last elapsed second
15343 34. rate_lim [.F..]: configured limit on new sessions per second
15344 35. rate_max [.FBS]: max number of new sessions per second
15345 36. check_status [...S]: status of last health check, one of:
Cyril Bontéf0c60612010-02-06 14:44:47 +010015346 UNK -> unknown
15347 INI -> initializing
15348 SOCKERR -> socket error
15349 L4OK -> check passed on layer 4, no upper layers testing enabled
Jason Harvey83104802015-04-16 11:13:21 -080015350 L4TOUT -> layer 1-4 timeout
Cyril Bontéf0c60612010-02-06 14:44:47 +010015351 L4CON -> layer 1-4 connection problem, for example
15352 "Connection refused" (tcp rst) or "No route to host" (icmp)
15353 L6OK -> check passed on layer 6
15354 L6TOUT -> layer 6 (SSL) timeout
15355 L6RSP -> layer 6 invalid response - protocol error
15356 L7OK -> check passed on layer 7
15357 L7OKC -> check conditionally passed on layer 7, for example 404 with
15358 disable-on-404
15359 L7TOUT -> layer 7 (HTTP/SMTP) timeout
15360 L7RSP -> layer 7 invalid response - protocol error
15361 L7STS -> layer 7 response error, for example HTTP 5xx
James Westbyebe62d62014-07-08 10:14:57 -040015362 37. check_code [...S]: layer5-7 code, if available
15363 38. check_duration [...S]: time in ms took to finish last health check
15364 39. hrsp_1xx [.FBS]: http responses with 1xx code
15365 40. hrsp_2xx [.FBS]: http responses with 2xx code
15366 41. hrsp_3xx [.FBS]: http responses with 3xx code
15367 42. hrsp_4xx [.FBS]: http responses with 4xx code
15368 43. hrsp_5xx [.FBS]: http responses with 5xx code
15369 44. hrsp_other [.FBS]: http responses with other codes (protocol error)
15370 45. hanafail [...S]: failed health checks details
15371 46. req_rate [.F..]: HTTP requests per second over last elapsed second
15372 47. req_rate_max [.F..]: max number of HTTP requests per second observed
15373 48. req_tot [.F..]: total number of HTTP requests received
15374 49. cli_abrt [..BS]: number of data transfers aborted by the client
15375 50. srv_abrt [..BS]: number of data transfers aborted by the server
15376 (inc. in eresp)
15377 51. comp_in [.FB.]: number of HTTP response bytes fed to the compressor
15378 52. comp_out [.FB.]: number of HTTP response bytes emitted by the compressor
15379 53. comp_byp [.FB.]: number of bytes that bypassed the HTTP compressor
15380 (CPU/BW limit)
15381 54. comp_rsp [.FB.]: number of HTTP responses that were compressed
15382 55. lastsess [..BS]: number of seconds since last session assigned to
15383 server/backend
15384 56. last_chk [...S]: last health check contents or textual error
15385 57. last_agt [...S]: last agent check contents or textual error
15386 58. qtime [..BS]: the average queue time in ms over the 1024 last requests
15387 59. ctime [..BS]: the average connect time in ms over the 1024 last requests
15388 60. rtime [..BS]: the average response time in ms over the 1024 last requests
15389 (0 for TCP)
15390 61. ttime [..BS]: the average total session time in ms over the 1024 last
15391 requests
Willy Tarreau844e3c52008-01-11 16:28:18 +010015392
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010015393
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200153949.2. Unix Socket commands
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010015395-------------------------
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010015396
Willy Tarreau468f4932013-08-01 16:50:16 +020015397The stats socket is not enabled by default. In order to enable it, it is
15398necessary to add one line in the global section of the haproxy configuration.
15399A second line is recommended to set a larger timeout, always appreciated when
15400issuing commands by hand :
15401
15402 global
15403 stats socket /var/run/haproxy.sock mode 600 level admin
15404 stats timeout 2m
15405
15406It is also possible to add multiple instances of the stats socket by repeating
15407the line, and make them listen to a TCP port instead of a UNIX socket. This is
15408never done by default because this is dangerous, but can be handy in some
15409situations :
15410
15411 global
15412 stats socket /var/run/haproxy.sock mode 600 level admin
15413 stats socket ipv4@192.168.0.1:9999 level admin
15414 stats timeout 2m
15415
Cyril Bonté307ee1e2015-09-28 23:16:06 +020015416To access the socket, an external utility such as "socat" is required. Socat is
15417a swiss-army knife to connect anything to anything. We use it to connect
15418terminals to the socket, or a couple of stdin/stdout pipes to it for scripts.
15419The two main syntaxes we'll use are the following :
Willy Tarreau468f4932013-08-01 16:50:16 +020015420
15421 # socat /var/run/haproxy.sock stdio
15422 # socat /var/run/haproxy.sock readline
15423
15424The first one is used with scripts. It is possible to send the output of a
15425script to haproxy, and pass haproxy's output to another script. That's useful
15426for retrieving counters or attack traces for example.
15427
15428The second one is only useful for issuing commands by hand. It has the benefit
15429that the terminal is handled by the readline library which supports line
15430editing and history, which is very convenient when issuing repeated commands
15431(eg: watch a counter).
