blob: 1a7114e4bdd9b5cd8558cc2796d8f4322c05591c [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 Tarreau991b4782015-10-13 21:48:10 +02005 version 1.7
Willy Tarreau6a06a402007-07-15 20:15:28 +02006 willy tarreau
Willy Tarreau41d5e3a2016-08-14 12:25:21 +02007 2016/08/14
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
Christopher Fauletc3fe5332016-04-07 15:30:10 +02001069. Supported filters
1079.1. Trace
1089.2. HTTP compression
109
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200110
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
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200528 - cpu-map
Willy Tarreau6a06a402007-07-15 20:15:28 +0200529 - daemon
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200530 - description
531 - deviceatlas-json-file
532 - deviceatlas-log-level
533 - deviceatlas-separator
534 - deviceatlas-properties-cookie
Simon Horman98637e52014-06-20 12:30:16 +0900535 - external-check
Willy Tarreau6a06a402007-07-15 20:15:28 +0200536 - gid
537 - group
538 - log
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200539 - log-tag
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100540 - log-send-hostname
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200541 - lua-load
Willy Tarreau6a06a402007-07-15 20:15:28 +0200542 - nbproc
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200543 - node
Willy Tarreau6a06a402007-07-15 20:15:28 +0200544 - pidfile
Willy Tarreau1d549722016-02-16 12:41:57 +0100545 - presetenv
546 - resetenv
Willy Tarreau6a06a402007-07-15 20:15:28 +0200547 - uid
548 - ulimit-n
549 - user
Willy Tarreau1d549722016-02-16 12:41:57 +0100550 - setenv
Willy Tarreaufbee7132007-10-18 13:53:22 +0200551 - stats
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200552 - ssl-default-bind-ciphers
553 - ssl-default-bind-options
554 - ssl-default-server-ciphers
555 - ssl-default-server-options
556 - ssl-dh-param-file
Emeric Brun850efd52014-01-29 12:24:34 +0100557 - ssl-server-verify
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100558 - unix-bind
Willy Tarreau1d549722016-02-16 12:41:57 +0100559 - unsetenv
Thomas Holmesdb04f192015-05-18 13:21:39 +0100560 - 51degrees-data-file
561 - 51degrees-property-name-list
Dragan Dosen93b38d92015-06-29 16:43:25 +0200562 - 51degrees-property-separator
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200563 - 51degrees-cache-size
Willy Tarreaud72758d2010-01-12 10:42:19 +0100564
Willy Tarreau6a06a402007-07-15 20:15:28 +0200565 * Performance tuning
Willy Tarreau1746eec2014-04-25 10:46:47 +0200566 - max-spread-checks
Willy Tarreau6a06a402007-07-15 20:15:28 +0200567 - maxconn
Willy Tarreau81c25d02011-09-07 15:17:21 +0200568 - maxconnrate
William Lallemandd85f9172012-11-09 17:05:39 +0100569 - maxcomprate
William Lallemand072a2bf2012-11-20 17:01:01 +0100570 - maxcompcpuusage
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100571 - maxpipes
Willy Tarreau93e7c002013-10-07 18:51:07 +0200572 - maxsessrate
Willy Tarreau403edff2012-09-06 11:58:37 +0200573 - maxsslconn
Willy Tarreaue43d5322013-10-07 20:01:52 +0200574 - maxsslrate
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200575 - maxzlibmem
Willy Tarreau6a06a402007-07-15 20:15:28 +0200576 - noepoll
577 - nokqueue
578 - nopoll
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100579 - nosplice
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300580 - nogetaddrinfo
Willy Tarreaufe255b72007-10-14 23:09:26 +0200581 - spread-checks
Baptiste Assmann5626f482015-08-23 10:00:10 +0200582 - server-state-base
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200583 - server-state-file
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200584 - tune.buffers.limit
585 - tune.buffers.reserve
Willy Tarreau27a674e2009-08-17 07:23:33 +0200586 - tune.bufsize
Willy Tarreau43961d52010-10-04 20:39:20 +0200587 - tune.chksize
William Lallemandf3747832012-11-09 12:33:10 +0100588 - tune.comp.maxlevel
Willy Tarreau193b8c62012-11-22 00:17:38 +0100589 - tune.http.cookielen
Willy Tarreauac1932d2011-10-24 19:14:41 +0200590 - tune.http.maxhdr
Willy Tarreau7e312732014-02-12 16:35:14 +0100591 - tune.idletimer
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100592 - tune.lua.forced-yield
Willy Tarreau32f61e22015-03-18 17:54:59 +0100593 - tune.lua.maxmem
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100594 - tune.lua.session-timeout
595 - tune.lua.task-timeout
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +0200596 - tune.lua.service-timeout
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100597 - tune.maxaccept
598 - tune.maxpollevents
Willy Tarreau27a674e2009-08-17 07:23:33 +0200599 - tune.maxrewrite
Willy Tarreauf3045d22015-04-29 16:24:50 +0200600 - tune.pattern.cache-size
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200601 - tune.pipesize
Willy Tarreaue803de22010-01-21 17:43:04 +0100602 - tune.rcvbuf.client
603 - tune.rcvbuf.server
Willy Tarreaub22fc302015-12-14 12:04:35 +0100604 - tune.recv_enough
Willy Tarreaue803de22010-01-21 17:43:04 +0100605 - tune.sndbuf.client
606 - tune.sndbuf.server
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100607 - tune.ssl.cachesize
Willy Tarreaubfd59462013-02-21 07:46:09 +0100608 - tune.ssl.lifetime
Emeric Brun8dc60392014-05-09 13:52:00 +0200609 - tune.ssl.force-private-cache
Willy Tarreaubfd59462013-02-21 07:46:09 +0100610 - tune.ssl.maxrecord
Remi Gacognef46cd6e2014-06-12 14:58:40 +0200611 - tune.ssl.default-dh-param
Christopher Faulet31af49d2015-06-09 17:29:50 +0200612 - tune.ssl.ssl-ctx-cache-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200613 - tune.vars.global-max-size
614 - tune.vars.reqres-max-size
615 - tune.vars.sess-max-size
616 - tune.vars.txn-max-size
William Lallemanda509e4c2012-11-07 16:54:34 +0100617 - tune.zlib.memlevel
618 - tune.zlib.windowsize
Willy Tarreaud72758d2010-01-12 10:42:19 +0100619
Willy Tarreau6a06a402007-07-15 20:15:28 +0200620 * Debugging
621 - debug
622 - quiet
Willy Tarreau6a06a402007-07-15 20:15:28 +0200623
624
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006253.1. Process management and security
Willy Tarreau6a06a402007-07-15 20:15:28 +0200626------------------------------------
627
Emeric Brunc8e8d122012-10-02 18:42:10 +0200628ca-base <dir>
629 Assigns a default directory to fetch SSL CA certificates and CRLs from when a
Emeric Brunfd33a262012-10-11 16:28:27 +0200630 relative path is used with "ca-file" or "crl-file" directives. Absolute
631 locations specified in "ca-file" and "crl-file" prevail and ignore "ca-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +0200632
Willy Tarreau6a06a402007-07-15 20:15:28 +0200633chroot <jail dir>
634 Changes current directory to <jail dir> and performs a chroot() there before
635 dropping privileges. This increases the security level in case an unknown
636 vulnerability would be exploited, since it would make it very hard for the
637 attacker to exploit the system. This only works when the process is started
638 with superuser privileges. It is important to ensure that <jail_dir> is both
639 empty and unwritable to anyone.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100640
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100641cpu-map <"all"|"odd"|"even"|process_num> <cpu-set>...
642 On Linux 2.6 and above, it is possible to bind a process to a specific CPU
643 set. This means that the process will never run on other CPUs. The "cpu-map"
644 directive specifies CPU sets for process sets. The first argument is the
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100645 process number to bind. This process must have a number between 1 and 32 or
646 64, depending on the machine's word size, and any process IDs above nbproc
647 are ignored. It is possible to specify all processes at once using "all",
648 only odd numbers using "odd" or even numbers using "even", just like with the
649 "bind-process" directive. The second and forthcoming arguments are CPU sets.
650 Each CPU set is either a unique number between 0 and 31 or 63 or a range with
651 two such numbers delimited by a dash ('-'). Multiple CPU numbers or ranges
652 may be specified, and the processes will be allowed to bind to all of them.
653 Obviously, multiple "cpu-map" directives may be specified. Each "cpu-map"
654 directive will replace the previous ones when they overlap.
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100655
Emeric Brunc8e8d122012-10-02 18:42:10 +0200656crt-base <dir>
657 Assigns a default directory to fetch SSL certificates from when a relative
658 path is used with "crtfile" directives. Absolute locations specified after
659 "crtfile" prevail and ignore "crt-base".
660
Willy Tarreau6a06a402007-07-15 20:15:28 +0200661daemon
662 Makes the process fork into background. This is the recommended mode of
663 operation. It is equivalent to the command line "-D" argument. It can be
664 disabled by the command line "-db" argument.
665
David Carlier8167f302015-06-01 13:50:06 +0200666deviceatlas-json-file <path>
667 Sets the path of the DeviceAtlas JSON data file to be loaded by the API.
668 The path must be a valid JSON data file and accessible by Haproxy process.
669
670deviceatlas-log-level <value>
671 Sets the level of informations returned by the API. This directive is
672 optional and set to 0 by default if not set.
673
674deviceatlas-separator <char>
675 Sets the character separator for the API properties results. This directive
676 is optional and set to | by default if not set.
677
Cyril Bonté0306c4a2015-10-26 22:37:38 +0100678deviceatlas-properties-cookie <name>
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200679 Sets the client cookie's name used for the detection if the DeviceAtlas
680 Client-side component was used during the request. This directive is optional
681 and set to DAPROPS by default if not set.
David Carlier29b3ca32015-09-25 14:09:21 +0100682
Simon Horman98637e52014-06-20 12:30:16 +0900683external-check
684 Allows the use of an external agent to perform health checks.
685 This is disabled by default as a security precaution.
686 See "option external-check".
687
Willy Tarreau6a06a402007-07-15 20:15:28 +0200688gid <number>
689 Changes the process' group ID to <number>. It is recommended that the group
690 ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
691 be started with a user belonging to this group, or with superuser privileges.
Michael Schererab012dd2013-01-12 18:35:19 +0100692 Note that if haproxy is started from a user having supplementary groups, it
693 will only be able to drop these groups if started with superuser privileges.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200694 See also "group" and "uid".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100695
Willy Tarreau6a06a402007-07-15 20:15:28 +0200696group <group name>
697 Similar to "gid" but uses the GID of group name <group name> from /etc/group.
698 See also "gid" and "user".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100699
Dragan Dosen7ad31542015-09-28 17:16:47 +0200700log <address> [len <length>] [format <format>] <facility> [max level [min level]]
Willy Tarreau6a06a402007-07-15 20:15:28 +0200701 Adds a global syslog server. Up to two global servers can be defined. They
702 will receive logs for startups and exits, as well as all logs from proxies
Robert Tsai81ae1952007-12-05 10:47:29 +0100703 configured with "log global".
704
705 <address> can be one of:
706
Willy Tarreau2769aa02007-12-27 18:26:09 +0100707 - An IPv4 address optionally followed by a colon and a UDP port. If
Robert Tsai81ae1952007-12-05 10:47:29 +0100708 no port is specified, 514 is used by default (the standard syslog
709 port).
710
David du Colombier24bb5f52011-03-17 10:40:23 +0100711 - An IPv6 address followed by a colon and optionally a UDP port. If
712 no port is specified, 514 is used by default (the standard syslog
713 port).
714
Robert Tsai81ae1952007-12-05 10:47:29 +0100715 - A filesystem path to a UNIX domain socket, keeping in mind
716 considerations for chroot (be sure the path is accessible inside
717 the chroot) and uid/gid (be sure the path is appropriately
718 writeable).
719
William Lallemandb2f07452015-05-12 14:27:13 +0200720 You may want to reference some environment variables in the address
721 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +0100722
Willy Tarreau18324f52014-06-27 18:10:07 +0200723 <length> is an optional maximum line length. Log lines larger than this value
724 will be truncated before being sent. The reason is that syslog
725 servers act differently on log line length. All servers support the
726 default value of 1024, but some servers simply drop larger lines
727 while others do log them. If a server supports long lines, it may
728 make sense to set this value here in order to avoid truncating long
729 lines. Similarly, if a server drops long lines, it is preferable to
730 truncate them before sending them. Accepted values are 80 to 65535
731 inclusive. The default value of 1024 is generally fine for all
732 standard usages. Some specific cases of long captures or
733 JSON-formated logs may require larger values.
734
Dragan Dosen7ad31542015-09-28 17:16:47 +0200735 <format> is the log format used when generating syslog messages. It may be
736 one of the following :
737
738 rfc3164 The RFC3164 syslog message format. This is the default.
739 (https://tools.ietf.org/html/rfc3164)
740
741 rfc5424 The RFC5424 syslog message format.
742 (https://tools.ietf.org/html/rfc5424)
743
Robert Tsai81ae1952007-12-05 10:47:29 +0100744 <facility> must be one of the 24 standard syslog facilities :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200745
746 kern user mail daemon auth syslog lpr news
747 uucp cron auth2 ftp ntp audit alert cron2
748 local0 local1 local2 local3 local4 local5 local6 local7
749
750 An optional level can be specified to filter outgoing messages. By default,
Willy Tarreauf7edefa2009-05-10 17:20:05 +0200751 all messages are sent. If a maximum level is specified, only messages with a
752 severity at least as important as this level will be sent. An optional minimum
753 level can be specified. If it is set, logs emitted with a more severe level
754 than this one will be capped to this level. This is used to avoid sending
755 "emerg" messages on all terminals on some default syslog configurations.
756 Eight levels are known :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200757
Cyril Bontédc4d9032012-04-08 21:57:39 +0200758 emerg alert crit err warning notice info debug
Willy Tarreau6a06a402007-07-15 20:15:28 +0200759
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100760log-send-hostname [<string>]
761 Sets the hostname field in the syslog header. If optional "string" parameter
762 is set the header is set to the string contents, otherwise uses the hostname
763 of the system. Generally used if one is not relaying logs through an
764 intermediate syslog server or for simply customizing the hostname printed in
765 the logs.
766
Kevinm48936af2010-12-22 16:08:21 +0000767log-tag <string>
768 Sets the tag field in the syslog header to this string. It defaults to the
769 program name as launched from the command line, which usually is "haproxy".
770 Sometimes it can be useful to differentiate between multiple processes
Willy Tarreau094af4e2015-01-07 15:03:42 +0100771 running on the same host. See also the per-proxy "log-tag" directive.
Kevinm48936af2010-12-22 16:08:21 +0000772
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100773lua-load <file>
774 This global directive loads and executes a Lua file. This directive can be
775 used multiple times.
776
Willy Tarreau6a06a402007-07-15 20:15:28 +0200777nbproc <number>
778 Creates <number> processes when going daemon. This requires the "daemon"
779 mode. By default, only one process is created, which is the recommended mode
780 of operation. For systems limited to small sets of file descriptors per
781 process, it may be needed to fork multiple daemons. USING MULTIPLE PROCESSES
782 IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. See also "daemon".
783
784pidfile <pidfile>
785 Writes pids of all daemons into file <pidfile>. This option is equivalent to
786 the "-p" command line argument. The file must be accessible to the user
787 starting the process. See also "daemon".
788
Willy Tarreau1d549722016-02-16 12:41:57 +0100789presetenv <name> <value>
790 Sets environment variable <name> to value <value>. If the variable exists, it
791 is NOT overwritten. The changes immediately take effect so that the next line
792 in the configuration file sees the new value. See also "setenv", "resetenv",
793 and "unsetenv".
794
795resetenv [<name> ...]
796 Removes all environment variables except the ones specified in argument. It
797 allows to use a clean controlled environment before setting new values with
798 setenv or unsetenv. Please note that some internal functions may make use of
799 some environment variables, such as time manipulation functions, but also
800 OpenSSL or even external checks. This must be used with extreme care and only
801 after complete validation. The changes immediately take effect so that the
802 next line in the configuration file sees the new environment. See also
803 "setenv", "presetenv", and "unsetenv".
804
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100805stats bind-process [ all | odd | even | <number 1-64>[-<number 1-64>] ] ...
Willy Tarreau35b7b162012-10-22 23:17:18 +0200806 Limits the stats socket to a certain set of processes numbers. By default the
807 stats socket is bound to all processes, causing a warning to be emitted when
808 nbproc is greater than 1 because there is no way to select the target process
809 when connecting. However, by using this setting, it becomes possible to pin
810 the stats socket to a specific set of processes, typically the first one. The
811 warning will automatically be disabled when this setting is used, whatever
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100812 the number of processes used. The maximum process ID depends on the machine's
Willy Tarreauae302532014-05-07 19:22:24 +0200813 word size (32 or 64). A better option consists in using the "process" setting
814 of the "stats socket" line to force the process on each line.
Willy Tarreau35b7b162012-10-22 23:17:18 +0200815
Baptiste Assmann5626f482015-08-23 10:00:10 +0200816server-state-base <directory>
817 Specifies the directory prefix to be prepended in front of all servers state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +0200818 file names which do not start with a '/'. See also "server-state-file",
819 "load-server-state-from-file" and "server-state-file-name".
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200820
821server-state-file <file>
822 Specifies the path to the file containing state of servers. If the path starts
823 with a slash ('/'), it is considered absolute, otherwise it is considered
824 relative to the directory specified using "server-state-base" (if set) or to
825 the current directory. Before reloading HAProxy, it is possible to save the
826 servers' current state using the stats command "show servers state". The
827 output of this command must be written in the file pointed by <file>. When
828 starting up, before handling traffic, HAProxy will read, load and apply state
829 for each server found in the file and available in its current running
Baptiste Assmann01c6cc32015-08-23 11:45:29 +0200830 configuration. See also "server-state-base" and "show servers state",
831 "load-server-state-from-file" and "server-state-file-name"
Baptiste Assmann5626f482015-08-23 10:00:10 +0200832
Willy Tarreau1d549722016-02-16 12:41:57 +0100833setenv <name> <value>
834 Sets environment variable <name> to value <value>. If the variable exists, it
835 is overwritten. The changes immediately take effect so that the next line in
836 the configuration file sees the new value. See also "presetenv", "resetenv",
837 and "unsetenv".
838
Willy Tarreau610f04b2014-02-13 11:36:41 +0100839ssl-default-bind-ciphers <ciphers>
840 This setting is only available when support for OpenSSL was built in. It sets
841 the default string describing the list of cipher algorithms ("cipher suite")
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300842 that are negotiated during the SSL/TLS handshake for all "bind" lines which
Willy Tarreau610f04b2014-02-13 11:36:41 +0100843 do not explicitly define theirs. The format of the string is defined in
844 "man 1 ciphers" from OpenSSL man pages, and can be for instance a string such
845 as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes). Please check the
846 "bind" keyword for more information.
847
Emeric Brun2c86cbf2014-10-30 15:56:50 +0100848ssl-default-bind-options [<option>]...
849 This setting is only available when support for OpenSSL was built in. It sets
850 default ssl-options to force on all "bind" lines. Please check the "bind"
851 keyword to see available options.
852
853 Example:
854 global
855 ssl-default-bind-options no-sslv3 no-tls-tickets
856
Willy Tarreau610f04b2014-02-13 11:36:41 +0100857ssl-default-server-ciphers <ciphers>
858 This setting is only available when support for OpenSSL was built in. It
859 sets the default string describing the list of cipher algorithms that are
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300860 negotiated during the SSL/TLS handshake with the server, for all "server"
Willy Tarreau610f04b2014-02-13 11:36:41 +0100861 lines which do not explicitly define theirs. The format of the string is
862 defined in "man 1 ciphers". Please check the "server" keyword for more
863 information.
864
Emeric Brun2c86cbf2014-10-30 15:56:50 +0100865ssl-default-server-options [<option>]...
866 This setting is only available when support for OpenSSL was built in. It sets
867 default ssl-options to force on all "server" lines. Please check the "server"
868 keyword to see available options.