15432
15433The socket supports two operation modes :
15434 - interactive
15435 - non-interactive
15436
15437The non-interactive mode is the default when socat connects to the socket. In
15438this mode, a single line may be sent. It is processed as a whole, responses are
15439sent back, and the connection closes after the end of the response. This is the
15440mode that scripts and monitoring tools use. It is possible to send multiple
15441commands in this mode, they need to be delimited by a semi-colon (';'). For
15442example :
15443
15444 # echo "show info;show stat;show table" | socat /var/run/haproxy stdio
15445
15446The interactive mode displays a prompt ('>') and waits for commands to be
15447entered on the line, then processes them, and displays the prompt again to wait
15448for a new command. This mode is entered via the "prompt" command which must be
15449sent on the first line in non-interactive mode. The mode is a flip switch, if
15450"prompt" is sent in interactive mode, it is disabled and the connection closes
15451after processing the last command of the same line.
15452
15453For this reason, when debugging by hand, it's quite common to start with the
15454"prompt" command :
15455
15456 # socat /var/run/haproxy readline
15457 prompt
15458 > show info
15459 ...
15460 >
15461
15462Since multiple commands may be issued at once, haproxy uses the empty line as a
15463delimiter to mark an end of output for each command, and takes care of ensuring
15464that no command can emit an empty line on output. A script can thus easily
15465parse the output even when multiple commands were pipelined on a single line.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010015466
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020015467It is important to understand that when multiple haproxy processes are started
15468on the same sockets, any process may pick up the request and will output its
15469own stats.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010015470
Willy Tarreau468f4932013-08-01 16:50:16 +020015471The list of commands currently supported on the stats socket is provided below.
15472If an unknown command is sent, haproxy displays the usage message which reminds
15473all supported commands. Some commands support a more complex syntax, generally
15474it will explain what part of the command is invalid when this happens.
15475
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010015476add acl <acl> <pattern>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010015477 Add an entry into the acl <acl>. <acl> is the #<id> or the <file> returned by
15478 "show acl". This command does not verify if the entry already exists. This
15479 command cannot be used if the reference <acl> is a file also used with a map.
15480 In this case, you must use the command "add map" in place of "add acl".
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010015481
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010015482add map <map> <key> <value>
15483 Add an entry into the map <map> to associate the value <value> to the key
15484 <key>. This command does not verify if the entry already exists. It is
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010015485 mainly used to fill a map after a clear operation. Note that if the reference
15486 <map> is a file and is shared with a map, this map will contain also a new
15487 pattern entry.
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010015488
Willy Tarreaud63335a2010-02-26 12:56:52 +010015489clear counters
15490 Clear the max values of the statistics counters in each proxy (frontend &
15491 backend) and in each server. The cumulated counters are not affected. This
15492 can be used to get clean counters after an incident, without having to
15493 restart nor to clear traffic counters. This command is restricted and can
15494 only be issued on sockets configured for levels "operator" or "admin".
15495
15496clear counters all
15497 Clear all statistics counters in each proxy (frontend & backend) and in each
15498 server. This has the same effect as restarting. This command is restricted
15499 and can only be issued on sockets configured for level "admin".
15500
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010015501clear acl <acl>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010015502 Remove all entries from the acl <acl>. <acl> is the #<id> or the <file>
15503 returned by "show acl". Note that if the reference <acl> is a file and is
15504 shared with a map, this map will be also cleared.
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010015505
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010015506clear map <map>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010015507 Remove all entries from the map <map>. <map> is the #<id> or the <file>
15508 returned by "show map". Note that if the reference <map> is a file and is
15509 shared with a acl, this acl will be also cleared.
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010015510
Simon Hormanc88b8872011-06-15 15:18:49 +090015511clear table <table> [ data.<type> <operator> <value> ] | [ key <key> ]
15512 Remove entries from the stick-table <table>.
15513
15514 This is typically used to unblock some users complaining they have been
15515 abusively denied access to a service, but this can also be used to clear some
15516 stickiness entries matching a server that is going to be replaced (see "show
15517 table" below for details). Note that sometimes, removal of an entry will be
15518 refused because it is currently tracked by a session. Retrying a few seconds
15519 later after the session ends is usual enough.
15520
15521 In the case where no options arguments are given all entries will be removed.
15522
15523 When the "data." form is used entries matching a filter applied using the
15524 stored data (see "stick-table" in section 4.2) are removed. A stored data
15525 type must be specified in <type>, and this data type must be stored in the
15526 table otherwise an error is reported. The data is compared according to
15527 <operator> with the 64-bit integer <value>. Operators are the same as with
15528 the ACLs :
15529
15530 - eq : match entries whose data is equal to this value
15531 - ne : match entries whose data is not equal to this value
15532 - le : match entries whose data is less than or equal to this value
15533 - ge : match entries whose data is greater than or equal to this value
15534 - lt : match entries whose data is less than this value
15535 - gt : match entries whose data is greater than this value
15536
15537 When the key form is used the entry <key> is removed. The key must be of the
Simon Horman619e3cc2011-06-15 15:18:52 +090015538 same type as the table, which currently is limited to IPv4, IPv6, integer and
15539 string.
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020015540
15541 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020015542 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Emeric Brun7c6b82e2010-09-24 16:34:28 +020015543 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020015544 >>> 0x80e6a4c: key=127.0.0.1 use=0 exp=3594729 gpc0=0 conn_rate(30000)=1 \
15545 bytes_out_rate(60000)=187
15546 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
15547 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020015548
15549 $ echo "clear table http_proxy key 127.0.0.1" | socat stdio /tmp/sock1
15550
15551 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Emeric Brun7c6b82e2010-09-24 16:34:28 +020015552 >>> # table: http_proxy, type: ip, size:204800, used:1
Willy Tarreau62a36c42010-08-17 15:53:10 +020015553 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
15554 bytes_out_rate(60000)=191
Simon Hormanc88b8872011-06-15 15:18:49 +090015555 $ echo "clear table http_proxy data.gpc0 eq 1" | socat stdio /tmp/sock1
15556 $ echo "show table http_proxy" | socat stdio /tmp/sock1
15557 >>> # table: http_proxy, type: ip, size:204800, used:1
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020015558
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010015559del acl <acl> [<key>|#<ref>]
15560 Delete all the acl entries from the acl <acl> corresponding to the key <key>.
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010015561 <acl> is the #<id> or the <file> returned by "show acl". If the <ref> is used,
15562 this command delete only the listed reference. The reference can be found with
15563 listing the content of the acl. Note that if the reference <acl> is a file and
15564 is shared with a map, the entry will be also deleted in the map.
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010015565
15566del map <map> [<key>|#<ref>]
Thierry FOURNIER0b90f312014-01-29 20:40:18 +010015567 Delete all the map entries from the map <map> corresponding to the key <key>.
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010015568 <map> is the #<id> or the <file> returned by "show map". If the <ref> is used,
15569 this command delete only the listed reference. The reference can be found with
15570 listing the content of the map. Note that if the reference <map> is a file and
15571 is shared with a acl, the entry will be also deleted in the map.
Thierry FOURNIER0b90f312014-01-29 20:40:18 +010015572
15573disable agent <backend>/<server>
Simon Horman671b6f02013-11-25 10:46:39 +090015574 Mark the auxiliary agent check as temporarily stopped.
15575
15576 In the case where an agent check is being run as a auxiliary check, due
15577 to the agent-check parameter of a server directive, new checks are only
15578 initialised when the agent is in the enabled. Thus, disable agent will
15579 prevent any new agent checks from begin initiated until the agent
15580 re-enabled using enable agent.
15581
15582 When an agent is disabled the processing of an auxiliary agent check that
15583 was initiated while the agent was set as enabled is as follows: All
15584 results that would alter the weight, specifically "drain" or a weight
15585 returned by the agent, are ignored. The processing of agent check is
15586 otherwise unchanged.
15587
15588 The motivation for this feature is to allow the weight changing effects
15589 of the agent checks to be paused to allow the weight of a server to be
15590 configured using set weight without being overridden by the agent.
15591
15592 This command is restricted and can only be issued on sockets configured for
15593 level "admin".
15594
Willy Tarreau532a4502011-09-07 22:37:44 +020015595disable frontend <frontend>
15596 Mark the frontend as temporarily stopped. This corresponds to the mode which
15597 is used during a soft restart : the frontend releases the port but can be
15598 enabled again if needed. This should be used with care as some non-Linux OSes
15599 are unable to enable it back. This is intended to be used in environments
15600 where stopping a proxy is not even imaginable but a misconfigured proxy must
15601 be fixed. That way it's possible to release the port and bind it into another
15602 process to restore operations. The frontend will appear with status "STOP"
15603 on the stats page.
15604
15605 The frontend may be specified either by its name or by its numeric ID,
15606 prefixed with a sharp ('#').
15607
15608 This command is restricted and can only be issued on sockets configured for
15609 level "admin".
15610
Willy Tarreau9b5aecd2014-05-23 11:53:10 +020015611disable health <backend>/<server>
15612 Mark the primary health check as temporarily stopped. This will disable
15613 sending of health checks, and the last health check result will be ignored.
15614 The server will be in unchecked state and considered UP unless an auxiliary
15615 agent check forces it down.
15616
15617 This command is restricted and can only be issued on sockets configured for
15618 level "admin".
15619
Willy Tarreaud63335a2010-02-26 12:56:52 +010015620disable server <backend>/<server>
15621 Mark the server DOWN for maintenance. In this mode, no more checks will be
15622 performed on the server until it leaves maintenance.
15623 If the server is tracked by other servers, those servers will be set to DOWN
15624 during the maintenance.
15625
15626 In the statistics page, a server DOWN for maintenance will appear with a
15627 "MAINT" status, its tracking servers with the "MAINT(via)" one.
15628
15629 Both the backend and the server may be specified either by their name or by
Willy Tarreauf5f31922011-08-02 11:32:07 +020015630 their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010015631
15632 This command is restricted and can only be issued on sockets configured for
15633 level "admin".
15634
Simon Horman671b6f02013-11-25 10:46:39 +090015635enable agent <backend>/<server>
15636 Resume auxiliary agent check that was temporarily stopped.
15637
15638 See "disable agent" for details of the effect of temporarily starting
15639 and stopping an auxiliary agent.
15640
15641 This command is restricted and can only be issued on sockets configured for
15642 level "admin".
15643
Willy Tarreau532a4502011-09-07 22:37:44 +020015644enable frontend <frontend>
15645 Resume a frontend which was temporarily stopped. It is possible that some of
15646 the listening ports won't be able to bind anymore (eg: if another process
15647 took them since the 'disable frontend' operation). If this happens, an error
15648 is displayed. Some operating systems might not be able to resume a frontend
15649 which was disabled.
15650
15651 The frontend may be specified either by its name or by its numeric ID,
15652 prefixed with a sharp ('#').