869
Remi Gacogne47783ef2015-05-29 15:53:22 +0200870ssl-dh-param-file <file>
871 This setting is only available when support for OpenSSL was built in. It sets
872 the default DH parameters that are used during the SSL/TLS handshake when
873 ephemeral Diffie-Hellman (DHE) key exchange is used, for all "bind" lines
874 which do not explicitely define theirs. It will be overridden by custom DH
875 parameters found in a bind certificate file if any. If custom DH parameters
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200876 are not specified either by using ssl-dh-param-file or by setting them
877 directly in the certificate file, pre-generated DH parameters of the size
878 specified by tune.ssl.default-dh-param will be used. Custom parameters are
879 known to be more secure and therefore their use is recommended.
Remi Gacogne47783ef2015-05-29 15:53:22 +0200880 Custom DH parameters may be generated by using the OpenSSL command
881 "openssl dhparam <size>", where size should be at least 2048, as 1024-bit DH
882 parameters should not be considered secure anymore.
883
Emeric Brun850efd52014-01-29 12:24:34 +0100884ssl-server-verify [none|required]
885 The default behavior for SSL verify on servers side. If specified to 'none',
886 servers certificates are not verified. The default is 'required' except if
887 forced using cmdline option '-dV'.
888
Willy Tarreauabb175f2012-09-24 12:43:26 +0200889stats socket [<address:port>|<path>] [param*]
890 Binds a UNIX socket to <path> or a TCPv4/v6 address to <address:port>.
891 Connections to this socket will return various statistics outputs and even
892 allow some commands to be issued to change some runtime settings. Please
Kevin Decherf949c7202015-10-13 23:26:44 +0200893 consult section 9.2 "Unix Socket commands" of Management Guide for more
894 details.
Willy Tarreau6162db22009-10-10 17:13:00 +0200895
Willy Tarreauabb175f2012-09-24 12:43:26 +0200896 All parameters supported by "bind" lines are supported, for instance to
897 restrict access to some users or their access rights. Please consult
898 section 5.1 for more information.
Willy Tarreaufbee7132007-10-18 13:53:22 +0200899
900stats timeout <timeout, in milliseconds>
901 The default timeout on the stats socket is set to 10 seconds. It is possible
902 to change this value with "stats timeout". The value must be passed in
Willy Tarreaubefdff12007-12-02 22:27:38 +0100903 milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.
Willy Tarreaufbee7132007-10-18 13:53:22 +0200904
905stats maxconn <connections>
906 By default, the stats socket is limited to 10 concurrent connections. It is
907 possible to change this value with "stats maxconn".
908
Willy Tarreau6a06a402007-07-15 20:15:28 +0200909uid <number>
910 Changes the process' user ID to <number>. It is recommended that the user ID
911 is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
912 be started with superuser privileges in order to be able to switch to another
913 one. See also "gid" and "user".
914
915ulimit-n <number>
916 Sets the maximum number of per-process file-descriptors to <number>. By
917 default, it is automatically computed, so it is recommended not to use this
918 option.
919
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100920unix-bind [ prefix <prefix> ] [ mode <mode> ] [ user <user> ] [ uid <uid> ]
921 [ group <group> ] [ gid <gid> ]
922
923 Fixes common settings to UNIX listening sockets declared in "bind" statements.
924 This is mainly used to simplify declaration of those UNIX sockets and reduce
925 the risk of errors, since those settings are most commonly required but are
926 also process-specific. The <prefix> setting can be used to force all socket
927 path to be relative to that directory. This might be needed to access another
928 component's chroot. Note that those paths are resolved before haproxy chroots
929 itself, so they are absolute. The <mode>, <user>, <uid>, <group> and <gid>
930 all have the same meaning as their homonyms used by the "bind" statement. If
931 both are specified, the "bind" statement has priority, meaning that the
932 "unix-bind" settings may be seen as process-wide default settings.
933
Willy Tarreau1d549722016-02-16 12:41:57 +0100934unsetenv [<name> ...]
935 Removes environment variables specified in arguments. This can be useful to
936 hide some sensitive information that are occasionally inherited from the
937 user's environment during some operations. Variables which did not exist are
938 silently ignored so that after the operation, it is certain that none of
939 these variables remain. The changes immediately take effect so that the next
940 line in the configuration file will not see these variables. See also
941 "setenv", "presetenv", and "resetenv".
942
Willy Tarreau6a06a402007-07-15 20:15:28 +0200943user <user name>
944 Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.
945 See also "uid" and "group".
946
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +0200947node <name>
948 Only letters, digits, hyphen and underscore are allowed, like in DNS names.
949
950 This statement is useful in HA configurations where two or more processes or
951 servers share the same IP address. By setting a different node-name on all
952 nodes, it becomes easy to immediately spot what server is handling the
953 traffic.
954
955description <text>
956 Add a text that describes the instance.
957
958 Please note that it is required to escape certain characters (# for example)
959 and this text is inserted into a html page so you should avoid using
960 "<" and ">" characters.
961
Thomas Holmesdb04f192015-05-18 13:21:39 +010096251degrees-data-file <file path>
963 The path of the 51Degrees data file to provide device detection services. The
964 file should be unzipped and accessible by HAProxy with relevavnt permissions.
965
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200966 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +0100967 compiled with USE_51DEGREES.
968
96951degrees-property-name-list [<string>]
970 A list of 51Degrees property names to be load from the dataset. A full list
971 of names is available on the 51Degrees website:
972 https://51degrees.com/resources/property-dictionary
973
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200974 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +0100975 compiled with USE_51DEGREES.
976
Dragan Dosen93b38d92015-06-29 16:43:25 +020097751degrees-property-separator <char>
Thomas Holmesdb04f192015-05-18 13:21:39 +0100978 A char that will be appended to every property value in a response header
979 containing 51Degrees results. If not set that will be set as ','.
980
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200981 Please note that this option is only available when haproxy has been
982 compiled with USE_51DEGREES.
983
98451degrees-cache-size <number>
985 Sets the size of the 51Degrees converter cache to <number> entries. This
986 is an LRU cache which reminds previous device detections and their results.
987 By default, this cache is disabled.
988
989 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +0100990 compiled with USE_51DEGREES.
991
Willy Tarreau6a06a402007-07-15 20:15:28 +0200992
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009933.2. Performance tuning
Willy Tarreau6a06a402007-07-15 20:15:28 +0200994-----------------------
995
Willy Tarreau1746eec2014-04-25 10:46:47 +0200996max-spread-checks <delay in milliseconds>
997 By default, haproxy tries to spread the start of health checks across the
998 smallest health check interval of all the servers in a farm. The principle is
999 to avoid hammering services running on the same server. But when using large
1000 check intervals (10 seconds or more), the last servers in the farm take some
1001 time before starting to be tested, which can be a problem. This parameter is
1002 used to enforce an upper bound on delay between the first and the last check,
1003 even if the servers' check intervals are larger. When servers run with
1004 shorter intervals, their intervals will be respected though.
1005
Willy Tarreau6a06a402007-07-15 20:15:28 +02001006maxconn <number>
1007 Sets the maximum per-process number of concurrent connections to <number>. It
1008 is equivalent to the command-line argument "-n". Proxies will stop accepting
1009 connections when this limit is reached. The "ulimit-n" parameter is
Willy Tarreau8274e102014-06-19 15:31:25 +02001010 automatically adjusted according to this value. See also "ulimit-n". Note:
1011 the "select" poller cannot reliably use more than 1024 file descriptors on
1012 some platforms. If your platform only supports select and reports "select
1013 FAILED" on startup, you need to reduce maxconn until it works (slightly
Willy Tarreaud0256482015-01-15 21:45:22 +01001014 below 500 in general). If this value is not set, it will default to the value
1015 set in DEFAULT_MAXCONN at build time (reported in haproxy -vv) if no memory
1016 limit is enforced, or will be computed based on the memory limit, the buffer
1017 size, memory allocated to compression, SSL cache size, and use or not of SSL
1018 and the associated maxsslconn (which can also be automatic).
Willy Tarreau6a06a402007-07-15 20:15:28 +02001019
Willy Tarreau81c25d02011-09-07 15:17:21 +02001020maxconnrate <number>
1021 Sets the maximum per-process number of connections per second to <number>.
1022 Proxies will stop accepting connections when this limit is reached. It can be
1023 used to limit the global capacity regardless of each frontend capacity. It is
1024 important to note that this can only be used as a service protection measure,
1025 as there will not necessarily be a fair share between frontends when the
1026 limit is reached, so it's a good idea to also limit each frontend to some
1027 value close to its expected share. Also, lowering tune.maxaccept can improve
1028 fairness.
1029
William Lallemandd85f9172012-11-09 17:05:39 +01001030maxcomprate <number>
1031 Sets the maximum per-process input compression rate to <number> kilobytes
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001032 per second. For each session, if the maximum is reached, the compression
William Lallemandd85f9172012-11-09 17:05:39 +01001033 level will be decreased during the session. If the maximum is reached at the
1034 beginning of a session, the session will not compress at all. If the maximum
1035 is not reached, the compression level will be increased up to
1036 tune.comp.maxlevel. A value of zero means there is no limit, this is the
1037 default value.
1038
William Lallemand072a2bf2012-11-20 17:01:01 +01001039maxcompcpuusage <number>
1040 Sets the maximum CPU usage HAProxy can reach before stopping the compression
1041 for new requests or decreasing the compression level of current requests.
1042 It works like 'maxcomprate' but measures CPU usage instead of incoming data
1043 bandwidth. The value is expressed in percent of the CPU used by haproxy. In
1044 case of multiple processes (nbproc > 1), each process manages its individual
1045 usage. A value of 100 disable the limit. The default value is 100. Setting
1046 a lower value will prevent the compression work from slowing the whole
1047 process down and from introducing high latencies.
1048
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001049maxpipes <number>
1050 Sets the maximum per-process number of pipes to <number>. Currently, pipes
1051 are only used by kernel-based tcp splicing. Since a pipe contains two file
1052 descriptors, the "ulimit-n" value will be increased accordingly. The default
1053 value is maxconn/4, which seems to be more than enough for most heavy usages.
1054 The splice code dynamically allocates and releases pipes, and can fall back
1055 to standard copy, so setting this value too low may only impact performance.
1056
Willy Tarreau93e7c002013-10-07 18:51:07 +02001057maxsessrate <number>
1058 Sets the maximum per-process number of sessions per second to <number>.
1059 Proxies will stop accepting connections when this limit is reached. It can be
1060 used to limit the global capacity regardless of each frontend capacity. It is
1061 important to note that this can only be used as a service protection measure,
1062 as there will not necessarily be a fair share between frontends when the
1063 limit is reached, so it's a good idea to also limit each frontend to some
1064 value close to its expected share. Also, lowering tune.maxaccept can improve
1065 fairness.
1066
Willy Tarreau403edff2012-09-06 11:58:37 +02001067maxsslconn <number>
1068 Sets the maximum per-process number of concurrent SSL connections to
1069 <number>. By default there is no SSL-specific limit, which means that the
1070 global maxconn setting will apply to all connections. Setting this limit
1071 avoids having openssl use too much memory and crash when malloc returns NULL
1072 (since it unfortunately does not reliably check for such conditions). Note
1073 that the limit applies both to incoming and outgoing connections, so one
1074 connection which is deciphered then ciphered accounts for 2 SSL connections.
Willy Tarreaud0256482015-01-15 21:45:22 +01001075 If this value is not set, but a memory limit is enforced, this value will be
1076 automatically computed based on the memory limit, maxconn, the buffer size,
1077 memory allocated to compression, SSL cache size, and use of SSL in either
1078 frontends, backends or both. If neither maxconn nor maxsslconn are specified
1079 when there is a memory limit, haproxy will automatically adjust these values
1080 so that 100% of the connections can be made over SSL with no risk, and will
1081 consider the sides where it is enabled (frontend, backend, both).
Willy Tarreau403edff2012-09-06 11:58:37 +02001082
Willy Tarreaue43d5322013-10-07 20:01:52 +02001083maxsslrate <number>
1084 Sets the maximum per-process number of SSL sessions per second to <number>.
1085 SSL listeners will stop accepting connections when this limit is reached. It
1086 can be used to limit the global SSL CPU usage regardless of each frontend
1087 capacity. It is important to note that this can only be used as a service
1088 protection measure, as there will not necessarily be a fair share between
1089 frontends when the limit is reached, so it's a good idea to also limit each
1090 frontend to some value close to its expected share. It is also important to
1091 note that the sessions are accounted before they enter the SSL stack and not
1092 after, which also protects the stack against bad handshakes. Also, lowering
1093 tune.maxaccept can improve fairness.
1094
William Lallemand9d5f5482012-11-07 16:12:57 +01001095maxzlibmem <number>
1096 Sets the maximum amount of RAM in megabytes per process usable by the zlib.
1097 When the maximum amount is reached, future sessions will not compress as long
1098 as RAM is unavailable. When sets to 0, there is no limit.
William Lallemande3a7d992012-11-20 11:25:20 +01001099 The default value is 0. The value is available in bytes on the UNIX socket
1100 with "show info" on the line "MaxZlibMemUsage", the memory used by zlib is
1101 "ZlibMemUsage" in bytes.
1102
Willy Tarreau6a06a402007-07-15 20:15:28 +02001103noepoll
1104 Disables the use of the "epoll" event polling system on Linux. It is
1105 equivalent to the command-line argument "-de". The next polling system
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001106 used will generally be "poll". See also "nopoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001107
1108nokqueue
1109 Disables the use of the "kqueue" event polling system on BSD. It is
1110 equivalent to the command-line argument "-dk". The next polling system
1111 used will generally be "poll". See also "nopoll".
1112
1113nopoll
1114 Disables the use of the "poll" event polling system. It is equivalent to the
1115 command-line argument "-dp". The next polling system used will be "select".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001116 It should never be needed to disable "poll" since it's available on all
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001117 platforms supported by HAProxy. See also "nokqueue" and "noepoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001118
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001119nosplice
1120 Disables the use of kernel tcp splicing between sockets on Linux. It is
1121 equivalent to the command line argument "-dS". Data will then be copied
1122 using conventional and more portable recv/send calls. Kernel tcp splicing is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001123 limited to some very recent instances of kernel 2.6. Most versions between
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001124 2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not
1125 be used. This option makes it easier to globally disable kernel splicing in
1126 case of doubt. See also "option splice-auto", "option splice-request" and
1127 "option splice-response".
1128
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001129nogetaddrinfo
1130 Disables the use of getaddrinfo(3) for name resolving. It is equivalent to
1131 the command line argument "-dG". Deprecated gethostbyname(3) will be used.
1132
Willy Tarreaufe255b72007-10-14 23:09:26 +02001133spread-checks <0..50, in percent>
Simon Hormand60d6912013-11-25 10:46:36 +09001134 Sometimes it is desirable to avoid sending agent and health checks to
1135 servers at exact intervals, for instance when many logical servers are
1136 located on the same physical server. With the help of this parameter, it
1137 becomes possible to add some randomness in the check interval between 0
1138 and +/- 50%. A value between 2 and 5 seems to show good results. The
1139 default value remains at 0.
Willy Tarreaufe255b72007-10-14 23:09:26 +02001140
Willy Tarreau33cb0652014-12-23 22:52:37 +01001141tune.buffers.limit <number>
1142 Sets a hard limit on the number of buffers which may be allocated per process.
1143 The default value is zero which means unlimited. The minimum non-zero value
1144 will always be greater than "tune.buffers.reserve" and should ideally always
1145 be about twice as large. Forcing this value can be particularly useful to
1146 limit the amount of memory a process may take, while retaining a sane
1147 behaviour. When this limit is reached, sessions which need a buffer wait for
1148 another one to be released by another session. Since buffers are dynamically
1149 allocated and released, the waiting time is very short and not perceptible
1150 provided that limits remain reasonable. In fact sometimes reducing the limit
1151 may even increase performance by increasing the CPU cache's efficiency. Tests
1152 have shown good results on average HTTP traffic with a limit to 1/10 of the
1153 expected global maxconn setting, which also significantly reduces memory
1154 usage. The memory savings come from the fact that a number of connections
1155 will not allocate 2*tune.bufsize. It is best not to touch this value unless
1156 advised to do so by an haproxy core developer.
1157
Willy Tarreau1058ae72014-12-23 22:40:40 +01001158tune.buffers.reserve <number>
1159 Sets the number of buffers which are pre-allocated and reserved for use only
1160 during memory shortage conditions resulting in failed memory allocations. The
1161 minimum value is 2 and is also the default. There is no reason a user would
1162 want to change this value, it's mostly aimed at haproxy core developers.
1163
Willy Tarreau27a674e2009-08-17 07:23:33 +02001164tune.bufsize <number>
1165 Sets the buffer size to this size (in bytes). Lower values allow more
1166 sessions to coexist in the same amount of RAM, and higher values allow some
1167 applications with very large cookies to work. The default value is 16384 and
1168 can be changed at build time. It is strongly recommended not to change this
1169 from the default value, as very low values will break some services such as
1170 statistics, and values larger than default size will increase memory usage,
1171 possibly causing the system to run out of memory. At least the global maxconn
1172 parameter should be decreased by the same factor as this one is increased.
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04001173 If HTTP request is larger than (tune.bufsize - tune.maxrewrite), haproxy will
1174 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
1175 than this size, haproxy will return HTTP 502 (Bad Gateway).
Willy Tarreau27a674e2009-08-17 07:23:33 +02001176
Willy Tarreau43961d52010-10-04 20:39:20 +02001177tune.chksize <number>
1178 Sets the check buffer size to this size (in bytes). Higher values may help
1179 find string or regex patterns in very large pages, though doing so may imply
1180 more memory and CPU usage. The default value is 16384 and can be changed at
1181 build time. It is not recommended to change this value, but to use better
1182 checks whenever possible.
1183
William Lallemandf3747832012-11-09 12:33:10 +01001184tune.comp.maxlevel <number>
1185 Sets the maximum compression level. The compression level affects CPU
1186 usage during compression. This value affects CPU usage during compression.
1187 Each session using compression initializes the compression algorithm with
1188 this value. The default value is 1.
1189
Willy Tarreau193b8c62012-11-22 00:17:38 +01001190tune.http.cookielen <number>
1191 Sets the maximum length of captured cookies. This is the maximum value that
1192 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
1193 will automatically be truncated to this one. It is important not to set too
1194 high a value because all cookie captures still allocate this size whatever
1195 their configured value (they share a same pool). This value is per request
1196 per response, so the memory allocated is twice this value per connection.
1197 When not specified, the limit is set to 63 characters. It is recommended not
1198 to change this value.
1199
Willy Tarreauac1932d2011-10-24 19:14:41 +02001200tune.http.maxhdr <number>
1201 Sets the maximum number of headers in a request. When a request comes with a
1202 number of headers greater than this value (including the first line), it is
1203 rejected with a "400 Bad Request" status code. Similarly, too large responses
1204 are blocked with "502 Bad Gateway". The default value is 101, which is enough
1205 for all usages, considering that the widely deployed Apache server uses the
1206 same limit. It can be useful to push this limit further to temporarily allow
1207 a buggy application to work by the time it gets fixed. Keep in mind that each
1208 new header consumes 32bits of memory for each session, so don't push this
1209 limit too high.
1210
Willy Tarreau7e312732014-02-12 16:35:14 +01001211tune.idletimer <timeout>
1212 Sets the duration after which haproxy will consider that an empty buffer is
1213 probably associated with an idle stream. This is used to optimally adjust
1214 some packet sizes while forwarding large and small data alternatively. The
1215 decision to use splice() or to send large buffers in SSL is modulated by this
1216 parameter. The value is in milliseconds between 0 and 65535. A value of zero
1217 means that haproxy will not try to detect idle streams. The default is 1000,
1218 which seems to correctly detect end user pauses (eg: read a page before
1219 clicking). There should be not reason for changing this value. Please check
1220 tune.ssl.maxrecord below.
1221
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001222tune.lua.forced-yield <number>
1223 This directive forces the Lua engine to execute a yield each <number> of
1224 instructions executed. This permits interruptng a long script and allows the
1225 HAProxy scheduler to process other tasks like accepting connections or
1226 forwarding traffic. The default value is 10000 instructions. If HAProxy often
1227 executes some Lua code but more reactivity is required, this value can be
1228 lowered. If the Lua code is quite long and its result is absolutely required
1229 to process the data, the <number> can be increased.