15653
15654 This command is restricted and can only be issued on sockets configured for
15655 level "admin".
15656
Willy Tarreau9b5aecd2014-05-23 11:53:10 +020015657enable health <backend>/<server>
15658 Resume a primary health check that was temporarily stopped. This will enable
15659 sending of health checks again. Please see "disable health" for details.
15660
15661 This command is restricted and can only be issued on sockets configured for
15662 level "admin".
15663
Willy Tarreaud63335a2010-02-26 12:56:52 +010015664enable server <backend>/<server>
15665 If the server was previously marked as DOWN for maintenance, this marks the
15666 server UP and checks are re-enabled.
15667
15668 Both the backend and the server may be specified either by their name or by
Willy Tarreauf5f31922011-08-02 11:32:07 +020015669 their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010015670
15671 This command is restricted and can only be issued on sockets configured for
15672 level "admin".
15673
Thierry FOURNIER5b16df72014-02-26 18:07:38 +010015674get map <map> <value>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010015675get acl <acl> <value>
15676 Lookup the value <value> in the map <map> or in the ACL <acl>. <map> or <acl>
15677 are the #<id> or the <file> returned by "show map" or "show acl". This command
15678 returns all the matching patterns associated with this map. This is useful for
15679 debugging maps and ACLs. The output format is composed by one line par
15680 matching type. Each line is composed by space-delimited series of words.
Thierry FOURNIER5b16df72014-02-26 18:07:38 +010015681
15682 The first two words are:
15683
15684 <match method>: The match method applied. It can be "found", "bool",
15685 "int", "ip", "bin", "len", "str", "beg", "sub", "dir",
15686 "dom", "end" or "reg".
15687
15688 <match result>: The result. Can be "match" or "no-match".
15689
15690 The following words are returned only if the pattern matches an entry.
15691
15692 <index type>: "tree" or "list". The internal lookup algorithm.
15693
15694 <case>: "case-insensitive" or "case-sensitive". The
15695 interpretation of the case.
15696
15697 <entry matched>: match="<entry>". Return the matched pattern. It is
15698 useful with regular expressions.
15699
15700 The two last word are used to show the returned value and its type. With the
15701 "acl" case, the pattern doesn't exist.
15702
15703 return=nothing: No return because there are no "map".
15704 return="<value>": The value returned in the string format.
15705 return=cannot-display: The value cannot be converted as string.
15706
15707 type="<type>": The type of the returned sample.
15708
Willy Tarreaud63335a2010-02-26 12:56:52 +010015709get weight <backend>/<server>
15710 Report the current weight and the initial weight of server <server> in
15711 backend <backend> or an error if either doesn't exist. The initial weight is
15712 the one that appears in the configuration file. Both are normally equal
15713 unless the current weight has been changed. Both the backend and the server
15714 may be specified either by their name or by their numeric ID, prefixed with a
Willy Tarreauf5f31922011-08-02 11:32:07 +020015715 sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010015716
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020015717help
15718 Print the list of known keywords and their basic usage. The same help screen
15719 is also displayed for unknown commands.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010015720
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020015721prompt
15722 Toggle the prompt at the beginning of the line and enter or leave interactive
15723 mode. In interactive mode, the connection is not closed after a command
15724 completes. Instead, the prompt will appear again, indicating the user that
15725 the interpreter is waiting for a new command. The prompt consists in a right
15726 angle bracket followed by a space "> ". This mode is particularly convenient
15727 when one wants to periodically check information such as stats or errors.
15728 It is also a good idea to enter interactive mode before issuing a "help"
15729 command.
15730
15731quit
15732 Close the connection when in interactive mode.
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010015733
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010015734set map <map> [<key>|#<ref>] <value>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010015735 Modify the value corresponding to each key <key> in a map <map>. <map> is the
15736 #<id> or <file> returned by "show map". If the <ref> is used in place of
15737 <key>, only the entry pointed by <ref> is changed. The new value is <value>.
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010015738
Willy Tarreau2a0f4d22011-08-02 11:49:05 +020015739set maxconn frontend <frontend> <value>
Willy Tarreau3c7a79d2012-09-26 21:07:15 +020015740 Dynamically change the specified frontend's maxconn setting. Any positive
15741 value is allowed including zero, but setting values larger than the global
15742 maxconn does not make much sense. If the limit is increased and connections
15743 were pending, they will immediately be accepted. If it is lowered to a value
15744 below the current number of connections, new connections acceptation will be
Willy Tarreau2a0f4d22011-08-02 11:49:05 +020015745 delayed until the threshold is reached. The frontend might be specified by
15746 either its name or its numeric ID prefixed with a sharp ('#').
15747
Willy Tarreau91886b62011-09-07 14:38:31 +020015748set maxconn global <maxconn>
15749 Dynamically change the global maxconn setting within the range defined by the
15750 initial global maxconn setting. If it is increased and connections were
15751 pending, they will immediately be accepted. If it is lowered to a value below
15752 the current number of connections, new connections acceptation will be
15753 delayed until the threshold is reached. A value of zero restores the initial
15754 setting.
15755
Willy Tarreauf5b22872011-09-07 16:13:44 +020015756set rate-limit connections global <value>
15757 Change the process-wide connection rate limit, which is set by the global
15758 'maxconnrate' setting. A value of zero disables the limitation. This limit
15759 applies to all frontends and the change has an immediate effect. The value
15760 is passed in number of connections per second.
15761
William Lallemandd85f9172012-11-09 17:05:39 +010015762set rate-limit http-compression global <value>
15763 Change the maximum input compression rate, which is set by the global
15764 'maxcomprate' setting. A value of zero disables the limitation. The value is
William Lallemand096f5542012-11-19 17:26:05 +010015765 passed in number of kilobytes per second. The value is available in the "show
15766 info" on the line "CompressBpsRateLim" in bytes.
William Lallemandd85f9172012-11-09 17:05:39 +010015767
Willy Tarreau93e7c002013-10-07 18:51:07 +020015768set rate-limit sessions global <value>
15769 Change the process-wide session rate limit, which is set by the global
15770 'maxsessrate' setting. A value of zero disables the limitation. This limit
15771 applies to all frontends and the change has an immediate effect. The value
15772 is passed in number of sessions per second.
15773
Willy Tarreaue43d5322013-10-07 20:01:52 +020015774set rate-limit ssl-sessions global <value>
15775 Change the process-wide SSL session rate limit, which is set by the global
15776 'maxsslrate' setting. A value of zero disables the limitation. This limit
15777 applies to all frontends and the change has an immediate effect. The value
15778 is passed in number of sessions per second sent to the SSL stack. It applies
15779 before the handshake in order to protect the stack against handshake abuses.
15780
Baptiste Assmann3d8f8312015-04-13 22:54:33 +020015781set server <backend>/<server> addr <ip4 or ip6 address>
15782 Replace the current IP address of a server by the one provided.
15783
Willy Tarreau2a4b70f2014-05-22 18:42:35 +020015784set server <backend>/<server> agent [ up | down ]
15785 Force a server's agent to a new state. This can be useful to immediately
15786 switch a server's state regardless of some slow agent checks for example.
15787 Note that the change is propagated to tracking servers if any.
15788
15789set server <backend>/<server> health [ up | stopping | down ]
15790 Force a server's health to a new state. This can be useful to immediately
15791 switch a server's state regardless of some slow health checks for example.
15792 Note that the change is propagated to tracking servers if any.
15793
15794set server <backend>/<server> state [ ready | drain | maint ]
15795 Force a server's administrative state to a new state. This can be useful to
15796 disable load balancing and/or any traffic to a server. Setting the state to
15797 "ready" puts the server in normal mode, and the command is the equivalent of
15798 the "enable server" command. Setting the state to "maint" disables any traffic
15799 to the server as well as any health checks. This is the equivalent of the
15800 "disable server" command. Setting the mode to "drain" only removes the server
15801 from load balancing but still allows it to be checked and to accept new
15802 persistent connections. Changes are propagated to tracking servers if any.
15803
15804set server <backend>/<server> weight <weight>[%]
15805 Change a server's weight to the value passed in argument. This is the exact
15806 equivalent of the "set weight" command below.
15807
Emeric Brun4147b2e2014-06-16 18:36:30 +020015808set ssl ocsp-response <response>
15809 This command is used to update an OCSP Response for a certificate (see "crt"
15810 on "bind" lines). Same controls are performed as during the initial loading of
15811 the response. The <response> must be passed as a base64 encoded string of the
15812 DER encoded response from the OCSP server.
15813
15814 Example:
15815 openssl ocsp -issuer issuer.pem -cert server.pem \
15816 -host ocsp.issuer.com:80 -respout resp.der
15817 echo "set ssl ocsp-response $(base64 -w 10000 resp.der)" | \
15818 socat stdio /var/run/haproxy.stat
15819
Nenad Merdanovicc6985f02015-05-09 08:46:02 +020015820set ssl tls-key <id> <tlskey>
15821 Set the next TLS key for the <id> listener to <tlskey>. This key becomes the
15822 ultimate key, while the penultimate one is used for encryption (others just
15823 decrypt). The oldest TLS key present is overwritten. <id> is either a numeric
15824 #<id> or <file> returned by "show tls-keys". <tlskey> is a base64 encoded 48
15825 bit TLS ticket key (ex. openssl rand -base64 48).
15826
Willy Tarreau47060b62013-08-01 21:11:42 +020015827set table <table> key <key> [data.<data_type> <value>]*
Willy Tarreau654694e2012-06-07 01:03:16 +020015828 Create or update a stick-table entry in the table. If the key is not present,
15829 an entry is inserted. See stick-table in section 4.2 to find all possible
15830 values for <data_type>. The most likely use consists in dynamically entering
15831 entries for source IP addresses, with a flag in gpc0 to dynamically block an
Willy Tarreau47060b62013-08-01 21:11:42 +020015832 IP address or affect its quality of service. It is possible to pass multiple
15833 data_types in a single call.
Willy Tarreau654694e2012-06-07 01:03:16 +020015834
Willy Tarreaud63335a2010-02-26 12:56:52 +010015835set timeout cli <delay>
15836 Change the CLI interface timeout for current connection. This can be useful
15837 during long debugging sessions where the user needs to constantly inspect
15838 some indicators without being disconnected. The delay is passed in seconds.
15839
15840set weight <backend>/<server> <weight>[%]
15841 Change a server's weight to the value passed in argument. If the value ends
15842 with the '%' sign, then the new weight will be relative to the initially
Simon Horman58b5d292013-02-12 10:45:52 +090015843 configured weight. Absolute weights are permitted between 0 and 256.