1230
Willy Tarreau32f61e22015-03-18 17:54:59 +01001231tune.lua.maxmem
1232 Sets the maximum amount of RAM in megabytes per process usable by Lua. By
1233 default it is zero which means unlimited. It is important to set a limit to
1234 ensure that a bug in a script will not result in the system running out of
1235 memory.
1236
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001237tune.lua.session-timeout <timeout>
1238 This is the execution timeout for the Lua sessions. This is useful for
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001239 preventing infinite loops or spending too much time in Lua. This timeout
1240 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
1241 not taked in account. The default timeout is 4s.
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001242
1243tune.lua.task-timeout <timeout>
1244 Purpose is the same as "tune.lua.session-timeout", but this timeout is
1245 dedicated to the tasks. By default, this timeout isn't set because a task may
1246 remain alive during of the lifetime of HAProxy. For example, a task used to
1247 check servers.
1248
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001249tune.lua.service-timeout <timeout>
1250 This is the execution timeout for the Lua services. This is useful for
1251 preventing infinite loops or spending too much time in Lua. This timeout
1252 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
1253 not taked in account. The default timeout is 4s.
1254
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001255tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +01001256 Sets the maximum number of consecutive connections a process may accept in a
1257 row before switching to other work. In single process mode, higher numbers
1258 give better performance at high connection rates. However in multi-process
1259 modes, keeping a bit of fairness between processes generally is better to
1260 increase performance. This value applies individually to each listener, so
1261 that the number of processes a listener is bound to is taken into account.
1262 This value defaults to 64. In multi-process mode, it is divided by twice
1263 the number of processes the listener is bound to. Setting this value to -1
1264 completely disables the limitation. It should normally not be needed to tweak
1265 this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001266
1267tune.maxpollevents <number>
1268 Sets the maximum amount of events that can be processed at once in a call to
1269 the polling system. The default value is adapted to the operating system. It
1270 has been noticed that reducing it below 200 tends to slightly decrease
1271 latency at the expense of network bandwidth, and increasing it above 200
1272 tends to trade latency for slightly increased bandwidth.
1273
Willy Tarreau27a674e2009-08-17 07:23:33 +02001274tune.maxrewrite <number>
1275 Sets the reserved buffer space to this size in bytes. The reserved space is
1276 used for header rewriting or appending. The first reads on sockets will never
1277 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
1278 bufsize, though that does not make much sense since there are rarely large
1279 numbers of headers to add. Setting it too high prevents processing of large
1280 requests or responses. Setting it too low prevents addition of new headers
1281 to already large requests or to POST requests. It is generally wise to set it
1282 to about 1024. It is automatically readjusted to half of bufsize if it is
1283 larger than that. This means you don't have to worry about it when changing
1284 bufsize.
1285
Willy Tarreauf3045d22015-04-29 16:24:50 +02001286tune.pattern.cache-size <number>
1287 Sets the size of the pattern lookup cache to <number> entries. This is an LRU
1288 cache which reminds previous lookups and their results. It is used by ACLs
1289 and maps on slow pattern lookups, namely the ones using the "sub", "reg",
1290 "dir", "dom", "end", "bin" match methods as well as the case-insensitive
1291 strings. It applies to pattern expressions which means that it will be able
1292 to memorize the result of a lookup among all the patterns specified on a
1293 configuration line (including all those loaded from files). It automatically
1294 invalidates entries which are updated using HTTP actions or on the CLI. The
1295 default cache size is set to 10000 entries, which limits its footprint to
1296 about 5 MB on 32-bit systems and 8 MB on 64-bit systems. There is a very low
1297 risk of collision in this cache, which is in the order of the size of the
1298 cache divided by 2^64. Typically, at 10000 requests per second with the
1299 default cache size of 10000 entries, there's 1% chance that a brute force
1300 attack could cause a single collision after 60 years, or 0.1% after 6 years.
1301 This is considered much lower than the risk of a memory corruption caused by
1302 aging components. If this is not acceptable, the cache can be disabled by
1303 setting this parameter to 0.
1304
Willy Tarreaubd9a0a72011-10-23 21:14:29 +02001305tune.pipesize <number>
1306 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
1307 are the default size for the system. But sometimes when using TCP splicing,
1308 it can improve performance to increase pipe sizes, especially if it is
1309 suspected that pipes are not filled and that many calls to splice() are
1310 performed. This has an impact on the kernel's memory footprint, so this must
1311 not be changed if impacts are not understood.
1312
Willy Tarreaue803de22010-01-21 17:43:04 +01001313tune.rcvbuf.client <number>
1314tune.rcvbuf.server <number>
1315 Forces the kernel socket receive buffer size on the client or the server side
1316 to the specified value in bytes. This value applies to all TCP/HTTP frontends
1317 and backends. It should normally never be set, and the default size (0) lets
1318 the kernel autotune this value depending on the amount of available memory.
1319 However it can sometimes help to set it to very low values (eg: 4096) in
1320 order to save kernel memory by preventing it from buffering too large amounts
1321 of received data. Lower values will significantly increase CPU usage though.
1322
Willy Tarreaub22fc302015-12-14 12:04:35 +01001323tune.recv_enough <number>
1324 Haproxy uses some hints to detect that a short read indicates the end of the
1325 socket buffers. One of them is that a read returns more than <recv_enough>
1326 bytes, which defaults to 10136 (7 segments of 1448 each). This default value
1327 may be changed by this setting to better deal with workloads involving lots
1328 of short messages such as telnet or SSH sessions.
1329
Willy Tarreaue803de22010-01-21 17:43:04 +01001330tune.sndbuf.client <number>
1331tune.sndbuf.server <number>
1332 Forces the kernel socket send buffer size on the client or the server side to
1333 the specified value in bytes. This value applies to all TCP/HTTP frontends
1334 and backends. It should normally never be set, and the default size (0) lets
1335 the kernel autotune this value depending on the amount of available memory.
1336 However it can sometimes help to set it to very low values (eg: 4096) in
1337 order to save kernel memory by preventing it from buffering too large amounts
1338 of received data. Lower values will significantly increase CPU usage though.
1339 Another use case is to prevent write timeouts with extremely slow clients due
1340 to the kernel waiting for a large part of the buffer to be read before
1341 notifying haproxy again.
1342
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001343tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +01001344 Sets the size of the global SSL session cache, in a number of blocks. A block
1345 is large enough to contain an encoded session without peer certificate.
1346 An encoded session with peer certificate is stored in multiple blocks
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001347 depending on the size of the peer certificate. A block uses approximately
Emeric Brunaf9619d2012-11-28 18:47:52 +01001348 200 bytes of memory. The default value may be forced at build time, otherwise
1349 defaults to 20000. When the cache is full, the most idle entries are purged
1350 and reassigned. Higher values reduce the occurrence of such a purge, hence
1351 the number of CPU-intensive SSL handshakes by ensuring that all users keep
1352 their session as long as possible. All entries are pre-allocated upon startup
Emeric Brun22890a12012-12-28 14:41:32 +01001353 and are shared between all processes if "nbproc" is greater than 1. Setting
1354 this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001355
Emeric Brun8dc60392014-05-09 13:52:00 +02001356tune.ssl.force-private-cache
1357 This boolean disables SSL session cache sharing between all processes. It
1358 should normally not be used since it will force many renegotiations due to
1359 clients hitting a random process. But it may be required on some operating
1360 systems where none of the SSL cache synchronization method may be used. In
1361 this case, adding a first layer of hash-based load balancing before the SSL
1362 layer might limit the impact of the lack of session sharing.
1363
Emeric Brun4f65bff2012-11-16 15:11:00 +01001364tune.ssl.lifetime <timeout>
1365 Sets how long a cached SSL session may remain valid. This time is expressed
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001366 in seconds and defaults to 300 (5 min). It is important to understand that it
Emeric Brun4f65bff2012-11-16 15:11:00 +01001367 does not guarantee that sessions will last that long, because if the cache is
1368 full, the longest idle sessions will be purged despite their configured
1369 lifetime. The real usefulness of this setting is to prevent sessions from
1370 being used for too long.
1371
Willy Tarreaubfd59462013-02-21 07:46:09 +01001372tune.ssl.maxrecord <number>
1373 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
1374 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
1375 data only once it has received a full record. With large records, it means
1376 that clients might have to download up to 16kB of data before starting to
1377 process them. Limiting the value can improve page load times on browsers
1378 located over high latency or low bandwidth networks. It is suggested to find
1379 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
1380 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
1381 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
1382 2859 gave good results during tests. Use "strace -e trace=write" to find the
Willy Tarreau7e312732014-02-12 16:35:14 +01001383 best value. Haproxy will automatically switch to this setting after an idle
1384 stream has been detected (see tune.idletimer above).
Willy Tarreaubfd59462013-02-21 07:46:09 +01001385
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001386tune.ssl.default-dh-param <number>
1387 Sets the maximum size of the Diffie-Hellman parameters used for generating
1388 the ephemeral/temporary Diffie-Hellman key in case of DHE key exchange. The
1389 final size will try to match the size of the server's RSA (or DSA) key (e.g,
1390 a 2048 bits temporary DH key for a 2048 bits RSA key), but will not exceed
1391 this maximum value. Default value if 1024. Only 1024 or higher values are
1392 allowed. Higher values will increase the CPU load, and values greater than
1393 1024 bits are not supported by Java 7 and earlier clients. This value is not
Remi Gacogne47783ef2015-05-29 15:53:22 +02001394 used if static Diffie-Hellman parameters are supplied either directly
1395 in the certificate file or by using the ssl-dh-param-file parameter.
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001396
Christopher Faulet31af49d2015-06-09 17:29:50 +02001397tune.ssl.ssl-ctx-cache-size <number>
1398 Sets the size of the cache used to store generated certificates to <number>
1399 entries. This is a LRU cache. Because generating a SSL certificate
1400 dynamically is expensive, they are cached. The default cache size is set to
1401 1000 entries.
1402
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001403tune.vars.global-max-size <size>
1404tune.vars.reqres-max-size <size>
1405tune.vars.sess-max-size <size>
1406tune.vars.txn-max-size <size>
Daniel Schneller0b547052016-03-21 20:46:57 +01001407 These four tunes help to manage the maximum amount of memory used by the
1408 variables system. "global" limits the overall amount of memory available
1409 for all scopes. "sess" limits the memory for the session scope, "txn" for
1410 the transaction scope, and "reqres" limits the memory for each request or
1411 response processing.
1412 Memory accounting is hierarchical, meaning more coarse grained limits
1413 include the finer grained ones: "sess" includes "txn", and "txn" includes
1414 "reqres".
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001415
Daniel Schneller0b547052016-03-21 20:46:57 +01001416 For example, when "tune.vars.sess-max-size" is limited to 100,
1417 "tune.vars.txn-max-size" and "tune.vars.reqres-max-size" cannot exceed
1418 100 either. If we create a variable "txn.var" that contains 100 bytes,
1419 all available space is consumed.
1420 Notice that exceeding the limits at runtime will not result in an error
1421 message, but values might be cut off or corrupted. So make sure to accurately
1422 plan for the amount of space needed to store all your variables.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001423
William Lallemanda509e4c2012-11-07 16:54:34 +01001424tune.zlib.memlevel <number>
1425 Sets the memLevel parameter in zlib initialization for each session. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001426 defines how much memory should be allocated for the internal compression
William Lallemanda509e4c2012-11-07 16:54:34 +01001427 state. A value of 1 uses minimum memory but is slow and reduces compression
1428 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
1429 between 1 and 9. The default value is 8.
1430
1431tune.zlib.windowsize <number>
1432 Sets the window size (the size of the history buffer) as a parameter of the
1433 zlib initialization for each session. Larger values of this parameter result
1434 in better compression at the expense of memory usage. Can be a value between
1435 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001436
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014373.3. Debugging
1438--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02001439
1440debug
1441 Enables debug mode which dumps to stdout all exchanges, and disables forking
1442 into background. It is the equivalent of the command-line argument "-d". It
1443 should never be used in a production configuration since it may prevent full
1444 system startup.
1445
1446quiet
1447 Do not display any message during startup. It is equivalent to the command-
1448 line argument "-q".
1449
Emeric Brunf099e792010-09-27 12:05:28 +02001450
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010014513.4. Userlists
1452--------------
1453It is possible to control access to frontend/backend/listen sections or to
1454http stats by allowing only authenticated and authorized users. To do this,
1455it is required to create at least one userlist and to define users.
1456
1457userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +01001458 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001459 used to store authentication & authorization data for independent customers.
1460
1461group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +01001462 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001463 attach users to this group by using a comma separated list of names
1464 proceeded by "users" keyword.
1465
Cyril Bontéf0c60612010-02-06 14:44:47 +01001466user <username> [password|insecure-password <password>]
1467 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001468 Adds user <username> to the current userlist. Both secure (encrypted) and
1469 insecure (unencrypted) passwords can be used. Encrypted passwords are
Cyril Bonté78caf842010-03-10 22:41:43 +01001470 evaluated using the crypt(3) function so depending of the system's
1471 capabilities, different algorithms are supported. For example modern Glibc
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001472 based Linux system supports MD5, SHA-256, SHA-512 and of course classic,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001473 DES-based method of encrypting passwords.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001474
1475
1476 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +01001477 userlist L1
1478 group G1 users tiger,scott
1479 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001480
Cyril Bontéf0c60612010-02-06 14:44:47 +01001481 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
1482 user scott insecure-password elgato
1483 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001484
Cyril Bontéf0c60612010-02-06 14:44:47 +01001485 userlist L2
1486 group G1
1487 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001488
Cyril Bontéf0c60612010-02-06 14:44:47 +01001489 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
1490 user scott insecure-password elgato groups G1,G2
1491 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001492
1493 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001494
Emeric Brunf099e792010-09-27 12:05:28 +02001495
14963.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02001497----------
Emeric Brun94900952015-06-11 18:25:54 +02001498It is possible to propagate entries of any data-types in stick-tables between
1499several haproxy instances over TCP connections in a multi-master fashion. Each
1500instance pushes its local updates and insertions to remote peers. The pushed
1501values overwrite remote ones without aggregation. Interrupted exchanges are
1502automatically detected and recovered from the last known point.
1503In addition, during a soft restart, the old process connects to the new one
1504using such a TCP connection to push all its entries before the new process
1505tries to connect to other peers. That ensures very fast replication during a
1506reload, it typically takes a fraction of a second even for large tables.
1507Note that Server IDs are used to identify servers remotely, so it is important
1508that configurations look similar or at least that the same IDs are forced on
1509each server on all participants.
Emeric Brunf099e792010-09-27 12:05:28 +02001510
1511peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001512 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02001513 which is referenced by one or more stick-tables.
1514
Willy Tarreau77e4bd12015-05-01 20:02:17 +02001515disabled
1516 Disables a peers section. It disables both listening and any synchronization
1517 related to this section. This is provided to disable synchronization of stick
1518 tables without having to comment out all "peers" references.
1519
1520enable
1521 This re-enables a disabled peers section which was previously disabled.
1522
Emeric Brunf099e792010-09-27 12:05:28 +02001523peer <peername> <ip>:<port>
1524 Defines a peer inside a peers section.
1525 If <peername> is set to the local peer name (by default hostname, or forced
1526 using "-L" command line option), haproxy will listen for incoming remote peer
1527 connection on <ip>:<port>. Otherwise, <ip>:<port> defines where to connect to
1528 to join the remote peer, and <peername> is used at the protocol level to
1529 identify and validate the remote peer on the server side.
1530
1531 During a soft restart, local peer <ip>:<port> is used by the old instance to
1532 connect the new one and initiate a complete replication (teaching process).
1533
1534 It is strongly recommended to have the exact same peers declaration on all
1535 peers and to only rely on the "-L" command line argument to change the local
1536 peer name. This makes it easier to maintain coherent configuration files
1537 across all peers.
1538
William Lallemandb2f07452015-05-12 14:27:13 +02001539 You may want to reference some environment variables in the address
1540 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01001541
Cyril Bontédc4d9032012-04-08 21:57:39 +02001542 Example:
Emeric Brunf099e792010-09-27 12:05:28 +02001543 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001544 peer haproxy1 192.168.0.1:1024
1545 peer haproxy2 192.168.0.2:1024
1546 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02001547
1548 backend mybackend
1549 mode tcp
1550 balance roundrobin
1551 stick-table type ip size 20k peers mypeers
1552 stick on src
1553
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001554 server srv1 192.168.0.30:80
1555 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02001556
1557
Simon Horman51a1cf62015-02-03 13:00:44 +090015583.6. Mailers
1559------------
1560It is possible to send email alerts when the state of servers changes.
1561If configured email alerts are sent to each mailer that is configured
1562in a mailers section. Email is sent to mailers using SMTP.
1563
Pieter Baauw386a1272015-08-16 15:26:24 +02001564mailers <mailersect>
Simon Horman51a1cf62015-02-03 13:00:44 +09001565 Creates a new mailer list with the name <mailersect>. It is an
1566 independent section which is referenced by one or more proxies.
1567
1568mailer <mailername> <ip>:<port>
1569 Defines a mailer inside a mailers section.
1570
1571 Example:
1572 mailers mymailers
1573 mailer smtp1 192.168.0.1:587
1574 mailer smtp2 192.168.0.2:587
1575
1576 backend mybackend
1577 mode tcp
1578 balance roundrobin
1579
1580 email-alert mailers mymailers
1581 email-alert from test1@horms.org
1582 email-alert to test2@horms.org
1583
1584 server srv1 192.168.0.30:80
1585 server srv2 192.168.0.31:80
1586
Pieter Baauw235fcfc2016-02-13 15:33:40 +01001587timeout mail <time>
1588 Defines the time available for a mail/connection to be made and send to
1589 the mail-server. If not defined the default value is 10 seconds. To allow
1590 for at least two SYN-ACK packets to be send during initial TCP handshake it
1591 is advised to keep this value above 4 seconds.
1592
1593 Example:
1594 mailers mymailers
1595 timeout mail 20s
1596 mailer smtp1 192.168.0.1:587
Simon Horman51a1cf62015-02-03 13:00:44 +09001597
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015984. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02001599----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001600
Willy Tarreau6a06a402007-07-15 20:15:28 +02001601Proxy configuration can be located in a set of sections :
William Lallemand6e62fb62015-04-28 16:55:23 +02001602 - defaults [<name>]
Willy Tarreau6a06a402007-07-15 20:15:28 +02001603 - frontend <name>
1604 - backend <name>
1605 - listen <name>
1606
1607A "defaults" section sets default parameters for all other sections following
1608its declaration. Those default parameters are reset by the next "defaults"
1609section. See below for the list of parameters which can be set in a "defaults"
Willy Tarreau0ba27502007-12-24 16:55:16 +01001610section. The name is optional but its use is encouraged for better readability.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001611
1612A "frontend" section describes a set of listening sockets accepting client
1613connections.
1614
1615A "backend" section describes a set of servers to which the proxy will connect
1616to forward incoming connections.
1617
1618A "listen" section defines a complete proxy with its frontend and backend
1619parts combined in one section. It is generally useful for TCP-only traffic.
1620
Willy Tarreau0ba27502007-12-24 16:55:16 +01001621All proxy names must be formed from upper and lower case letters, digits,
1622'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
1623case-sensitive, which means that "www" and "WWW" are two different proxies.
1624
1625Historically, all proxy names could overlap, it just caused troubles in the
1626logs. Since the introduction of content switching, it is mandatory that two
1627proxies with overlapping capabilities (frontend/backend) have different names.
1628However, it is still permitted that a frontend and a backend share the same
1629name, as this configuration seems to be commonly encountered.
1630
1631Right now, two major proxy modes are supported : "tcp", also known as layer 4,
1632and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001633bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01001634protocol, and can interact with it by allowing, blocking, switching, adding,
1635modifying, or removing arbitrary contents in requests or responses, based on
1636arbitrary criteria.