15844 Relative weights must be positive with the resulting absolute weight is
15845 capped at 256. Servers which are part of a farm running a static
15846 load-balancing algorithm have stricter limitations because the weight
15847 cannot change once set. Thus for these servers, the only accepted values
15848 are 0 and 100% (or 0 and the initial weight). Changes take effect
15849 immediately, though certain LB algorithms require a certain amount of
15850 requests to consider changes. A typical usage of this command is to
15851 disable a server during an update by setting its weight to zero, then to
15852 enable it again after the update by setting it back to 100%. This command
15853 is restricted and can only be issued on sockets configured for level
15854 "admin". Both the backend and the server may be specified either by their
15855 name or by their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010015856
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010015857show errors [<iid>]
15858 Dump last known request and response errors collected by frontends and
15859 backends. If <iid> is specified, the limit the dump to errors concerning
Willy Tarreau6162db22009-10-10 17:13:00 +020015860 either frontend or backend whose ID is <iid>. This command is restricted
15861 and can only be issued on sockets configured for levels "operator" or
15862 "admin".
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010015863
15864 The errors which may be collected are the last request and response errors
15865 caused by protocol violations, often due to invalid characters in header
15866 names. The report precisely indicates what exact character violated the
15867 protocol. Other important information such as the exact date the error was
15868 detected, frontend and backend names, the server name (when known), the
15869 internal session ID and the source address which has initiated the session
15870 are reported too.
15871
15872 All characters are returned, and non-printable characters are encoded. The
15873 most common ones (\t = 9, \n = 10, \r = 13 and \e = 27) are encoded as one
15874 letter following a backslash. The backslash itself is encoded as '\\' to
15875 avoid confusion. Other non-printable characters are encoded '\xNN' where
15876 NN is the two-digits hexadecimal representation of the character's ASCII
15877 code.
15878
15879 Lines are prefixed with the position of their first character, starting at 0
15880 for the beginning of the buffer. At most one input line is printed per line,
15881 and large lines will be broken into multiple consecutive output lines so that
15882 the output never goes beyond 79 characters wide. It is easy to detect if a
15883 line was broken, because it will not end with '\n' and the next line's offset
15884 will be followed by a '+' sign, indicating it is a continuation of previous
15885 line.
15886
15887 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020015888 $ echo "show errors" | socat stdio /tmp/sock1
15889 >>> [04/Mar/2009:15:46:56.081] backend http-in (#2) : invalid response
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010015890 src 127.0.0.1, session #54, frontend fe-eth0 (#1), server s2 (#1)
15891 response length 213 bytes, error at position 23:
15892
15893 00000 HTTP/1.0 200 OK\r\n
15894 00017 header/bizarre:blah\r\n
15895 00038 Location: blah\r\n
15896 00054 Long-line: this is a very long line which should b
15897 00104+ e broken into multiple lines on the output buffer,
15898 00154+ otherwise it would be too large to print in a ter
15899 00204+ minal\r\n
15900 00211 \r\n
15901
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015902 In the example above, we see that the backend "http-in" which has internal
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010015903 ID 2 has blocked an invalid response from its server s2 which has internal
15904 ID 1. The request was on session 54 initiated by source 127.0.0.1 and
15905 received by frontend fe-eth0 whose ID is 1. The total response length was
15906 213 bytes when the error was detected, and the error was at byte 23. This
15907 is the slash ('/') in header name "header/bizarre", which is not a valid
15908 HTTP character for a header name.
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010015909
Baptiste Assmann9b6857e2015-09-18 14:49:12 +020015910show backend
15911 Dump the list of backends available in the running process
15912
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020015913show info
15914 Dump info about haproxy status on current process.
15915
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010015916show map [<map>]
15917 Dump info about map converters. Without argument, the list of all available
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010015918 maps is returned. If a <map> is specified, its contents are dumped. <map> is
15919 the #<id> or <file>. The first column is a unique identifier. It can be used
15920 as reference for the operation "del map" and "set map". The second column is
15921 the pattern and the third column is the sample if available. The data returned
15922 are not directly a list of available maps, but are the list of all patterns
15923 composing any map. Many of these patterns can be shared with ACL.
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010015924
15925show acl [<acl>]
15926 Dump info about acl converters. Without argument, the list of all available
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010015927 acls is returned. If a <acl> is specified, its contents are dumped. <acl> if
15928 the #<id> or <file>. The dump format is the same than the map even for the
15929 sample value. The data returned are not a list of available ACL, but are the
15930 list of all patterns composing any ACL. Many of these patterns can be shared
15931 with maps.
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010015932
Willy Tarreau12833bb2014-01-28 16:49:56 +010015933show pools
15934 Dump the status of internal memory pools. This is useful to track memory
15935 usage when suspecting a memory leak for example. It does exactly the same
15936 as the SIGQUIT when running in foreground except that it does not flush
15937 the pools.
15938
Baptiste Assmann6f9225a2015-05-08 19:35:08 +020015939show servers state [<backend>]
15940 Dump the state of the servers found in the running configuration. A backend
15941 name or identifier may be provided to limit the output to this backend only.
15942
15943 The dump has the following format:
15944 - first line contains the format version (1 in this specification);
15945 - second line contains the column headers, prefixed by a sharp ('#');
15946 - third line and next ones contain data;
15947 - each line starting by a sharp ('#') is considered as a comment.
15948
15949 Since multiple versions of the ouptput may co-exist, below is the list of
15950 fields and their order per file format version :
15951 1:
15952 be_id: Backend unique id.
15953 be_name: Backend label.
15954 srv_id: Server unique id (in the backend).
15955 srv_name: Server label.
15956 srv_addr: Server IP address.
15957 srv_op_state: Server operational state (UP/DOWN/...).
15958 In source code: SRV_ST_*.
15959 srv_admin_state: Server administrative state (MAINT/DRAIN/...).
15960 In source code: SRV_ADMF_*.
15961 srv_uweight: User visible server's weight.
15962 srv_iweight: Server's initial weight.
15963 srv_time_since_last_change: Time since last operational change.
15964 srv_check_status: Last health check status.
15965 srv_check_result: Last check result (FAILED/PASSED/...).
15966 In source code: CHK_RES_*.
15967 srv_check_health: Checks rise / fall current counter.
15968 srv_check_state: State of the check (ENABLED/PAUSED/...).
15969 In source code: CHK_ST_*.
15970 srv_agent_state: State of the agent check (ENABLED/PAUSED/...).
15971 In source code: CHK_ST_*.
15972 bk_f_forced_id: Flag to know if the backend ID is forced by
15973 configuration.
15974 srv_f_forced_id: Flag to know if the server's ID is forced by
15975 configuration.
15976
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020015977show sess
15978 Dump all known sessions. Avoid doing this on slow connections as this can
Willy Tarreau6162db22009-10-10 17:13:00 +020015979 be huge. This command is restricted and can only be issued on sockets
15980 configured for levels "operator" or "admin".
15981
Willy Tarreau66dc20a2010-03-05 17:53:32 +010015982show sess <id>
15983 Display a lot of internal information about the specified session identifier.
15984 This identifier is the first field at the beginning of the lines in the dumps
15985 of "show sess" (it corresponds to the session pointer). Those information are
15986 useless to most users but may be used by haproxy developers to troubleshoot a
15987 complex bug. The output format is intentionally not documented so that it can
Olivierce31e6e2014-09-05 18:49:10 +020015988 freely evolve depending on demands. You may find a description of all fields
15989 returned in src/dumpstats.c
15990
15991 The special id "all" dumps the states of all sessions, which must be avoided
15992 as much as possible as it is highly CPU intensive and can take a lot of time.
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020015993
15994show stat [<iid> <type> <sid>]
15995 Dump statistics in the CSV format. By passing <id>, <type> and <sid>, it is
15996 possible to dump only selected items :
15997 - <iid> is a proxy ID, -1 to dump everything
15998 - <type> selects the type of dumpable objects : 1 for frontends, 2 for
15999 backends, 4 for servers, -1 for everything. These values can be ORed,
16000 for example:
16001 1 + 2 = 3 -> frontend + backend.
16002 1 + 2 + 4 = 7 -> frontend + backend + server.
16003 - <sid> is a server ID, -1 to dump everything from the selected proxy.
16004
16005 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020016006 $ echo "show info;show stat" | socat stdio unix-connect:/tmp/sock1
16007 >>> Name: HAProxy
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020016008 Version: 1.4-dev2-49
16009 Release_date: 2009/09/23
16010 Nbproc: 1
16011 Process_num: 1
16012 (...)
16013
16014 # pxname,svname,qcur,qmax,scur,smax,slim,stot,bin,bout,dreq, (...)
16015 stats,FRONTEND,,,0,0,1000,0,0,0,0,0,0,,,,,OPEN,,,,,,,,,1,1,0, (...)
16016 stats,BACKEND,0,0,0,0,1000,0,0,0,0,0,,0,0,0,0,UP,0,0,0,,0,250,(...)
16017 (...)
16018 www1,BACKEND,0,0,0,0,1000,0,0,0,0,0,,0,0,0,0,UP,1,1,0,,0,250, (...)
16019
16020 $
16021
16022 Here, two commands have been issued at once. That way it's easy to find
16023 which process the stats apply to in multi-process mode. Notice the empty
16024 line after the information output which marks the end of the first block.
16025 A similar empty line appears at the end of the second block (stats) so that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010016026 the reader knows the output has not been truncated.
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020016027
Baptiste Assmann3863f972015-05-17 00:33:24 +020016028show stat resolvers <resolvers section id>
16029 Dump statistics for the given resolvers section.
16030 For each name server, the following counters are reported:
16031 sent: number of DNS requests sent to this server
16032 valid: number of DNS valid responses received from this server
16033 update: number of DNS responses used to update the server's IP address
16034 cname: number of CNAME responses
16035 cname_error: CNAME errors encountered with this server
16036 any_err: number of empty response (IE: server does not support ANY type)
16037 nx: non existent domain response received from this server
16038 timeout: how many time this server did not answer in time
16039 refused: number of requests refused by this server
16040 other: any other DNS errors
16041 invalid: invalid DNS response (from a protocol point of view)
16042 too_big: too big response
16043 outdated: number of response arrived too late (after an other name server)
16044
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020016045show table
16046 Dump general information on all known stick-tables. Their name is returned
16047 (the name of the proxy which holds them), their type (currently zero, always
16048 IP), their size in maximum possible number of entries, and the number of
16049 entries currently in use.