1637
Willy Tarreau70dffda2014-01-30 03:07:23 +01001638In HTTP mode, the processing applied to requests and responses flowing over
1639a connection depends in the combination of the frontend's HTTP options and
1640the backend's. HAProxy supports 5 connection modes :
1641
1642 - KAL : keep alive ("option http-keep-alive") which is the default mode : all
1643 requests and responses are processed, and connections remain open but idle
1644 between responses and new requests.
1645
1646 - TUN: tunnel ("option http-tunnel") : this was the default mode for versions
1647 1.0 to 1.5-dev21 : only the first request and response are processed, and
1648 everything else is forwarded with no analysis at all. This mode should not
1649 be used as it creates lots of trouble with logging and HTTP processing.
1650
1651 - PCL: passive close ("option httpclose") : exactly the same as tunnel mode,
1652 but with "Connection: close" appended in both directions to try to make
1653 both ends close after the first request/response exchange.
1654
1655 - SCL: server close ("option http-server-close") : the server-facing
1656 connection is closed after the end of the response is received, but the
1657 client-facing connection remains open.
1658
1659 - FCL: forced close ("option forceclose") : the connection is actively closed
1660 after the end of the response.
1661
1662The effective mode that will be applied to a connection passing through a
1663frontend and a backend can be determined by both proxy modes according to the
1664following matrix, but in short, the modes are symmetric, keep-alive is the
1665weakest option and force close is the strongest.
1666
1667 Backend mode
1668
1669 | KAL | TUN | PCL | SCL | FCL
1670 ----+-----+-----+-----+-----+----
1671 KAL | KAL | TUN | PCL | SCL | FCL
1672 ----+-----+-----+-----+-----+----
1673 TUN | TUN | TUN | PCL | SCL | FCL
1674 Frontend ----+-----+-----+-----+-----+----
1675 mode PCL | PCL | PCL | PCL | FCL | FCL
1676 ----+-----+-----+-----+-----+----
1677 SCL | SCL | SCL | FCL | SCL | FCL
1678 ----+-----+-----+-----+-----+----
1679 FCL | FCL | FCL | FCL | FCL | FCL
1680
Willy Tarreau0ba27502007-12-24 16:55:16 +01001681
Willy Tarreau70dffda2014-01-30 03:07:23 +01001682
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016834.1. Proxy keywords matrix
1684--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001685
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001686The following list of keywords is supported. Most of them may only be used in a
1687limited set of section types. Some of them are marked as "deprecated" because
1688they are inherited from an old syntax which may be confusing or functionally
1689limited, and there are new recommended keywords to replace them. Keywords
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001690marked with "(*)" can be optionally inverted using the "no" prefix, eg. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001691option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02001692and must be disabled for a specific instance. Such options may also be prefixed
1693with "default" in order to restore default settings regardless of what has been
1694specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001695
Willy Tarreau6a06a402007-07-15 20:15:28 +02001696
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001697 keyword defaults frontend listen backend
1698------------------------------------+----------+----------+---------+---------
1699acl - X X X
Willy Tarreau294d0f02015-08-10 19:40:12 +02001700appsession - - - -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001701backlog X X X -
1702balance X - X X
1703bind - X X -
1704bind-process X X X X
1705block - X X X
1706capture cookie - X X -
1707capture request header - X X -
1708capture response header - X X -
1709clitimeout (deprecated) X X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02001710compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001711contimeout (deprecated) X - X X
1712cookie X - X X
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02001713declare capture - X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001714default-server X - X X
1715default_backend X X X -
1716description - X X X
1717disabled X X X X
1718dispatch - - X X
Simon Horman51a1cf62015-02-03 13:00:44 +09001719email-alert from X X X X
Simon Horman64e34162015-02-06 11:11:57 +09001720email-alert level X X X X
Simon Horman51a1cf62015-02-03 13:00:44 +09001721email-alert mailers X X X X
1722email-alert myhostname X X X X
1723email-alert to X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001724enabled X X X X
1725errorfile X X X X
1726errorloc X X X X
1727errorloc302 X X X X
1728-- keyword -------------------------- defaults - frontend - listen -- backend -
1729errorloc303 X X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001730force-persist - X X X
Christopher Fauletc3fe5332016-04-07 15:30:10 +02001731filter - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001732fullconn X - X X
1733grace X X X X
1734hash-type X - X X
1735http-check disable-on-404 X - X X
Willy Tarreaubd741542010-03-16 18:46:54 +01001736http-check expect - - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02001737http-check send-state X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001738http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02001739http-response - X X X
Willy Tarreau30631952015-08-06 15:05:24 +02001740http-reuse X - X X
Baptiste Assmann2c42ef52013-10-09 21:57:02 +02001741http-send-name-header - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001742id - X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001743ignore-persist - X X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001744load-server-state-from-file X - X X
William Lallemand0f99e342011-10-12 17:50:54 +02001745log (*) X X X X
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01001746log-format X X X -
Dragan Dosen7ad31542015-09-28 17:16:47 +02001747log-format-sd X X X -
Willy Tarreau094af4e2015-01-07 15:03:42 +01001748log-tag X X X X
Willy Tarreauc35362a2014-04-25 13:58:37 +02001749max-keep-alive-queue X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001750maxconn X X X -
1751mode X X X X
1752monitor fail - X X -
1753monitor-net X X X -
1754monitor-uri X X X -
1755option abortonclose (*) X - X X
1756option accept-invalid-http-request (*) X X X -
1757option accept-invalid-http-response (*) X - X X
1758option allbackups (*) X - X X
1759option checkcache (*) X - X X
1760option clitcpka (*) X X X -
1761option contstats (*) X X X -
1762option dontlog-normal (*) X X X -
1763option dontlognull (*) X X X -
1764option forceclose (*) X X X X
1765-- keyword -------------------------- defaults - frontend - listen -- backend -
1766option forwardfor X X X X
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02001767option http-buffer-request (*) X X X X
Willy Tarreau82649f92015-05-01 22:40:51 +02001768option http-ignore-probes (*) X X X -
Willy Tarreau16bfb022010-01-16 19:48:41 +01001769option http-keep-alive (*) X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02001770option http-no-delay (*) X X X X
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02001771option http-pretend-keepalive (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001772option http-server-close (*) X X X X
Willy Tarreau02bce8b2014-01-30 00:15:28 +01001773option http-tunnel (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001774option http-use-proxy-header (*) X X X -
1775option httpchk X - X X
1776option httpclose (*) X X X X
1777option httplog X X X X
1778option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001779option independent-streams (*) X X X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02001780option ldap-check X - X X
Simon Horman98637e52014-06-20 12:30:16 +09001781option external-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001782option log-health-checks (*) X - X X
1783option log-separate-errors (*) X X X -
1784option logasap (*) X X X -
1785option mysql-check X - X X
1786option nolinger (*) X X X X
1787option originalto X X X X
1788option persist (*) X - X X
Baptiste Assmann809e22a2015-10-12 20:22:55 +02001789option pgsql-check X - X X
1790option prefer-last-server (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001791option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02001792option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001793option smtpchk X - X X
1794option socket-stats (*) X X X -
1795option splice-auto (*) X X X X
1796option splice-request (*) X X X X
1797option splice-response (*) X X X X
1798option srvtcpka (*) X - X X
1799option ssl-hello-chk X - X X
1800-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreaued179852013-12-16 01:07:00 +01001801option tcp-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001802option tcp-smart-accept (*) X X X -
1803option tcp-smart-connect (*) X - X X
1804option tcpka X X X X
1805option tcplog X X X X
1806option transparent (*) X - X X
Simon Horman98637e52014-06-20 12:30:16 +09001807external-check command X - X X
1808external-check path X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001809persist rdp-cookie X - X X
1810rate-limit sessions X X X -
1811redirect - X X X
1812redisp (deprecated) X - X X
1813redispatch (deprecated) X - X X
1814reqadd - X X X
1815reqallow - X X X
1816reqdel - X X X
1817reqdeny - X X X
1818reqiallow - X X X
1819reqidel - X X X
1820reqideny - X X X
1821reqipass - X X X
1822reqirep - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001823reqitarpit - X X X
1824reqpass - X X X
1825reqrep - X X X
1826-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001827reqtarpit - X X X
1828retries X - X X
1829rspadd - X X X
1830rspdel - X X X
1831rspdeny - X X X
1832rspidel - X X X
1833rspideny - X X X
1834rspirep - X X X
1835rsprep - X X X
1836server - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001837server-state-file-name X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001838source X - X X
1839srvtimeout (deprecated) X - X X
Baptiste Assmann5a549212015-10-12 20:30:24 +02001840stats admin - X X X
1841stats auth X X X X
1842stats enable X X X X
1843stats hide-version X X X X
1844stats http-request - X X X
1845stats realm X X X X
1846stats refresh X X X X
1847stats scope X X X X
1848stats show-desc X X X X
1849stats show-legends X X X X
1850stats show-node X X X X
1851stats uri X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001852-- keyword -------------------------- defaults - frontend - listen -- backend -
1853stick match - - X X
1854stick on - - X X
1855stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02001856stick store-response - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001857stick-table - - X X
Willy Tarreau938c7fe2014-04-25 14:21:39 +02001858tcp-check connect - - X X
1859tcp-check expect - - X X
1860tcp-check send - - X X
1861tcp-check send-binary - - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02001862tcp-request connection - X X -
1863tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02001864tcp-request inspect-delay - X X X
Emeric Brun0a3b67f2010-09-24 15:34:53 +02001865tcp-response content - - X X
1866tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001867timeout check X - X X
1868timeout client X X X -
Willy Tarreau05cdd962014-05-10 14:30:07 +02001869timeout client-fin X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001870timeout clitimeout (deprecated) X X X -
1871timeout connect X - X X
1872timeout contimeout (deprecated) X - X X
1873timeout http-keep-alive X X X X
1874timeout http-request X X X X
1875timeout queue X - X X
1876timeout server X - X X
Willy Tarreau05cdd962014-05-10 14:30:07 +02001877timeout server-fin X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001878timeout srvtimeout (deprecated) X - X X
1879timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02001880timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001881transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01001882unique-id-format X X X -
1883unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001884use_backend - X X -
Willy Tarreau4a5cade2012-04-05 21:09:48 +02001885use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001886------------------------------------+----------+----------+---------+---------
1887 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02001888
Willy Tarreau0ba27502007-12-24 16:55:16 +01001889
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018904.2. Alphabetically sorted keywords reference
1891---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001892
1893This section provides a description of each keyword and its usage.
1894
1895
1896acl <aclname> <criterion> [flags] [operator] <value> ...
1897 Declare or complete an access list.
1898 May be used in sections : defaults | frontend | listen | backend
1899 no | yes | yes | yes
1900 Example:
1901 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
1902 acl invalid_src src_port 0:1023
1903 acl local_dst hdr(host) -i localhost
1904
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001905 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001906
1907
Cyril Bontéb21570a2009-11-29 20:04:48 +01001908appsession <cookie> len <length> timeout <holdtime>
1909 [request-learn] [prefix] [mode <path-parameters|query-string>]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001910 Define session stickiness on an existing application cookie.
1911 May be used in sections : defaults | frontend | listen | backend
1912 no | no | yes | yes
1913 Arguments :
1914 <cookie> this is the name of the cookie used by the application and which
1915 HAProxy will have to learn for each new session.
1916
Cyril Bontéb21570a2009-11-29 20:04:48 +01001917 <length> this is the max number of characters that will be memorized and
Willy Tarreau0ba27502007-12-24 16:55:16 +01001918 checked in each cookie value.
1919
1920 <holdtime> this is the time after which the cookie will be removed from
1921 memory if unused. If no unit is specified, this time is in
1922 milliseconds.
1923
Cyril Bontébf47aeb2009-10-15 00:15:40 +02001924 request-learn
1925 If this option is specified, then haproxy will be able to learn
1926 the cookie found in the request in case the server does not
1927 specify any in response. This is typically what happens with
1928 PHPSESSID cookies, or when haproxy's session expires before
1929 the application's session and the correct server is selected.
1930 It is recommended to specify this option to improve reliability.
1931
Cyril Bontéb21570a2009-11-29 20:04:48 +01001932 prefix When this option is specified, haproxy will match on the cookie
1933 prefix (or URL parameter prefix). The appsession value is the
1934 data following this prefix.
1935
1936 Example :
1937 appsession ASPSESSIONID len 64 timeout 3h prefix
1938
1939 This will match the cookie ASPSESSIONIDXXXX=XXXXX,
1940 the appsession value will be XXXX=XXXXX.
1941
1942 mode This option allows to change the URL parser mode.
1943 2 modes are currently supported :
1944 - path-parameters :
1945 The parser looks for the appsession in the path parameters
1946 part (each parameter is separated by a semi-colon), which is
1947 convenient for JSESSIONID for example.
1948 This is the default mode if the option is not set.
1949 - query-string :
1950 In this mode, the parser will look for the appsession in the
1951 query string.
1952
Willy Tarreau294d0f02015-08-10 19:40:12 +02001953 As of version 1.6, appsessions was removed. It is more flexible and more
1954 convenient to use stick-tables instead, and stick-tables support multi-master
1955 replication and data conservation across reloads, which appsessions did not.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001956
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01001957 See also : "cookie", "capture cookie", "balance", "stick", "stick-table",
1958 "ignore-persist", "nbproc" and "bind-process".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001959
1960
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01001961backlog <conns>
1962 Give hints to the system about the approximate listen backlog desired size
1963 May be used in sections : defaults | frontend | listen | backend
1964 yes | yes | yes | no
1965 Arguments :
1966 <conns> is the number of pending connections. Depending on the operating
1967 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02001968 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01001969
1970 In order to protect against SYN flood attacks, one solution is to increase
1971 the system's SYN backlog size. Depending on the system, sometimes it is just
1972 tunable via a system parameter, sometimes it is not adjustable at all, and
1973 sometimes the system relies on hints given by the application at the time of
1974 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
1975 to the listen() syscall. On systems which can make use of this value, it can
1976 sometimes be useful to be able to specify a different value, hence this
1977 backlog parameter.
1978
1979 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
1980 used as a hint and the system accepts up to the smallest greater power of
1981 two, and never more than some limits (usually 32768).
1982
1983 See also : "maxconn" and the target operating system's tuning guide.
1984
1985
Willy Tarreau0ba27502007-12-24 16:55:16 +01001986balance <algorithm> [ <arguments> ]
Willy Tarreau226071e2014-04-10 11:55:45 +02001987balance url_param <param> [check_post]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001988 Define the load balancing algorithm to be used in a backend.
1989 May be used in sections : defaults | frontend | listen | backend
1990 yes | no | yes | yes
1991 Arguments :
1992 <algorithm> is the algorithm used to select a server when doing load
1993 balancing. This only applies when no persistence information
1994 is available, or when a connection is redispatched to another
1995 server. <algorithm> may be one of the following :
1996
1997 roundrobin Each server is used in turns, according to their weights.
1998 This is the smoothest and fairest algorithm when the server's
1999 processing time remains equally distributed. This algorithm
2000 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02002001 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08002002 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02002003 large farms, when a server becomes up after having been down
2004 for a very short time, it may sometimes take a few hundreds
2005 requests for it to be re-integrated into the farm and start
2006 receiving traffic. This is normal, though very rare. It is
2007 indicated here in case you would have the chance to observe
2008 it, so that you don't worry.
2009
2010 static-rr Each server is used in turns, according to their weights.
2011 This algorithm is as similar to roundrobin except that it is
2012 static, which means that changing a server's weight on the
2013 fly will have no effect. On the other hand, it has no design
2014 limitation on the number of servers, and when a server goes
2015 up, it is always immediately reintroduced into the farm, once
2016 the full map is recomputed. It also uses slightly less CPU to
2017 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01002018
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01002019 leastconn The server with the lowest number of connections receives the
2020 connection. Round-robin is performed within groups of servers
2021 of the same load to ensure that all servers will be used. Use
2022 of this algorithm is recommended where very long sessions are
2023 expected, such as LDAP, SQL, TSE, etc... but is not very well
2024 suited for protocols using short sessions such as HTTP. This
2025 algorithm is dynamic, which means that server weights may be
2026 adjusted on the fly for slow starts for instance.
2027
Willy Tarreauf09c6602012-02-13 17:12:08 +01002028 first The first server with available connection slots receives the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002029 connection. The servers are chosen from the lowest numeric
Willy Tarreauf09c6602012-02-13 17:12:08 +01002030 identifier to the highest (see server parameter "id"), which
2031 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02002032 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01002033 not make sense to use this algorithm without setting maxconn.
2034 The purpose of this algorithm is to always use the smallest
2035 number of servers so that extra servers can be powered off
2036 during non-intensive hours. This algorithm ignores the server
2037 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02002038 or IMAP than HTTP, though it can be useful there too. In
2039 order to use this algorithm efficiently, it is recommended
2040 that a cloud controller regularly checks server usage to turn
2041 them off when unused, and regularly checks backend queue to
2042 turn new servers on when the queue inflates. Alternatively,
2043 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01002044
Willy Tarreau0ba27502007-12-24 16:55:16 +01002045 source The source IP address is hashed and divided by the total
2046 weight of the running servers to designate which server will
2047 receive the request. This ensures that the same client IP
2048 address will always reach the same server as long as no
2049 server goes down or up. If the hash result changes due to the
2050 number of running servers changing, many clients will be
2051 directed to a different server. This algorithm is generally
2052 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002053 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01002054 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002055 static by default, which means that changing a server's
2056 weight on the fly will have no effect, but this can be
2057 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002058
Oskar Stolc8dc41842012-05-19 10:19:54 +01002059 uri This algorithm hashes either the left part of the URI (before
2060 the question mark) or the whole URI (if the "whole" parameter
2061 is present) and divides the hash value by the total weight of
2062 the running servers. The result designates which server will
2063 receive the request. This ensures that the same URI will
2064 always be directed to the same server as long as no server
2065 goes up or down. This is used with proxy caches and
2066 anti-virus proxies in order to maximize the cache hit rate.
2067 Note that this algorithm may only be used in an HTTP backend.
2068 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".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002071
Oskar Stolc8dc41842012-05-19 10:19:54 +01002072 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002073 "depth", both followed by a positive integer number. These
2074 options may be helpful when it is needed to balance servers
2075 based on the beginning of the URI only. The "len" parameter
2076 indicates that the algorithm should only consider that many
2077 characters at the beginning of the URI to compute the hash.
2078 Note that having "len" set to 1 rarely makes sense since most
2079 URIs start with a leading "/".
2080
2081 The "depth" parameter indicates the maximum directory depth
2082 to be used to compute the hash. One level is counted for each
2083 slash in the request. If both parameters are specified, the
2084 evaluation stops when either is reached.
2085
Willy Tarreau0ba27502007-12-24 16:55:16 +01002086 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002087 the query string of each HTTP GET request.
2088
2089 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02002090 request entity will be searched for the parameter argument,
2091 when it is not found in a query string after a question mark
Willy Tarreau226071e2014-04-10 11:55:45 +02002092 ('?') in the URL. The message body will only start to be
2093 analyzed once either the advertised amount of data has been
2094 received or the request buffer is full. In the unlikely event
2095 that chunked encoding is used, only the first chunk is
Cyril Bontédc4d9032012-04-08 21:57:39 +02002096 scanned. Parameter values separated by a chunk boundary, may
Willy Tarreau226071e2014-04-10 11:55:45 +02002097 be randomly balanced if at all. This keyword used to support
2098 an optional <max_wait> parameter which is now ignored.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002099
2100 If the parameter is found followed by an equal sign ('=') and
2101 a value, then the value is hashed and divided by the total
2102 weight of the running servers. The result designates which
2103 server will receive the request.
2104
2105 This is used to track user identifiers in requests and ensure
2106 that a same user ID will always be sent to the same server as
2107 long as no server goes up or down. If no value is found or if
2108 the parameter is not found, then a round robin algorithm is
2109 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002110 backend. This algorithm is static by default, which means
2111 that changing a server's weight on the fly will have no
2112 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002113
Cyril Bontédc4d9032012-04-08 21:57:39 +02002114 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
2115 request. Just as with the equivalent ACL 'hdr()' function,
2116 the header name in parenthesis is not case sensitive. If the
2117 header is absent or if it does not contain any value, the
2118 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01002119
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002120 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01002121 reducing the hash algorithm to the main domain part with some
2122 specific headers such as 'Host'. For instance, in the Host
2123 value "haproxy.1wt.eu", only "1wt" will be considered.