16050
16051 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020016052 $ echo "show table" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090016053 >>> # table: front_pub, type: ip, size:204800, used:171454
16054 >>> # table: back_rdp, type: ip, size:204800, used:0
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020016055
Simon Horman17bce342011-06-15 15:18:47 +090016056show table <name> [ data.<type> <operator> <value> ] | [ key <key> ]
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020016057 Dump contents of stick-table <name>. In this mode, a first line of generic
16058 information about the table is reported as with "show table", then all
16059 entries are dumped. Since this can be quite heavy, it is possible to specify
Simon Horman17bce342011-06-15 15:18:47 +090016060 a filter in order to specify what entries to display.
16061
16062 When the "data." form is used the filter applies to the stored data (see
16063 "stick-table" in section 4.2). A stored data type must be specified
16064 in <type>, and this data type must be stored in the table otherwise an
16065 error is reported. The data is compared according to <operator> with the
16066 64-bit integer <value>. Operators are the same as with the ACLs :
16067
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020016068 - eq : match entries whose data is equal to this value
16069 - ne : match entries whose data is not equal to this value
16070 - le : match entries whose data is less than or equal to this value
16071 - ge : match entries whose data is greater than or equal to this value
16072 - lt : match entries whose data is less than this value
16073 - gt : match entries whose data is greater than this value
16074
Simon Hormanc88b8872011-06-15 15:18:49 +090016075
16076 When the key form is used the entry <key> is shown. The key must be of the
Simon Horman619e3cc2011-06-15 15:18:52 +090016077 same type as the table, which currently is limited to IPv4, IPv6, integer,
16078 and string.
Simon Horman17bce342011-06-15 15:18:47 +090016079
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020016080 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020016081 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090016082 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020016083 >>> 0x80e6a4c: key=127.0.0.1 use=0 exp=3594729 gpc0=0 conn_rate(30000)=1 \
16084 bytes_out_rate(60000)=187
16085 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
16086 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020016087
Willy Tarreau62a36c42010-08-17 15:53:10 +020016088 $ echo "show table http_proxy data.gpc0 gt 0" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090016089 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020016090 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
16091 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020016092
Willy Tarreau62a36c42010-08-17 15:53:10 +020016093 $ echo "show table http_proxy data.conn_rate gt 5" | \
16094 socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090016095 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020016096 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
16097 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020016098
Simon Horman17bce342011-06-15 15:18:47 +090016099 $ echo "show table http_proxy key 127.0.0.2" | \
16100 socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090016101 >>> # table: http_proxy, type: ip, size:204800, used:2
Simon Horman17bce342011-06-15 15:18:47 +090016102 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
16103 bytes_out_rate(60000)=191
16104
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020016105 When the data criterion applies to a dynamic value dependent on time such as
16106 a bytes rate, the value is dynamically computed during the evaluation of the
16107 entry in order to decide whether it has to be dumped or not. This means that
16108 such a filter could match for some time then not match anymore because as
16109 time goes, the average event rate drops.
16110
16111 It is possible to use this to extract lists of IP addresses abusing the
16112 service, in order to monitor them or even blacklist them in a firewall.
16113 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020016114 $ echo "show table http_proxy data.gpc0 gt 0" \
16115 | socat stdio /tmp/sock1 \
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020016116 | fgrep 'key=' | cut -d' ' -f2 | cut -d= -f2 > abusers-ip.txt
16117 ( or | awk '/key/{ print a[split($2,a,"=")]; }' )
Krzysztof Piotr Oledzki719e7262009-10-04 15:02:46 +020016118
Nenad Merdanovicc6985f02015-05-09 08:46:02 +020016119show tls-keys
16120 Dump all loaded TLS ticket keys. The TLS ticket key reference ID and the
16121 file from which the keys have been loaded is shown. Both of those can be
16122 used to update the TLS keys using "set ssl tls-key".
16123
Willy Tarreau532a4502011-09-07 22:37:44 +020016124shutdown frontend <frontend>
16125 Completely delete the specified frontend. All the ports it was bound to will
16126 be released. It will not be possible to enable the frontend anymore after
16127 this operation. This is intended to be used in environments where stopping a
16128 proxy is not even imaginable but a misconfigured proxy must be fixed. That
16129 way it's possible to release the port and bind it into another process to
16130 restore operations. The frontend will not appear at all on the stats page
16131 once it is terminated.
16132
16133 The frontend may be specified either by its name or by its numeric ID,
16134 prefixed with a sharp ('#').
16135
16136 This command is restricted and can only be issued on sockets configured for
16137 level "admin".
16138
Willy Tarreaua295edc2011-09-07 23:21:03 +020016139shutdown session <id>
16140 Immediately terminate the session matching the specified session identifier.
16141 This identifier is the first field at the beginning of the lines in the dumps
16142 of "show sess" (it corresponds to the session pointer). This can be used to
16143 terminate a long-running session without waiting for a timeout or when an
16144 endless transfer is ongoing. Such terminated sessions are reported with a 'K'
16145 flag in the logs.
16146
Cyril Bontée63a1eb2014-07-12 18:22:42 +020016147shutdown sessions server <backend>/<server>
Willy Tarreau52b2d222011-09-07 23:48:48 +020016148 Immediately terminate all the sessions attached to the specified server. This
16149 can be used to terminate long-running sessions after a server is put into
16150 maintenance mode, for instance. Such terminated sessions are reported with a
16151 'K' flag in the logs.
16152
Willy Tarreau0ba27502007-12-24 16:55:16 +010016153/*
16154 * Local variables:
16155 * fill-column: 79
16156 * End:
16157 */