2124
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002125 This algorithm is static by default, which means that
2126 changing a server's weight on the fly will have no effect,
2127 but this can be changed using "hash-type".
2128
Emeric Brun736aa232009-06-30 17:56:00 +02002129 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02002130 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02002131 The RDP cookie <name> (or "mstshash" if omitted) will be
2132 looked up and hashed for each incoming TCP request. Just as
2133 with the equivalent ACL 'req_rdp_cookie()' function, the name
2134 is not case-sensitive. This mechanism is useful as a degraded
2135 persistence mode, as it makes it possible to always send the
2136 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002137 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02002138 used instead.
2139
2140 Note that for this to work, the frontend must ensure that an
2141 RDP cookie is already present in the request buffer. For this
2142 you must use 'tcp-request content accept' rule combined with
2143 a 'req_rdp_cookie_cnt' ACL.
2144
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002145 This algorithm is static by default, which means that
2146 changing a server's weight on the fly will have no effect,
2147 but this can be changed using "hash-type".
2148
Cyril Bontédc4d9032012-04-08 21:57:39 +02002149 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09002150
Willy Tarreau0ba27502007-12-24 16:55:16 +01002151 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002152 algorithms. Right now, only "url_param" and "uri" support an
2153 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002154
Willy Tarreau3cd9af22009-03-15 14:06:41 +01002155 The load balancing algorithm of a backend is set to roundrobin when no other
2156 algorithm, mode nor option have been set. The algorithm may only be set once
2157 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002158
2159 Examples :
2160 balance roundrobin
2161 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002162 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01002163 balance hdr(User-Agent)
2164 balance hdr(host)
2165 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002166
2167 Note: the following caveats and limitations on using the "check_post"
2168 extension with "url_param" must be considered :
2169
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002170 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002171 to determine if the parameters will be found in the body or entity which
2172 may contain binary data. Therefore another method may be required to
2173 restrict consideration of POST requests that have no URL parameters in
2174 the body. (see acl reqideny http_end)
2175
2176 - using a <max_wait> value larger than the request buffer size does not
2177 make sense and is useless. The buffer size is set at build time, and
2178 defaults to 16 kB.
2179
2180 - Content-Encoding is not supported, the parameter search will probably
2181 fail; and load balancing will fall back to Round Robin.
2182
2183 - Expect: 100-continue is not supported, load balancing will fall back to
2184 Round Robin.
2185
2186 - Transfer-Encoding (RFC2616 3.6.1) is only supported in the first chunk.
2187 If the entire parameter value is not present in the first chunk, the
2188 selection of server is undefined (actually, defined by how little
2189 actually appeared in the first chunk).
2190
2191 - This feature does not support generation of a 100, 411 or 501 response.
2192
2193 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002194 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002195 white space or control characters are found, indicating the end of what
2196 might be a URL parameter list. This is probably not a concern with SGML
2197 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002198
Willy Tarreau294d0f02015-08-10 19:40:12 +02002199 See also : "dispatch", "cookie", "transparent", "hash-type" and "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002200
2201
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002202bind [<address>]:<port_range> [, ...] [param*]
2203bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002204 Define one or several listening addresses and/or ports in a frontend.
2205 May be used in sections : defaults | frontend | listen | backend
2206 no | yes | yes | no
2207 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002208 <address> is optional and can be a host name, an IPv4 address, an IPv6
2209 address, or '*'. It designates the address the frontend will
2210 listen on. If unset, all IPv4 addresses of the system will be
2211 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01002212 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01002213 Optionally, an address family prefix may be used before the
2214 address to force the family regardless of the address format,
2215 which can be useful to specify a path to a unix socket with
2216 no slash ('/'). Currently supported prefixes are :
2217 - 'ipv4@' -> address is always IPv4
2218 - 'ipv6@' -> address is always IPv6
2219 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau70f72e02014-07-08 00:37:50 +02002220 - 'abns@' -> address is in abstract namespace (Linux only).
2221 Note: since abstract sockets are not "rebindable", they
2222 do not cope well with multi-process mode during
2223 soft-restart, so it is better to avoid them if
2224 nbproc is greater than 1. The effect is that if the
2225 new process fails to start, only one of the old ones
2226 will be able to rebind to the socket.
Willy Tarreau40aa0702013-03-10 23:51:38 +01002227 - 'fd@<n>' -> use file descriptor <n> inherited from the
2228 parent. The fd must be bound and may or may not already
2229 be listening.
William Lallemandb2f07452015-05-12 14:27:13 +02002230 You may want to reference some environment variables in the
2231 address parameter, see section 2.3 about environment
2232 variables.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002233
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002234 <port_range> is either a unique TCP port, or a port range for which the
2235 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002236 above. The port is mandatory for TCP listeners. Note that in
2237 the case of an IPv6 address, the port is always the number
2238 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002239 - a numerical port (ex: '80')
2240 - a dash-delimited ports range explicitly stating the lower
2241 and upper bounds (ex: '2000-2100') which are included in
2242 the range.
2243
2244 Particular care must be taken against port ranges, because
2245 every <address:port> couple consumes one socket (= a file
2246 descriptor), so it's easy to consume lots of descriptors
2247 with a simple range, and to run out of sockets. Also, each
2248 <address:port> couple must be used only once among all
2249 instances running on a same system. Please note that binding
2250 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002251 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002252 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002253
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002254 <path> is a UNIX socket path beginning with a slash ('/'). This is
2255 alternative to the TCP listening port. Haproxy will then
2256 receive UNIX connections on the socket located at this place.
2257 The path must begin with a slash and by default is absolute.
2258 It can be relative to the prefix defined by "unix-bind" in
2259 the global section. Note that the total length of the prefix
2260 followed by the socket path cannot exceed some system limits
2261 for UNIX sockets, which commonly are set to 107 characters.
2262
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002263 <param*> is a list of parameters common to all sockets declared on the
2264 same line. These numerous parameters depend on OS and build
2265 options and have a complete section dedicated to them. Please
2266 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002267
Willy Tarreau0ba27502007-12-24 16:55:16 +01002268 It is possible to specify a list of address:port combinations delimited by
2269 commas. The frontend will then listen on all of these addresses. There is no
2270 fixed limit to the number of addresses and ports which can be listened on in
2271 a frontend, as well as there is no limit to the number of "bind" statements
2272 in a frontend.
2273
2274 Example :
2275 listen http_proxy
2276 bind :80,:443
2277 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002278 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01002279
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002280 listen http_https_proxy
2281 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02002282 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002283
Willy Tarreau24709282013-03-10 21:32:12 +01002284 listen http_https_proxy_explicit
2285 bind ipv6@:80
2286 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
2287 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
2288
Willy Tarreaudad36a32013-03-11 01:20:04 +01002289 listen external_bind_app1
William Lallemandb2f07452015-05-12 14:27:13 +02002290 bind "fd@${FD_APP1}"
Willy Tarreaudad36a32013-03-11 01:20:04 +01002291
Willy Tarreau55dcaf62015-09-27 15:03:15 +02002292 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
2293 sun_path length is used for the address length. Some other programs
2294 such as socat use the string length only by default. Pass the option
2295 ",unix-tightsocklen=0" to any abstract socket definition in socat to
2296 make it compatible with HAProxy's.
2297
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002298 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002299 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002300
2301
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002302bind-process [ all | odd | even | <number 1-64>[-<number 1-64>] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002303 Limit visibility of an instance to a certain set of processes numbers.
2304 May be used in sections : defaults | frontend | listen | backend
2305 yes | yes | yes | yes
2306 Arguments :
2307 all All process will see this instance. This is the default. It
2308 may be used to override a default value.
2309
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002310 odd This instance will be enabled on processes 1,3,5,...63. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002311 option may be combined with other numbers.
2312
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002313 even This instance will be enabled on processes 2,4,6,...64. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002314 option may be combined with other numbers. Do not use it
2315 with less than 2 processes otherwise some instances might be
2316 missing from all processes.
2317
Willy Tarreau110ecc12012-11-15 17:50:01 +01002318 number The instance will be enabled on this process number or range,
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002319 whose values must all be between 1 and 32 or 64 depending on
Willy Tarreau102df612014-05-07 23:56:38 +02002320 the machine's word size. If a proxy is bound to process
2321 numbers greater than the configured global.nbproc, it will
2322 either be forced to process #1 if a single process was
2323 specified, or to all processes otherwise.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002324
2325 This keyword limits binding of certain instances to certain processes. This
2326 is useful in order not to have too many processes listening to the same
2327 ports. For instance, on a dual-core machine, it might make sense to set
2328 'nbproc 2' in the global section, then distributes the listeners among 'odd'
2329 and 'even' instances.
2330
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002331 At the moment, it is not possible to reference more than 32 or 64 processes
2332 using this keyword, but this should be more than enough for most setups.
2333 Please note that 'all' really means all processes regardless of the machine's
2334 word size, and is not limited to the first 32 or 64.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002335
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002336 Each "bind" line may further be limited to a subset of the proxy's processes,
2337 please consult the "process" bind keyword in section 5.1.
2338
Willy Tarreaub369a042014-09-16 13:21:03 +02002339 When a frontend has no explicit "bind-process" line, it tries to bind to all
2340 the processes referenced by its "bind" lines. That means that frontends can
2341 easily adapt to their listeners' processes.
2342
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002343 If some backends are referenced by frontends bound to other processes, the
2344 backend automatically inherits the frontend's processes.
2345
2346 Example :
2347 listen app_ip1
2348 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002349 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002350
2351 listen app_ip2
2352 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002353 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002354
2355 listen management
2356 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002357 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002358
Willy Tarreau110ecc12012-11-15 17:50:01 +01002359 listen management
2360 bind 10.0.0.4:80
2361 bind-process 1-4
2362
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002363 See also : "nbproc" in global section, and "process" in section 5.1.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002364
2365
Willy Tarreau0ba27502007-12-24 16:55:16 +01002366block { if | unless } <condition>
2367 Block a layer 7 request if/unless a condition is matched
2368 May be used in sections : defaults | frontend | listen | backend
2369 no | yes | yes | yes
2370
2371 The HTTP request will be blocked very early in the layer 7 processing
2372 if/unless <condition> is matched. A 403 error will be returned if the request
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002373 is blocked. The condition has to reference ACLs (see section 7). This is
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02002374 typically used to deny access to certain sensitive resources if some
Willy Tarreau0ba27502007-12-24 16:55:16 +01002375 conditions are met or not met. There is no fixed limit to the number of
2376 "block" statements per instance.
2377
2378 Example:
2379 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2380 acl invalid_src src_port 0:1023
2381 acl local_dst hdr(host) -i localhost
2382 block if invalid_src || local_dst
2383
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002384 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002385
2386
2387capture cookie <name> len <length>
2388 Capture and log a cookie in the request and in the response.
2389 May be used in sections : defaults | frontend | listen | backend
2390 no | yes | yes | no
2391 Arguments :
2392 <name> is the beginning of the name of the cookie to capture. In order
2393 to match the exact name, simply suffix the name with an equal
2394 sign ('='). The full name will appear in the logs, which is
2395 useful with application servers which adjust both the cookie name
2396 and value (eg: ASPSESSIONXXXXX).
2397
2398 <length> is the maximum number of characters to report in the logs, which
2399 include the cookie name, the equal sign and the value, all in the
2400 standard "name=value" form. The string will be truncated on the
2401 right if it exceeds <length>.
2402
2403 Only the first cookie is captured. Both the "cookie" request headers and the
2404 "set-cookie" response headers are monitored. This is particularly useful to
2405 check for application bugs causing session crossing or stealing between
2406 users, because generally the user's cookies can only change on a login page.
2407
2408 When the cookie was not presented by the client, the associated log column
2409 will report "-". When a request does not cause a cookie to be assigned by the
2410 server, a "-" is reported in the response column.
2411
2412 The capture is performed in the frontend only because it is necessary that
2413 the log format does not change for a given frontend depending on the
2414 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01002415 "capture cookie" statement in a frontend. The maximum capture length is set
2416 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
2417 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002418
2419 Example:
2420 capture cookie ASPSESSION len 32
2421
2422 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002423 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002424
2425
2426capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01002427 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002428 May be used in sections : defaults | frontend | listen | backend
2429 no | yes | yes | no
2430 Arguments :
2431 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002432 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01002433 appear in the requests, with the first letter of each word in
2434 upper case. The header name will not appear in the logs, only the
2435 value is reported, but the position in the logs is respected.
2436
2437 <length> is the maximum number of characters to extract from the value and
2438 report in the logs. The string will be truncated on the right if
2439 it exceeds <length>.
2440
Willy Tarreau4460d032012-11-21 23:37:37 +01002441 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01002442 value will be added to the logs between braces ('{}'). If multiple headers
2443 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002444 in the same order they were declared in the configuration. Non-existent
2445 headers will be logged just as an empty string. Common uses for request
2446 header captures include the "Host" field in virtual hosting environments, the
2447 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002448 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002449 environments to find where the request came from.
2450
2451 Note that when capturing headers such as "User-agent", some spaces may be
2452 logged, making the log analysis more difficult. Thus be careful about what
2453 you log if you know your log parser is not smart enough to rely on the
2454 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002455
Willy Tarreau0900abb2012-11-22 00:21:46 +01002456 There is no limit to the number of captured request headers nor to their
2457 length, though it is wise to keep them low to limit memory usage per session.
2458 In order to keep log format consistent for a same frontend, header captures
2459 can only be declared in a frontend. It is not possible to specify a capture
2460 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002461
2462 Example:
2463 capture request header Host len 15
2464 capture request header X-Forwarded-For len 15
Cyril Bontéd1b0f7c2015-10-26 22:37:39 +01002465 capture request header Referer len 15
Willy Tarreau0ba27502007-12-24 16:55:16 +01002466
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002467 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01002468 about logging.
2469
2470
2471capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01002472 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002473 May be used in sections : defaults | frontend | listen | backend
2474 no | yes | yes | no
2475 Arguments :
2476 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002477 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01002478 appear in the response, with the first letter of each word in
2479 upper case. The header name will not appear in the logs, only the
2480 value is reported, but the position in the logs is respected.
2481
2482 <length> is the maximum number of characters to extract from the value and
2483 report in the logs. The string will be truncated on the right if
2484 it exceeds <length>.
2485
Willy Tarreau4460d032012-11-21 23:37:37 +01002486 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01002487 result will be added to the logs between braces ('{}') after the captured
2488 request headers. If multiple headers are captured, they will be delimited by
2489 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002490 the configuration. Non-existent headers will be logged just as an empty
2491 string. Common uses for response header captures include the "Content-length"
2492 header which indicates how many bytes are expected to be returned, the
2493 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002494
Willy Tarreau0900abb2012-11-22 00:21:46 +01002495 There is no limit to the number of captured response headers nor to their
2496 length, though it is wise to keep them low to limit memory usage per session.
2497 In order to keep log format consistent for a same frontend, header captures
2498 can only be declared in a frontend. It is not possible to specify a capture
2499 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002500
2501 Example:
2502 capture response header Content-length len 9
2503 capture response header Location len 15
2504
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002505 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01002506 about logging.
2507
2508
Cyril Bontéf0c60612010-02-06 14:44:47 +01002509clitimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002510 Set the maximum inactivity time on the client side.
2511 May be used in sections : defaults | frontend | listen | backend
2512 yes | yes | yes | no
2513 Arguments :
2514 <timeout> is the timeout value is specified in milliseconds by default, but
2515 can be in any other unit if the number is suffixed by the unit,
2516 as explained at the top of this document.
2517
2518 The inactivity timeout applies when the client is expected to acknowledge or
2519 send data. In HTTP mode, this timeout is particularly important to consider
2520 during the first phase, when the client sends the request, and during the
2521 response while it is reading data sent by the server. The value is specified
2522 in milliseconds by default, but can be in any other unit if the number is
2523 suffixed by the unit, as specified at the top of this document. In TCP mode
2524 (and to a lesser extent, in HTTP mode), it is highly recommended that the
2525 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002526 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01002527 losses by specifying timeouts that are slightly above multiples of 3 seconds
2528 (eg: 4 or 5 seconds).
2529
2530 This parameter is specific to frontends, but can be specified once for all in
2531 "defaults" sections. This is in fact one of the easiest solutions not to
2532 forget about it. An unspecified timeout results in an infinite timeout, which
2533 is not recommended. Such a usage is accepted and works but reports a warning
2534 during startup because it may results in accumulation of expired sessions in
2535 the system if the system's timeouts are not configured either.
2536
2537 This parameter is provided for compatibility but is currently deprecated.
2538 Please use "timeout client" instead.
2539
Willy Tarreau036fae02008-01-06 13:24:40 +01002540 See also : "timeout client", "timeout http-request", "timeout server", and
2541 "srvtimeout".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002542
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002543compression algo <algorithm> ...
2544compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02002545compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02002546 Enable HTTP compression.
2547 May be used in sections : defaults | frontend | listen | backend
2548 yes | yes | yes | yes
2549 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002550 algo is followed by the list of supported compression algorithms.
2551 type is followed by the list of MIME types that will be compressed.
2552 offload makes haproxy work as a compression offloader only (see notes).
2553
2554 The currently supported algorithms are :
Willy Tarreauc91840a2015-03-28 17:00:39 +01002555 identity this is mostly for debugging, and it was useful for developing
2556 the compression feature. Identity does not apply any change on
2557 data.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002558
Willy Tarreauc91840a2015-03-28 17:00:39 +01002559 gzip applies gzip compression. This setting is only available when
Baptiste Assmannf085d632015-12-21 17:57:32 +01002560 support for zlib or libslz was built in.
Willy Tarreauc91840a2015-03-28 17:00:39 +01002561
2562 deflate same as "gzip", but with deflate algorithm and zlib format.
2563 Note that this algorithm has ambiguous support on many
2564 browsers and no support at all from recent ones. It is
2565 strongly recommended not to use it for anything else than
2566 experimentation. This setting is only available when support
Baptiste Assmannf085d632015-12-21 17:57:32 +01002567 for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002568
Willy Tarreauc91840a2015-03-28 17:00:39 +01002569 raw-deflate same as "deflate" without the zlib wrapper, and used as an
2570 alternative when the browser wants "deflate". All major
2571 browsers understand it and despite violating the standards,
2572 it is known to work better than "deflate", at least on MSIE
2573 and some versions of Safari. Do not use it in conjunction
2574 with "deflate", use either one or the other since both react
2575 to the same Accept-Encoding token. This setting is only
Baptiste Assmannf085d632015-12-21 17:57:32 +01002576 available when support for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002577
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04002578 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002579 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04002580 If backend servers support HTTP compression, these directives
2581 will be no-op: haproxy will see the compressed response and will not
2582 compress again. If backend servers do not support HTTP compression and
2583 there is Accept-Encoding header in request, haproxy will compress the
2584 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02002585
2586 The "offload" setting makes haproxy remove the Accept-Encoding header to
2587 prevent backend servers from compressing responses. It is strongly
2588 recommended not to do this because this means that all the compression work
2589 will be done on the single point where haproxy is located. However in some
2590 deployment scenarios, haproxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04002591 with broken HTTP compression implementation which can't be turned off.
2592 In that case haproxy can be used to prevent that gateway from emitting
2593 invalid payloads. In this case, simply removing the header in the
2594 configuration does not work because it applies before the header is parsed,
2595 so that prevents haproxy from compressing. The "offload" setting should
Willy Tarreauffea9fd2014-07-12 16:37:02 +02002596 then be used for such scenarios. Note: for now, the "offload" setting is
2597 ignored when set in a defaults section.
William Lallemand82fe75c2012-10-23 10:25:10 +02002598
William Lallemand05097442012-11-20 12:14:28 +01002599 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002600 * the request does not advertise a supported compression algorithm in the
2601 "Accept-Encoding" header
2602 * the response message is not HTTP/1.1
William Lallemandd3002612012-11-26 14:34:47 +01002603 * HTTP status code is not 200
William Lallemand8bb4e342013-12-10 17:28:48 +01002604 * response header "Transfer-Encoding" contains "chunked" (Temporary
2605 Workaround)
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002606 * response contain neither a "Content-Length" header nor a
2607 "Transfer-Encoding" whose last value is "chunked"
2608 * response contains a "Content-Type" header whose first value starts with
2609 "multipart"
2610 * the response contains the "no-transform" value in the "Cache-control"
2611 header
2612 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
2613 and later
2614 * The response contains a "Content-Encoding" header, indicating that the
2615 response is already compressed (see compression offload)
William Lallemand05097442012-11-20 12:14:28 +01002616
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002617 Note: The compression does not rewrite Etag headers, and does not emit the
2618 Warning header.
William Lallemand05097442012-11-20 12:14:28 +01002619
William Lallemand82fe75c2012-10-23 10:25:10 +02002620 Examples :
2621 compression algo gzip
2622 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01002623
Christopher Fauletc3fe5332016-04-07 15:30:10 +02002624
Cyril Bontéf0c60612010-02-06 14:44:47 +01002625contimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002626 Set the maximum time to wait for a connection attempt to a server to succeed.
2627 May be used in sections : defaults | frontend | listen | backend
2628 yes | no | yes | yes
2629 Arguments :
2630 <timeout> is the timeout value is specified in milliseconds by default, but
2631 can be in any other unit if the number is suffixed by the unit,
2632 as explained at the top of this document.
2633
2634 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002635 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01002636 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01002637 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
2638 connect timeout also presets the queue timeout to the same value if this one
2639 has not been specified. Historically, the contimeout was also used to set the
2640 tarpit timeout in a listen section, which is not possible in a pure frontend.
2641
2642 This parameter is specific to backends, but can be specified once for all in
2643 "defaults" sections. This is in fact one of the easiest solutions not to
2644 forget about it. An unspecified timeout results in an infinite timeout, which
2645 is not recommended. Such a usage is accepted and works but reports a warning
2646 during startup because it may results in accumulation of failed sessions in
2647 the system if the system's timeouts are not configured either.
2648
2649 This parameter is provided for backwards compatibility but is currently
2650 deprecated. Please use "timeout connect", "timeout queue" or "timeout tarpit"
2651 instead.
2652
2653 See also : "timeout connect", "timeout queue", "timeout tarpit",
2654 "timeout server", "contimeout".
2655
2656
Willy Tarreau55165fe2009-05-10 12:02:55 +02002657cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02002658 [ postonly ] [ preserve ] [ httponly ] [ secure ]
2659 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002660 Enable cookie-based persistence in a backend.
2661 May be used in sections : defaults | frontend | listen | backend
2662 yes | no | yes | yes
2663 Arguments :
2664 <name> is the name of the cookie which will be monitored, modified or
2665 inserted in order to bring persistence. This cookie is sent to
2666 the client via a "Set-Cookie" header in the response, and is
2667 brought back by the client in a "Cookie" header in all requests.
2668 Special care should be taken to choose a name which does not
2669 conflict with any likely application cookie. Also, if the same
2670 backends are subject to be used by the same clients (eg:
2671 HTTP/HTTPS), care should be taken to use different cookie names
2672 between all backends if persistence between them is not desired.
2673
2674 rewrite This keyword indicates that the cookie will be provided by the
2675 server and that haproxy will have to modify its value to set the
2676 server's identifier in it. This mode is handy when the management
2677 of complex combinations of "Set-cookie" and "Cache-control"
2678 headers is left to the application. The application can then
2679 decide whether or not it is appropriate to emit a persistence
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01002680 cookie. Since all responses should be monitored, this mode
2681 doesn't work in HTTP tunnel mode. Unless the application
2682 behaviour is very complex and/or broken, it is advised not to
2683 start with this mode for new deployments. This keyword is
2684 incompatible with "insert" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002685
2686 insert This keyword indicates that the persistence cookie will have to
Willy Tarreaua79094d2010-08-31 22:54:15 +02002687 be inserted by haproxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002688
Willy Tarreaua79094d2010-08-31 22:54:15 +02002689 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002690 server. When used without the "preserve" option, if the server
2691 emits a cookie with the same name, it will be remove before
2692 processing. For this reason, this mode can be used to upgrade
2693 existing configurations running in the "rewrite" mode. The cookie
2694 will only be a session cookie and will not be stored on the
2695 client's disk. By default, unless the "indirect" option is added,
2696 the server will see the cookies emitted by the client. Due to
2697 caching effects, it is generally wise to add the "nocache" or
2698 "postonly" keywords (see below). The "insert" keyword is not
2699 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002700
2701 prefix This keyword indicates that instead of relying on a dedicated
2702 cookie for the persistence, an existing one will be completed.
2703 This may be needed in some specific environments where the client
2704 does not support more than one single cookie and the application
2705 already needs it. In this case, whenever the server sets a cookie
2706 named <name>, it will be prefixed with the server's identifier
2707 and a delimiter. The prefix will be removed from all client
2708 requests so that the server still finds the cookie it emitted.
2709 Since all requests and responses are subject to being modified,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01002710 this mode doesn't work with tunnel mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02002711 not compatible with "rewrite" and "insert". Note: it is highly
2712 recommended not to use "indirect" with "prefix", otherwise server
2713 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002714
Willy Tarreaua79094d2010-08-31 22:54:15 +02002715 indirect When this option is specified, no cookie will be emitted to a
2716 client which already has a valid one for the server which has
2717 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002718 it will be removed, unless the "preserve" option is also set. In
2719 "insert" mode, this will additionally remove cookies from the
2720 requests transmitted to the server, making the persistence
2721 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02002722 Note: it is highly recommended not to use "indirect" with
2723 "prefix", otherwise server cookie updates would not be sent to
2724 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002725
2726 nocache This option is recommended in conjunction with the insert mode
2727 when there is a cache between the client and HAProxy, as it
2728 ensures that a cacheable response will be tagged non-cacheable if
2729 a cookie needs to be inserted. This is important because if all
2730 persistence cookies are added on a cacheable home page for
2731 instance, then all customers will then fetch the page from an
2732 outer cache and will all share the same persistence cookie,
2733 leading to one server receiving much more traffic than others.
2734 See also the "insert" and "postonly" options.
2735
2736 postonly This option ensures that cookie insertion will only be performed
2737 on responses to POST requests. It is an alternative to the
2738 "nocache" option, because POST responses are not cacheable, so
2739 this ensures that the persistence cookie will never get cached.
2740 Since most sites do not need any sort of persistence before the
2741 first POST which generally is a login request, this is a very
2742 efficient method to optimize caching without risking to find a
2743 persistence cookie in the cache.
2744 See also the "insert" and "nocache" options.
2745
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002746 preserve This option may only be used with "insert" and/or "indirect". It
2747 allows the server to emit the persistence cookie itself. In this
2748 case, if a cookie is found in the response, haproxy will leave it
2749 untouched. This is useful in order to end persistence after a
2750 logout request for instance. For this, the server just has to
2751 emit a cookie with an invalid value (eg: empty) or with a date in
2752 the past. By combining this mechanism with the "disable-on-404"
2753 check option, it is possible to perform a completely graceful
2754 shutdown because users will definitely leave the server after
2755 they logout.
2756
Willy Tarreau4992dd22012-05-31 21:02:17 +02002757 httponly This option tells haproxy to add an "HttpOnly" cookie attribute
2758 when a cookie is inserted. This attribute is used so that a
2759 user agent doesn't share the cookie with non-HTTP components.
2760 Please check RFC6265 for more information on this attribute.
2761
2762 secure This option tells haproxy to add a "Secure" cookie attribute when
2763 a cookie is inserted. This attribute is used so that a user agent
2764 never emits this cookie over non-secure channels, which means
2765 that a cookie learned with this flag will be presented only over
2766 SSL/TLS connections. Please check RFC6265 for more information on
2767 this attribute.
2768
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002769 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002770 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01002771 name. If the domain begins with a dot, the browser is allowed to
2772 use it for any host ending with that name. It is also possible to
2773 specify several domain names by invoking this option multiple
2774 times. Some browsers might have small limits on the number of
2775 domains, so be careful when doing that. For the record, sending
2776 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002777
Willy Tarreau996a92c2010-10-13 19:30:47 +02002778 maxidle This option allows inserted cookies to be ignored after some idle
2779 time. It only works with insert-mode cookies. When a cookie is
2780 sent to the client, the date this cookie was emitted is sent too.
2781 Upon further presentations of this cookie, if the date is older
2782 than the delay indicated by the parameter (in seconds), it will
2783 be ignored. Otherwise, it will be refreshed if needed when the
2784 response is sent to the client. This is particularly useful to
2785 prevent users who never close their browsers from remaining for
2786 too long on the same server (eg: after a farm size change). When
2787 this option is set and a cookie has no date, it is always
2788 accepted, but gets refreshed in the response. This maintains the
2789 ability for admins to access their sites. Cookies that have a
2790 date in the future further than 24 hours are ignored. Doing so
2791 lets admins fix timezone issues without risking kicking users off
2792 the site.
2793
2794 maxlife This option allows inserted cookies to be ignored after some life
2795 time, whether they're in use or not. It only works with insert
2796 mode cookies. When a cookie is first sent to the client, the date
2797 this cookie was emitted is sent too. Upon further presentations
2798 of this cookie, if the date is older than the delay indicated by
2799 the parameter (in seconds), it will be ignored. If the cookie in
2800 the request has no date, it is accepted and a date will be set.
2801 Cookies that have a date in the future further than 24 hours are
2802 ignored. Doing so lets admins fix timezone issues without risking
2803 kicking users off the site. Contrary to maxidle, this value is
2804 not refreshed, only the first visit date counts. Both maxidle and
2805 maxlife may be used at the time. This is particularly useful to
2806 prevent users who never close their browsers from remaining for
2807 too long on the same server (eg: after a farm size change). This
2808 is stronger than the maxidle method in that it forces a
2809 redispatch after some absolute delay.
2810
Willy Tarreau0ba27502007-12-24 16:55:16 +01002811 There can be only one persistence cookie per HTTP backend, and it can be
2812 declared in a defaults section. The value of the cookie will be the value
2813 indicated after the "cookie" keyword in a "server" statement. If no cookie
2814 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002815
Willy Tarreau0ba27502007-12-24 16:55:16 +01002816 Examples :
2817 cookie JSESSIONID prefix
2818 cookie SRV insert indirect nocache
2819 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02002820 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01002821
Willy Tarreau294d0f02015-08-10 19:40:12 +02002822 See also : "balance source", "capture cookie", "server" and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002823
Willy Tarreau983e01e2010-01-11 18:42:06 +01002824
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02002825declare capture [ request | response ] len <length>
2826 Declares a capture slot.
2827 May be used in sections : defaults | frontend | listen | backend
2828 no | yes | yes | no
2829 Arguments:
2830 <length> is the length allowed for the capture.
2831
2832 This declaration is only available in the frontend or listen section, but the
2833 reserved slot can be used in the backends. The "request" keyword allocates a
2834 capture slot for use in the request, and "response" allocates a capture slot
2835 for use in the response.
2836
2837 See also: "capture-req", "capture-res" (sample converters),
Baptiste Assmann5ac425c2015-10-21 23:13:46 +02002838 "capture.req.hdr", "capture.res.hdr" (sample fetches),
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02002839 "http-request capture" and "http-response capture".
2840
2841
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002842default-server [param*]
2843 Change default options for a server in a backend
2844 May be used in sections : defaults | frontend | listen | backend
2845 yes | no | yes | yes
2846 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01002847 <param*> is a list of parameters for this server. The "default-server"
2848 keyword accepts an important number of options and has a complete
2849 section dedicated to it. Please refer to section 5 for more
2850 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002851
Willy Tarreau983e01e2010-01-11 18:42:06 +01002852 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002853 default-server inter 1000 weight 13
2854
2855 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01002856
Willy Tarreau983e01e2010-01-11 18:42:06 +01002857
Willy Tarreau0ba27502007-12-24 16:55:16 +01002858default_backend <backend>
2859 Specify the backend to use when no "use_backend" rule has been matched.
2860 May be used in sections : defaults | frontend | listen | backend
2861 yes | yes | yes | no
2862 Arguments :
2863 <backend> is the name of the backend to use.
2864
2865 When doing content-switching between frontend and backends using the
2866 "use_backend" keyword, it is often useful to indicate which backend will be
2867 used when no rule has matched. It generally is the dynamic backend which
2868 will catch all undetermined requests.
2869
Willy Tarreau0ba27502007-12-24 16:55:16 +01002870 Example :
2871
2872 use_backend dynamic if url_dyn
2873 use_backend static if url_css url_img extension_img
2874 default_backend dynamic
2875
Willy Tarreau98d04852015-05-26 12:18:29 +02002876 See also : "use_backend"
Willy Tarreau2769aa02007-12-27 18:26:09 +01002877
Willy Tarreau0ba27502007-12-24 16:55:16 +01002878
Baptiste Assmann27f51342013-10-09 06:51:49 +02002879description <string>
2880 Describe a listen, frontend or backend.
2881 May be used in sections : defaults | frontend | listen | backend
2882 no | yes | yes | yes
2883 Arguments : string
2884
2885 Allows to add a sentence to describe the related object in the HAProxy HTML
2886 stats page. The description will be printed on the right of the object name
2887 it describes.
2888 No need to backslash spaces in the <string> arguments.
2889
2890
Willy Tarreau0ba27502007-12-24 16:55:16 +01002891disabled
2892 Disable a proxy, frontend or backend.
2893 May be used in sections : defaults | frontend | listen | backend
2894 yes | yes | yes | yes
2895 Arguments : none
2896
2897 The "disabled" keyword is used to disable an instance, mainly in order to
2898 liberate a listening port or to temporarily disable a service. The instance
2899 will still be created and its configuration will be checked, but it will be
2900 created in the "stopped" state and will appear as such in the statistics. It
2901 will not receive any traffic nor will it send any health-checks or logs. It
2902 is possible to disable many instances at once by adding the "disabled"
2903 keyword in a "defaults" section.
2904
2905 See also : "enabled"
2906
2907
Willy Tarreau5ce94572010-06-07 14:35:41 +02002908dispatch <address>:<port>
2909 Set a default server address
2910 May be used in sections : defaults | frontend | listen | backend
2911 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02002912 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02002913
2914 <address> is the IPv4 address of the default server. Alternatively, a
2915 resolvable hostname is supported, but this name will be resolved
2916 during start-up.
2917
2918 <ports> is a mandatory port specification. All connections will be sent
2919 to this port, and it is not permitted to use port offsets as is
2920 possible with normal servers.
2921
Willy Tarreau787aed52011-04-15 06:45:37 +02002922 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02002923 server can take the connection. In the past it was used to forward non
2924 persistent connections to an auxiliary load balancer. Due to its simple
2925 syntax, it has also been used for simple TCP relays. It is recommended not to
2926 use it for more clarity, and to use the "server" directive instead.
2927
2928 See also : "server"
2929
2930
Willy Tarreau0ba27502007-12-24 16:55:16 +01002931enabled
2932 Enable a proxy, frontend or backend.
2933 May be used in sections : defaults | frontend | listen | backend
2934 yes | yes | yes | yes
2935 Arguments : none
2936
2937 The "enabled" keyword is used to explicitly enable an instance, when the
2938 defaults has been set to "disabled". This is very rarely used.
2939
2940 See also : "disabled"
2941
2942
2943errorfile <code> <file>
2944 Return a file contents instead of errors generated by HAProxy
2945 May be used in sections : defaults | frontend | listen | backend
2946 yes | yes | yes | yes
2947 Arguments :
2948 <code> is the HTTP status code. Currently, HAProxy is capable of
CJ Ess108b1dd2015-04-07 12:03:37 -04002949 generating codes 200, 400, 403, 405, 408, 429, 500, 502, 503, and
2950 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002951
2952 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002953 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01002954 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01002955 error pages, and to use absolute paths, since files are read
2956 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002957
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 Tarreau0ba27502007-12-24 16:55:16 +01002964 The files are returned verbatim on the TCP socket. This allows any trick such
2965 as redirections to another URL or site, as well as tricks to clean cookies,
2966 force enable or disable caching, etc... The package provides default error
2967 files returning the same contents as default errors.
2968
Willy Tarreau59140a22009-02-22 12:02:30 +01002969 The files should not exceed the configured buffer size (BUFSIZE), which
2970 generally is 8 or 16 kB, otherwise they will be truncated. It is also wise
2971 not to put any reference to local contents (eg: images) in order to avoid
2972 loops between the client and HAProxy when all servers are down, causing an
2973 error to be returned instead of an image. For better HTTP compliance, it is
2974 recommended that all header lines end with CR-LF and not LF alone.
2975
Willy Tarreau0ba27502007-12-24 16:55:16 +01002976 The files are read at the same time as the configuration and kept in memory.
2977 For this reason, the errors continue to be returned even when the process is
2978 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01002979 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01002980 403 status code and interrogating a blocked URL.
2981
2982 See also : "errorloc", "errorloc302", "errorloc303"
2983
Willy Tarreau59140a22009-02-22 12:02:30 +01002984 Example :
2985 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
Willy Tarreau989222a2016-01-15 10:26:26 +01002986 errorfile 408 /dev/null # work around Chrome pre-connect bug
Willy Tarreau59140a22009-02-22 12:02:30 +01002987 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
2988 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
2989
Willy Tarreau2769aa02007-12-27 18:26:09 +01002990
2991errorloc <code> <url>
2992errorloc302 <code> <url>
2993 Return an HTTP redirection to a URL instead of errors generated by HAProxy
2994 May be used in sections : defaults | frontend | listen | backend
2995 yes | yes | yes | yes
2996 Arguments :
2997 <code> is the HTTP status code. Currently, HAProxy is capable of
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002998 generating codes 200, 400, 403, 408, 500, 502, 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002999
3000 <url> it is the exact contents of the "Location" header. It may contain
3001 either a relative URI to an error page hosted on the same site,
3002 or an absolute URI designating an error page on another site.
3003 Special care should be given to relative URIs to avoid redirect
3004 loops if the URI itself may generate the same error (eg: 500).
3005
3006 It is important to understand that this keyword is not meant to rewrite
3007 errors returned by the server, but errors detected and returned by HAProxy.
3008 This is why the list of supported errors is limited to a small set.
3009
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003010 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3011
Willy Tarreau2769aa02007-12-27 18:26:09 +01003012 Note that both keyword return the HTTP 302 status code, which tells the
3013 client to fetch the designated URL using the same HTTP method. This can be
3014 quite problematic in case of non-GET methods such as POST, because the URL
3015 sent to the client might not be allowed for something other than GET. To
Willy Tarreau989222a2016-01-15 10:26:26 +01003016 work around this problem, please use "errorloc303" which send the HTTP 303
Willy Tarreau2769aa02007-12-27 18:26:09 +01003017 status code, indicating to the client that the URL must be fetched with a GET
3018 request.
3019
3020 See also : "errorfile", "errorloc303"
3021
3022
3023errorloc303 <code> <url>
3024 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3025 May be used in sections : defaults | frontend | listen | backend
3026 yes | yes | yes | yes
3027 Arguments :
3028 <code> is the HTTP status code. Currently, HAProxy is capable of
3029 generating codes 400, 403, 408, 500, 502, 503, and 504.
3030
3031 <url> it is the exact contents of the "Location" header. It may contain
3032 either a relative URI to an error page hosted on the same site,
3033 or an absolute URI designating an error page on another site.
3034 Special care should be given to relative URIs to avoid redirect
3035 loops if the URI itself may generate the same error (eg: 500).
3036
3037 It is important to understand that this keyword is not meant to rewrite
3038 errors returned by the server, but errors detected and returned by HAProxy.
3039 This is why the list of supported errors is limited to a small set.
3040
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003041 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3042
Willy Tarreau2769aa02007-12-27 18:26:09 +01003043 Note that both keyword return the HTTP 303 status code, which tells the
3044 client to fetch the designated URL using the same HTTP GET method. This
3045 solves the usual problems associated with "errorloc" and the 302 code. It is
3046 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003047 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003048
3049 See also : "errorfile", "errorloc", "errorloc302"
3050
3051
Simon Horman51a1cf62015-02-03 13:00:44 +09003052email-alert from <emailaddr>
3053 Declare the from email address to be used in both the envelope and header
3054 of email alerts. This is the address that email alerts are sent from.
3055 May be used in sections: defaults | frontend | listen | backend
3056 yes | yes | yes | yes
3057
3058 Arguments :
3059
3060 <emailaddr> is the from email address to use when sending email alerts
3061
3062 Also requires "email-alert mailers" and "email-alert to" to be set
3063 and if so sending email alerts is enabled for the proxy.
3064
Simon Horman64e34162015-02-06 11:11:57 +09003065 See also : "email-alert level", "email-alert mailers",
Cyril Bonté307ee1e2015-09-28 23:16:06 +02003066 "email-alert myhostname", "email-alert to", section 3.6 about
3067 mailers.
Simon Horman64e34162015-02-06 11:11:57 +09003068
3069
3070email-alert level <level>
3071 Declare the maximum log level of messages for which email alerts will be
3072 sent. This acts as a filter on the sending of email alerts.
3073 May be used in sections: defaults | frontend | listen | backend
3074 yes | yes | yes | yes
3075
3076 Arguments :
3077
3078 <level> One of the 8 syslog levels:
3079 emerg alert crit err warning notice info debug
3080 The above syslog levels are ordered from lowest to highest.
3081
3082 By default level is alert
3083
3084 Also requires "email-alert from", "email-alert mailers" and
3085 "email-alert to" to be set and if so sending email alerts is enabled
3086 for the proxy.
3087
Simon Horman1421e212015-04-30 13:10:35 +09003088 Alerts are sent when :
3089
3090 * An un-paused server is marked as down and <level> is alert or lower
3091 * A paused server is marked as down and <level> is notice or lower
3092 * A server is marked as up or enters the drain state and <level>
3093 is notice or lower
3094 * "option log-health-checks" is enabled, <level> is info or lower,
3095 and a health check status update occurs
3096
Simon Horman64e34162015-02-06 11:11:57 +09003097 See also : "email-alert from", "email-alert mailers",
3098 "email-alert myhostname", "email-alert to",
Simon Horman51a1cf62015-02-03 13:00:44 +09003099 section 3.6 about mailers.
3100
3101
3102email-alert mailers <mailersect>
3103 Declare the mailers to be used when sending email alerts
3104 May be used in sections: defaults | frontend | listen | backend
3105 yes | yes | yes | yes
3106
3107 Arguments :
3108
3109 <mailersect> is the name of the mailers section to send email alerts.
3110
3111 Also requires "email-alert from" and "email-alert to" to be set
3112 and if so sending email alerts is enabled for the proxy.
3113
Simon Horman64e34162015-02-06 11:11:57 +09003114 See also : "email-alert from", "email-alert level", "email-alert myhostname",
3115 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003116
3117
3118email-alert myhostname <hostname>
3119 Declare the to hostname address to be used when communicating with
3120 mailers.
3121 May be used in sections: defaults | frontend | listen | backend
3122 yes | yes | yes | yes
3123
3124 Arguments :
3125
Baptiste Assmann738bad92015-12-21 15:27:53 +01003126 <hostname> is the hostname to use when communicating with mailers
Simon Horman51a1cf62015-02-03 13:00:44 +09003127
3128 By default the systems hostname is used.
3129
3130 Also requires "email-alert from", "email-alert mailers" and
3131 "email-alert to" to be set and if so sending email alerts is enabled
3132 for the proxy.
3133
Simon Horman64e34162015-02-06 11:11:57 +09003134 See also : "email-alert from", "email-alert level", "email-alert mailers",
3135 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003136
3137
3138email-alert to <emailaddr>
3139 Declare both the recipent address in the envelope and to address in the
3140 header of email alerts. This is the address that email alerts are sent to.
3141 May be used in sections: defaults | frontend | listen | backend
3142 yes | yes | yes | yes
3143
3144 Arguments :
3145
3146 <emailaddr> is the to email address to use when sending email alerts
3147
3148 Also requires "email-alert mailers" and "email-alert to" to be set
3149 and if so sending email alerts is enabled for the proxy.
3150
Simon Horman64e34162015-02-06 11:11:57 +09003151 See also : "email-alert from", "email-alert level", "email-alert mailers",
Simon Horman51a1cf62015-02-03 13:00:44 +09003152 "email-alert myhostname", section 3.6 about mailers.
3153
3154
Willy Tarreau4de91492010-01-22 19:10:05 +01003155force-persist { if | unless } <condition>
3156 Declare a condition to force persistence on down servers
3157 May be used in sections: defaults | frontend | listen | backend
3158 no | yes | yes | yes
3159
3160 By default, requests are not dispatched to down servers. It is possible to
3161 force this using "option persist", but it is unconditional and redispatches
3162 to a valid server if "option redispatch" is set. That leaves with very little
3163 possibilities to force some requests to reach a server which is artificially
3164 marked down for maintenance operations.
3165
3166 The "force-persist" statement allows one to declare various ACL-based
3167 conditions which, when met, will cause a request to ignore the down status of
3168 a server and still try to connect to it. That makes it possible to start a
3169 server, still replying an error to the health checks, and run a specially
3170 configured browser to test the service. Among the handy methods, one could
3171 use a specific source IP address, or a specific cookie. The cookie also has
3172 the advantage that it can easily be added/removed on the browser from a test
3173 page. Once the service is validated, it is then possible to open the service
3174 to the world by returning a valid response to health checks.
3175
3176 The forced persistence is enabled when an "if" condition is met, or unless an
3177 "unless" condition is met. The final redispatch is always disabled when this
3178 is used.
3179
Cyril Bonté0d4bf012010-04-25 23:21:46 +02003180 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02003181 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01003182
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003183
3184filter <name> [param*]
3185 Add the filter <name> in the filter list attached to the proxy.
3186 May be used in sections : defaults | frontend | listen | backend
3187 no | yes | yes | yes
3188 Arguments :
3189 <name> is the name of the filter. Officially supported filters are
3190 referenced in section 9.
3191
3192 <param*> is a list of parameters accpeted by the filter <name>. The
3193 parsing of these parameters are the responsibility of the
3194 filter. Please refer to the documention of the corresponding
3195 filter (section 9) from all details on the supported parameters.
3196
3197 Multiple occurrences of the filter line can be used for the same proxy. The
3198 same filter can be referenced many times if needed.
3199
3200 Example:
3201 listen
3202 bind *:80
3203
3204 filter trace name BEFORE-HTTP-COMP
3205 filter compression
3206 filter trace name AFTER-HTTP-COMP
3207
3208 compression algo gzip
3209 compression offload
3210
3211 server srv1 192.168.0.1:80
3212
3213 See also : section 9.
3214
Willy Tarreau4de91492010-01-22 19:10:05 +01003215
Willy Tarreau2769aa02007-12-27 18:26:09 +01003216fullconn <conns>
3217 Specify at what backend load the servers will reach their maxconn
3218 May be used in sections : defaults | frontend | listen | backend
3219 yes | no | yes | yes
3220 Arguments :
3221 <conns> is the number of connections on the backend which will make the
3222 servers use the maximal number of connections.
3223
Willy Tarreau198a7442008-01-17 12:05:32 +01003224 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01003225 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01003226 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01003227 load. The server will then always accept at least <minconn> connections,
3228 never more than <maxconn>, and the limit will be on the ramp between both
3229 values when the backend has less than <conns> concurrent connections. This
3230 makes it possible to limit the load on the servers during normal loads, but
3231 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003232 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003233
Willy Tarreaufbb78422011-06-05 15:38:35 +02003234 Since it's hard to get this value right, haproxy automatically sets it to
3235 10% of the sum of the maxconns of all frontends that may branch to this
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01003236 backend (based on "use_backend" and "default_backend" rules). That way it's
3237 safe to leave it unset. However, "use_backend" involving dynamic names are
3238 not counted since there is no way to know if they could match or not.
Willy Tarreaufbb78422011-06-05 15:38:35 +02003239
Willy Tarreau2769aa02007-12-27 18:26:09 +01003240 Example :
3241 # The servers will accept between 100 and 1000 concurrent connections each
3242 # and the maximum of 1000 will be reached when the backend reaches 10000
3243 # connections.
3244 backend dynamic
3245 fullconn 10000
3246 server srv1 dyn1:80 minconn 100 maxconn 1000
3247 server srv2 dyn2:80 minconn 100 maxconn 1000
3248
3249 See also : "maxconn", "server"
3250
3251
3252grace <time>
3253 Maintain a proxy operational for some time after a soft stop
3254 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01003255 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01003256 Arguments :
3257 <time> is the time (by default in milliseconds) for which the instance
3258 will remain operational with the frontend sockets still listening
3259 when a soft-stop is received via the SIGUSR1 signal.
3260
3261 This may be used to ensure that the services disappear in a certain order.
3262 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003263 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01003264 needed by the equipment to detect the failure.
3265
3266 Note that currently, there is very little benefit in using this parameter,
3267 and it may in fact complicate the soft-reconfiguration process more than
3268 simplify it.
3269
Willy Tarreau0ba27502007-12-24 16:55:16 +01003270
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003271hash-type <method> <function> <modifier>
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003272 Specify a method to use for mapping hashes to servers
3273 May be used in sections : defaults | frontend | listen | backend
3274 yes | no | yes | yes
3275 Arguments :
Bhaskar98634f02013-10-29 23:30:51 -04003276 <method> is the method used to select a server from the hash computed by
3277 the <function> :
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003278
Bhaskar98634f02013-10-29 23:30:51 -04003279 map-based the hash table is a static array containing all alive servers.
3280 The hashes will be very smooth, will consider weights, but
3281 will be static in that weight changes while a server is up
3282 will be ignored. This means that there will be no slow start.
3283 Also, since a server is selected by its position in the array,
3284 most mappings are changed when the server count changes. This
3285 means that when a server goes up or down, or when a server is
3286 added to a farm, most connections will be redistributed to
3287 different servers. This can be inconvenient with caches for
3288 instance.
Willy Tarreau798a39c2010-11-24 15:04:29 +01003289
Bhaskar98634f02013-10-29 23:30:51 -04003290 consistent the hash table is a tree filled with many occurrences of each
3291 server. The hash key is looked up in the tree and the closest
3292 server is chosen. This hash is dynamic, it supports changing
3293 weights while the servers are up, so it is compatible with the
3294 slow start feature. It has the advantage that when a server
3295 goes up or down, only its associations are moved. When a
3296 server is added to the farm, only a few part of the mappings
3297 are redistributed, making it an ideal method for caches.
3298 However, due to its principle, the distribution will never be
3299 very smooth and it may sometimes be necessary to adjust a
3300 server's weight or its ID to get a more balanced distribution.
3301 In order to get the same distribution on multiple load
3302 balancers, it is important that all servers have the exact
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003303 same IDs. Note: consistent hash uses sdbm and avalanche if no
3304 hash function is specified.
Bhaskar98634f02013-10-29 23:30:51 -04003305
3306 <function> is the hash function to be used :
3307
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003308 sdbm this function was created initially for sdbm (a public-domain
Bhaskar98634f02013-10-29 23:30:51 -04003309 reimplementation of ndbm) database library. It was found to do
3310 well in scrambling bits, causing better distribution of the keys
3311 and fewer splits. It also happens to be a good general hashing
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003312 function with good distribution, unless the total server weight
3313 is a multiple of 64, in which case applying the avalanche
3314 modifier may help.
Bhaskar98634f02013-10-29 23:30:51 -04003315
3316 djb2 this function was first proposed by Dan Bernstein many years ago
3317 on comp.lang.c. Studies have shown that for certain workload this
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003318 function provides a better distribution than sdbm. It generally
3319 works well with text-based inputs though it can perform extremely
3320 poorly with numeric-only input or when the total server weight is
3321 a multiple of 33, unless the avalanche modifier is also used.
3322
Willy Tarreaua0f42712013-11-14 14:30:35 +01003323 wt6 this function was designed for haproxy while testing other
3324 functions in the past. It is not as smooth as the other ones, but
3325 is much less sensible to the input data set or to the number of
3326 servers. It can make sense as an alternative to sdbm+avalanche or
3327 djb2+avalanche for consistent hashing or when hashing on numeric
3328 data such as a source IP address or a visitor identifier in a URL
3329 parameter.
3330
Willy Tarreau324f07f2015-01-20 19:44:50 +01003331 crc32 this is the most common CRC32 implementation as used in Ethernet,
3332 gzip, PNG, etc. It is slower than the other ones but may provide
3333 a better distribution or less predictable results especially when
3334 used on strings.
3335
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003336 <modifier> indicates an optional method applied after hashing the key :
3337
3338 avalanche This directive indicates that the result from the hash
3339 function above should not be used in its raw form but that
3340 a 4-byte full avalanche hash must be applied first. The
3341 purpose of this step is to mix the resulting bits from the
3342 previous hash in order to avoid any undesired effect when
3343 the input contains some limited values or when the number of
3344 servers is a multiple of one of the hash's components (64
3345 for SDBM, 33 for DJB2). Enabling avalanche tends to make the
3346 result less predictable, but it's also not as smooth as when
3347 using the original function. Some testing might be needed
3348 with some workloads. This hash is one of the many proposed
3349 by Bob Jenkins.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003350
Bhaskar98634f02013-10-29 23:30:51 -04003351 The default hash type is "map-based" and is recommended for most usages. The
3352 default function is "sdbm", the selection of a function should be based on
3353 the range of the values being hashed.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003354
3355 See also : "balance", "server"
3356
3357
Willy Tarreau0ba27502007-12-24 16:55:16 +01003358http-check disable-on-404
3359 Enable a maintenance mode upon HTTP/404 response to health-checks
3360 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01003361 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01003362 Arguments : none
3363
3364 When this option is set, a server which returns an HTTP code 404 will be
3365 excluded from further load-balancing, but will still receive persistent
3366 connections. This provides a very convenient method for Web administrators
3367 to perform a graceful shutdown of their servers. It is also important to note
3368 that a server which is detected as failed while it was in this mode will not
3369 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
3370 will immediately be reinserted into the farm. The status on the stats page
3371 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01003372 option only works in conjunction with the "httpchk" option. If this option
3373 is used with "http-check expect", then it has precedence over it so that 404
3374 responses will still be considered as soft-stop.
3375
3376 See also : "option httpchk", "http-check expect"
3377
3378
3379http-check expect [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003380 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01003381 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02003382 yes | no | yes | yes
Willy Tarreaubd741542010-03-16 18:46:54 +01003383 Arguments :
3384 <match> is a keyword indicating how to look for a specific pattern in the
3385 response. The keyword may be one of "status", "rstatus",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003386 "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01003387 exclamation mark ("!") to negate the match. Spaces are allowed
3388 between the exclamation mark and the keyword. See below for more
3389 details on the supported keywords.
3390
3391 <pattern> is the pattern to look for. It may be a string or a regular
3392 expression. If the pattern contains spaces, they must be escaped
3393 with the usual backslash ('\').
3394
3395 By default, "option httpchk" considers that response statuses 2xx and 3xx
3396 are valid, and that others are invalid. When "http-check expect" is used,
3397 it defines what is considered valid or invalid. Only one "http-check"
3398 statement is supported in a backend. If a server fails to respond or times
3399 out, the check obviously fails. The available matches are :
3400
3401 status <string> : test the exact string match for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003402 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003403 response's status code is exactly this string. If the
3404 "status" keyword is prefixed with "!", then the response
3405 will be considered invalid if the status code matches.
3406
3407 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003408 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003409 response's status code matches the expression. If the
3410 "rstatus" keyword is prefixed with "!", then the response
3411 will be considered invalid if the status code matches.
3412 This is mostly used to check for multiple codes.
3413
3414 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003415 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003416 response's body contains this exact string. If the
3417 "string" keyword is prefixed with "!", then the response
3418 will be considered invalid if the body contains this
3419 string. This can be used to look for a mandatory word at
3420 the end of a dynamic page, or to detect a failure when a
3421 specific error appears on the check page (eg: a stack
3422 trace).
3423
3424 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003425 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003426 response's body matches this expression. If the "rstring"
3427 keyword is prefixed with "!", then the response will be
3428 considered invalid if the body matches the expression.
3429 This can be used to look for a mandatory word at the end
3430 of a dynamic page, or to detect a failure when a specific
3431 error appears on the check page (eg: a stack trace).
3432
3433 It is important to note that the responses will be limited to a certain size
3434 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
3435 Thus, too large responses may not contain the mandatory pattern when using
3436 "string" or "rstring". If a large response is absolutely required, it is
3437 possible to change the default max size by setting the global variable.
3438 However, it is worth keeping in mind that parsing very large responses can
3439 waste some CPU cycles, especially when regular expressions are used, and that
3440 it is always better to focus the checks on smaller resources.
3441
Cyril Bonté32602d22015-01-30 00:07:07 +01003442 Also "http-check expect" doesn't support HTTP keep-alive. Keep in mind that it
3443 will automatically append a "Connection: close" header, meaning that this
3444 header should not be present in the request provided by "option httpchk".
3445
Willy Tarreaubd741542010-03-16 18:46:54 +01003446 Last, if "http-check expect" is combined with "http-check disable-on-404",
3447 then this last one has precedence when the server responds with 404.
3448
3449 Examples :
3450 # only accept status 200 as valid
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003451 http-check expect status 200
Willy Tarreaubd741542010-03-16 18:46:54 +01003452
3453 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003454 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01003455
3456 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003457 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01003458
3459 # check that we have a correct hexadecimal tag before /html
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003460 http-check expect rstring <!--tag:[0-9a-f]*</html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01003461
Willy Tarreaubd741542010-03-16 18:46:54 +01003462 See also : "option httpchk", "http-check disable-on-404"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003463
3464
Willy Tarreauef781042010-01-27 11:53:01 +01003465http-check send-state
3466 Enable emission of a state header with HTTP health checks
3467 May be used in sections : defaults | frontend | listen | backend
3468 yes | no | yes | yes
3469 Arguments : none
3470
3471 When this option is set, haproxy will systematically send a special header
3472 "X-Haproxy-Server-State" with a list of parameters indicating to each server
3473 how they are seen by haproxy. This can be used for instance when a server is
3474 manipulated without access to haproxy and the operator needs to know whether
3475 haproxy still sees it up or not, or if the server is the last one in a farm.
3476
3477 The header is composed of fields delimited by semi-colons, the first of which
3478 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
3479 checks on the total number before transition, just as appears in the stats
3480 interface. Next headers are in the form "<variable>=<value>", indicating in
3481 no specific order some values available in the stats interface :
Joseph Lynch514061c2015-01-15 17:52:59 -08003482 - a variable "address", containing the address of the backend server.
3483 This corresponds to the <address> field in the server declaration. For
3484 unix domain sockets, it will read "unix".
3485
3486 - a variable "port", containing the port of the backend server. This
3487 corresponds to the <port> field in the server declaration. For unix
3488 domain sockets, it will read "unix".
3489
Willy Tarreauef781042010-01-27 11:53:01 +01003490 - a variable "name", containing the name of the backend followed by a slash
3491 ("/") then the name of the server. This can be used when a server is
3492 checked in multiple backends.
3493
3494 - a variable "node" containing the name of the haproxy node, as set in the
3495 global "node" variable, otherwise the system's hostname if unspecified.
3496
3497 - a variable "weight" indicating the weight of the server, a slash ("/")
3498 and the total weight of the farm (just counting usable servers). This
3499 helps to know if other servers are available to handle the load when this
3500 one fails.
3501
3502 - a variable "scur" indicating the current number of concurrent connections
3503 on the server, followed by a slash ("/") then the total number of
3504 connections on all servers of the same backend.
3505
3506 - a variable "qcur" indicating the current number of requests in the
3507 server's queue.
3508
3509 Example of a header received by the application server :
3510 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
3511 scur=13/22; qcur=0
3512
3513 See also : "option httpchk", "http-check disable-on-404"
3514
Willy Tarreaube1d34d2016-06-26 19:37:59 +02003515http-request { allow | tarpit | auth [realm <realm>] | redirect <rule> |
3516 deny [deny_status <status>] |
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003517 add-header <name> <fmt> | set-header <name> <fmt> |
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003518 capture <sample> [ len <length> | id <id> ] |
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003519 del-header <name> | set-nice <nice> | set-log-level <level> |
Sasha Pachev218f0642014-06-16 12:05:59 -06003520 replace-header <name> <match-regex> <replace-fmt> |
3521 replace-value <name> <match-regex> <replace-fmt> |
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01003522 set-method <fmt> | set-path <fmt> | set-query <fmt> |
3523 set-uri <fmt> | set-tos <tos> | set-mark <mark> |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003524 add-acl(<file name>) <key fmt> |
3525 del-acl(<file name>) <key fmt> |
3526 del-map(<file name>) <key fmt> |
Baptiste Assmannbb7e86a2014-09-03 18:29:47 +02003527 set-map(<file name>) <key fmt> <value fmt> |
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003528 set-var(<var name>) <expr> |
Thierry FOURNIER90da1912015-03-05 11:17:06 +01003529 { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] |
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02003530 sc-inc-gpc0(<sc-id>) |
Thierry FOURNIER236657b2015-08-19 08:25:14 +02003531 sc-set-gpt0(<sc-id>) <int> |
Willy Tarreau2d392c22015-08-24 01:43:45 +02003532 silent-drop |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003533 }
Cyril Bontéf0c60612010-02-06 14:44:47 +01003534 [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003535 Access control for Layer 7 requests
3536
3537 May be used in sections: defaults | frontend | listen | backend
3538 no | yes | yes | yes
3539
Willy Tarreau20b0de52012-12-24 15:45:22 +01003540 The http-request statement defines a set of rules which apply to layer 7
3541 processing. The rules are evaluated in their declaration order when they are
3542 met in a frontend, listen or backend section. Any rule may optionally be
3543 followed by an ACL-based condition, in which case it will only be evaluated
3544 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003545
Willy Tarreau20b0de52012-12-24 15:45:22 +01003546 The first keyword is the rule's action. Currently supported actions include :
3547 - "allow" : this stops the evaluation of the rules and lets the request
3548 pass the check. No further "http-request" rules are evaluated.
3549
3550 - "deny" : this stops the evaluation of the rules and immediately rejects
Willy Tarreaube1d34d2016-06-26 19:37:59 +02003551 the request and emits an HTTP 403 error, or optionally the status code
3552 specified as an argument to "deny_status". The list of permitted status
3553 codes is limited to those that can be overridden by the "errorfile"
3554 directive. No further "http-request" rules are evaluated.
Willy Tarreau20b0de52012-12-24 15:45:22 +01003555
Willy Tarreauccbcc372012-12-27 12:37:57 +01003556 - "tarpit" : this stops the evaluation of the rules and immediately blocks
3557 the request without responding for a delay specified by "timeout tarpit"
3558 or "timeout connect" if the former is not set. After that delay, if the
3559 client is still connected, an HTTP error 500 is returned so that the
3560 client does not suspect it has been tarpitted. Logs will report the flags
3561 "PT". The goal of the tarpit rule is to slow down robots during an attack
3562 when they're limited on the number of concurrent requests. It can be very
3563 efficient against very dumb robots, and will significantly reduce the
3564 load on firewalls compared to a "deny" rule. But when facing "correctly"
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003565 developed robots, it can make things worse by forcing haproxy and the
Willy Tarreau2d392c22015-08-24 01:43:45 +02003566 front firewall to support insane number of concurrent connections. See
3567 also the "silent-drop" action below.
Willy Tarreauccbcc372012-12-27 12:37:57 +01003568
Willy Tarreau20b0de52012-12-24 15:45:22 +01003569 - "auth" : this stops the evaluation of the rules and immediately responds
3570 with an HTTP 401 or 407 error code to invite the user to present a valid
3571 user name and password. No further "http-request" rules are evaluated. An
3572 optional "realm" parameter is supported, it sets the authentication realm
3573 that is returned with the response (typically the application's name).
3574
Willy Tarreau81499eb2012-12-27 12:19:02 +01003575 - "redirect" : this performs an HTTP redirection based on a redirect rule.
3576 This is exactly the same as the "redirect" statement except that it
3577 inserts a redirect rule which can be processed in the middle of other
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01003578 "http-request" rules and that these rules use the "log-format" strings.
3579 See the "redirect" keyword for the rule's syntax.
Willy Tarreau81499eb2012-12-27 12:19:02 +01003580
Willy Tarreau20b0de52012-12-24 15:45:22 +01003581 - "add-header" appends an HTTP header field whose name is specified in
3582 <name> and whose value is defined by <fmt> which follows the log-format
3583 rules (see Custom Log Format in section 8.2.4). This is particularly
3584 useful to pass connection-specific information to the server (eg: the
3585 client's SSL certificate), or to combine several headers into one. This
3586 rule is not final, so it is possible to add other similar rules. Note
3587 that header addition is performed immediately, so one rule might reuse
3588 the resulting header from a previous rule.
3589
3590 - "set-header" does the same as "add-header" except that the header name
3591 is first removed if it existed. This is useful when passing security
3592 information to the server, where the header must not be manipulated by
Willy Tarreau85603282015-01-21 20:39:27 +01003593 external users. Note that the new value is computed before the removal so
3594 it is possible to concatenate a value to an existing header.
Willy Tarreau20b0de52012-12-24 15:45:22 +01003595
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003596 - "del-header" removes all HTTP header fields whose name is specified in
3597 <name>.
3598
Sasha Pachev218f0642014-06-16 12:05:59 -06003599 - "replace-header" matches the regular expression in all occurrences of
3600 header field <name> according to <match-regex>, and replaces them with
3601 the <replace-fmt> argument. Format characters are allowed in replace-fmt
3602 and work like in <fmt> arguments in "add-header". The match is only
3603 case-sensitive. It is important to understand that this action only
3604 considers whole header lines, regardless of the number of values they
3605 may contain. This usage is suited to headers naturally containing commas
3606 in their value, such as If-Modified-Since and so on.
3607
3608 Example:
3609
3610 http-request replace-header Cookie foo=([^;]*);(.*) foo=\1;ip=%bi;\2
3611
3612 applied to:
3613
3614 Cookie: foo=foobar; expires=Tue, 14-Jun-2016 01:40:45 GMT;
3615
3616 outputs:
3617
3618 Cookie: foo=foobar;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT;
3619
3620 assuming the backend IP is 192.168.1.20
3621
3622 - "replace-value" works like "replace-header" except that it matches the
3623 regex against every comma-delimited value of the header field <name>
3624 instead of the entire header. This is suited for all headers which are
3625 allowed to carry more than one value. An example could be the Accept
3626 header.
3627
3628 Example:
3629
3630 http-request replace-value X-Forwarded-For ^192\.168\.(.*)$ 172.16.\1
3631
3632 applied to:
3633
3634 X-Forwarded-For: 192.168.10.1, 192.168.13.24, 10.0.0.37
3635
3636 outputs:
3637
3638 X-Forwarded-For: 172.16.10.1, 172.16.13.24, 10.0.0.37
3639
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01003640 - "set-method" rewrites the request method with the result of the
3641 evaluation of format string <fmt>. There should be very few valid reasons
3642 for having to do so as this is more likely to break something than to fix
3643 it.
3644
3645 - "set-path" rewrites the request path with the result of the evaluation of
3646 format string <fmt>. The query string, if any, is left intact. If a
3647 scheme and authority is found before the path, they are left intact as
3648 well. If the request doesn't have a path ("*"), this one is replaced with
3649 the format. This can be used to prepend a directory component in front of
3650 a path for example. See also "set-query" and "set-uri".
3651
3652 Example :
3653 # prepend the host name before the path
3654 http-request set-path /%[hdr(host)]%[path]
3655
3656 - "set-query" rewrites the request's query string which appears after the
3657 first question mark ("?") with the result of the evaluation of format
3658 string <fmt>. The part prior to the question mark is left intact. If the
3659 request doesn't contain a question mark and the new value is not empty,
3660 then one is added at the end of the URI, followed by the new value. If
3661 a question mark was present, it will never be removed even if the value
3662 is empty. This can be used to add or remove parameters from the query
3663 string. See also "set-query" and "set-uri".
3664
3665 Example :
3666 # replace "%3D" with "=" in the query string
3667 http-request set-query %[query,regsub(%3D,=,g)]
3668
3669 - "set-uri" rewrites the request URI with the result of the evaluation of
3670 format string <fmt>. The scheme, authority, path and query string are all
3671 replaced at once. This can be used to rewrite hosts in front of proxies,
3672 or to perform complex modifications to the URI such as moving parts
3673 between the path and the query string. See also "set-path" and
3674 "set-query".
3675
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003676 - "set-nice" sets the "nice" factor of the current request being processed.
3677 It only has effect against the other requests being processed at the same
3678 time. The default value is 0, unless altered by the "nice" setting on the
3679 "bind" line. The accepted range is -1024..1024. The higher the value, the
3680 nicest the request will be. Lower values will make the request more
3681 important than other ones. This can be useful to improve the speed of
3682 some requests, or lower the priority of non-important requests. Using
3683 this setting without prior experimentation can cause some major slowdown.
3684
Willy Tarreau9a355ec2013-06-11 17:45:46 +02003685 - "set-log-level" is used to change the log level of the current request
3686 when a certain condition is met. Valid levels are the 8 syslog levels
3687 (see the "log" keyword) plus the special level "silent" which disables
3688 logging for this request. This rule is not final so the last matching
3689 rule wins. This rule can be useful to disable health checks coming from
3690 another equipment.
3691
Willy Tarreau42cf39e2013-06-11 18:51:32 +02003692 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
3693 the client to the value passed in <tos> on platforms which support this.
3694 This value represents the whole 8 bits of the IP TOS field, and can be
3695 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
3696 that only the 6 higher bits are used in DSCP or TOS, and the two lower
3697 bits are always 0. This can be used to adjust some routing behaviour on
3698 border routers based on some information from the request. See RFC 2474,
3699 2597, 3260 and 4594 for more information.
3700
Willy Tarreau51347ed2013-06-11 19:34:13 +02003701 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
3702 client to the value passed in <mark> on platforms which support it. This
3703 value is an unsigned 32 bit value which can be matched by netfilter and
3704 by the routing table. It can be expressed both in decimal or hexadecimal
3705 format (prefixed by "0x"). This can be useful to force certain packets to
3706 take a different route (for example a cheaper network path for bulk
3707 downloads). This works on Linux kernels 2.6.32 and above and requires
3708 admin privileges.
3709
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003710 - "add-acl" is used to add a new entry into an ACL. The ACL must be loaded
3711 from a file (even a dummy empty file). The file name of the ACL to be
3712 updated is passed between parentheses. It takes one argument: <key fmt>,
3713 which follows log-format rules, to collect content of the new entry. It
3714 performs a lookup in the ACL before insertion, to avoid duplicated (or
3715 more) values. This lookup is done by a linear search and can be expensive
3716 with large lists! It is the equivalent of the "add acl" command from the
3717 stats socket, but can be triggered by an HTTP request.
3718
3719 - "del-acl" is used to delete an entry from an ACL. The ACL must be loaded
3720 from a file (even a dummy empty file). The file name of the ACL to be
3721 updated is passed between parentheses. It takes one argument: <key fmt>,
3722 which follows log-format rules, to collect content of the entry to delete.
3723 It is the equivalent of the "del acl" command from the stats socket, but
3724 can be triggered by an HTTP request.
3725
3726 - "del-map" is used to delete an entry from a MAP. The MAP must be loaded
3727 from a file (even a dummy empty file). The file name of the MAP to be
3728 updated is passed between parentheses. It takes one argument: <key fmt>,
3729 which follows log-format rules, to collect content of the entry to delete.
3730 It takes one argument: "file name" It is the equivalent of the "del map"
3731 command from the stats socket, but can be triggered by an HTTP request.
3732
3733 - "set-map" is used to add a new entry into a MAP. The MAP must be loaded
3734 from a file (even a dummy empty file). The file name of the MAP to be
3735 updated is passed between parentheses. It takes 2 arguments: <key fmt>,
3736 which follows log-format rules, used to collect MAP key, and <value fmt>,
3737 which follows log-format rules, used to collect content for the new entry.
3738 It performs a lookup in the MAP before insertion, to avoid duplicated (or
3739 more) values. This lookup is done by a linear search and can be expensive
3740 with large lists! It is the equivalent of the "set map" command from the
3741 stats socket, but can be triggered by an HTTP request.
3742
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003743 - capture <sample> [ len <length> | id <id> ] :
Willy Tarreaua9083d02015-05-08 15:27:59 +02003744 captures sample expression <sample> from the request buffer, and converts
3745 it to a string of at most <len> characters. The resulting string is
3746 stored into the next request "capture" slot, so it will possibly appear
3747 next to some captured HTTP headers. It will then automatically appear in
3748 the logs, and it will be possible to extract it using sample fetch rules
3749 to feed it into headers or anything. The length should be limited given
3750 that this size will be allocated for each capture during the whole
3751 session life. Please check section 7.3 (Fetching samples) and "capture
3752 request header" for more information.
3753
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003754 If the keyword "id" is used instead of "len", the action tries to store
3755 the captured string in a previously declared capture slot. This is useful
3756 to run captures in backends. The slot id can be declared by a previous
3757 directive "http-request capture" or with the "declare capture" keyword.
Baptiste Assmanne9544932015-11-03 23:31:35 +01003758 If the slot <id> doesn't exist, then HAProxy fails parsing the
3759 configuration to prevent unexpected behavior at run time.
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003760
Willy Tarreau09448f72014-06-25 18:12:15 +02003761 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
3762 enables tracking of sticky counters from current request. These rules
3763 do not stop evaluation and do not change default action. Three sets of
3764 counters may be simultaneously tracked by the same connection. The first
3765 "track-sc0" rule executed enables tracking of the counters of the
3766 specified table as the first set. The first "track-sc1" rule executed
3767 enables tracking of the counters of the specified table as the second
3768 set. The first "track-sc2" rule executed enables tracking of the
3769 counters of the specified table as the third set. It is a recommended
3770 practice to use the first set of counters for the per-frontend counters
3771 and the second set for the per-backend ones. But this is just a
3772 guideline, all may be used everywhere.
3773
3774 These actions take one or two arguments :
3775 <key> is mandatory, and is a sample expression rule as described
3776 in section 7.3. It describes what elements of the incoming
3777 request or connection will be analysed, extracted, combined,
3778 and used to select which table entry to update the counters.
3779
3780 <table> is an optional table to be used instead of the default one,
3781 which is the stick-table declared in the current proxy. All
3782 the counters for the matches and updates for the key will
3783 then be performed in that table until the session ends.
3784
3785 Once a "track-sc*" rule is executed, the key is looked up in the table
3786 and if it is not found, an entry is allocated for it. Then a pointer to
3787 that entry is kept during all the session's life, and this entry's
3788 counters are updated as often as possible, every time the session's
3789 counters are updated, and also systematically when the session ends.
3790 Counters are only updated for events that happen after the tracking has
3791 been started. As an exception, connection counters and request counters
3792 are systematically updated so that they reflect useful information.
3793
3794 If the entry tracks concurrent connection counters, one connection is
3795 counted for as long as the entry is tracked, and the entry will not
3796 expire during that time. Tracking counters also provides a performance
3797 advantage over just checking the keys, because only one table lookup is
3798 performed for all ACL checks that make use of it.
3799
Thierry FOURNIER236657b2015-08-19 08:25:14 +02003800 - sc-set-gpt0(<sc-id>) <int> :
3801 This action sets the GPT0 tag according to the sticky counter designated
3802 by <sc-id> and the value of <int>. The expected result is a boolean. If
3803 an error occurs, this action silently fails and the actions evaluation
3804 continues.
3805
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02003806 - sc-inc-gpc0(<sc-id>):
3807 This action increments the GPC0 counter according with the sticky counter
3808 designated by <sc-id>. If an error occurs, this action silently fails and
3809 the actions evaluation continues.
3810
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003811 - set-var(<var-name>) <expr> :
3812 Is used to set the contents of a variable. The variable is declared
3813 inline.
3814
Daniel Schneller0b547052016-03-21 20:46:57 +01003815 <var-name> The name of the variable starts with an indication about
3816 its scope. The scopes allowed are:
3817 "sess" : the variable is shared with the whole session
3818 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003819 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +01003820 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003821 processing
Daniel Schneller0b547052016-03-21 20:46:57 +01003822 "res" : the variable is shared only during response
3823 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003824 This prefix is followed by a name. The separator is a '.'.
Daniel Schneller0b547052016-03-21 20:46:57 +01003825 The name may only contain characters 'a-z', 'A-Z', '0-9'
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003826 and '_'.
3827
3828 <expr> Is a standard HAProxy expression formed by a sample-fetch
3829 followed by some converters.
3830
3831 Example:
3832
3833 http-request set-var(req.my_var) req.fhdr(user-agent),lower
3834
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02003835 - set-src <expr> :
3836 Is used to set the source IP address to the value of specified
3837 expression. Useful when a proxy in front of HAProxy rewrites source IP,
3838 but provides the correct IP in a HTTP header; or you want to mask
3839 source IP for privacy.
3840
3841 <expr> Is a standard HAProxy expression formed by a sample-fetch
3842 followed by some converters.
3843
3844 Example:
3845
3846 http-request set-src hdr(x-forwarded-for)
3847 http-request set-src src,ipmask(24)
3848
3849 When set-src is successful, the source port is set to 0.
3850
William Lallemand44be6402016-05-25 01:51:35 +02003851 - set-src-port <expr> :
3852 Is used to set the source port address to the value of specified
3853 expression.
3854
3855 <expr> Is a standard HAProxy expression formed by a sample-fetch
3856 followed by some converters.
3857
3858 Example:
3859
3860 http-request set-src-port hdr(x-port)
3861 http-request set-src-port int(4000)
3862
3863 Be careful to use "set-src-port" after "set-src", because "set-src" sets
3864 the source port to 0.
3865
William Lallemand13e9b0c2016-05-25 02:34:07 +02003866 - set-dst <expr> :
3867 Is used to set the destination IP address to the value of specified
3868 expression. Useful when a proxy in front of HAProxy rewrites destination
3869 IP, but provides the correct IP in a HTTP header; or you want to mask
3870 the IP for privacy. If you want to connect to the new address/port, use
3871 '0.0.0.0:0' as a server address in the backend.
3872
3873 <expr> Is a standard HAProxy expression formed by a sample-fetch
3874 followed by some converters.
3875
3876 Example:
3877
3878 http-request set-dst hdr(x-dst)
3879 http-request set-dst dst,ipmask(24)
3880
3881 - set-dst-port <expr> :
3882 Is used to set the destination port address to the value of specified
3883 expression. If you want to connect to the new address/port, use
3884 '0.0.0.0:0' as a server address in the backend.
3885
3886 <expr> Is a standard HAProxy expression formed by a sample-fetch
3887 followed by some converters.
3888
3889 Example:
3890
3891 http-request set-dst-port hdr(x-port)
3892 http-request set-dst-port int(4000)
3893
Willy Tarreau2d392c22015-08-24 01:43:45 +02003894 - "silent-drop" : this stops the evaluation of the rules and makes the
3895 client-facing connection suddenly disappear using a system-dependant way
3896 that tries to prevent the client from being notified. The effect it then
3897 that the client still sees an established connection while there's none
3898 on HAProxy. The purpose is to achieve a comparable effect to "tarpit"
3899 except that it doesn't use any local resource at all on the machine
3900 running HAProxy. It can resist much higher loads than "tarpit", and slow
3901 down stronger attackers. It is important to undestand the impact of using
3902 this mechanism. All stateful equipments placed between the client and
3903 HAProxy (firewalls, proxies, load balancers) will also keep the
3904 established connection for a long time and may suffer from this action.
3905 On modern Linux systems running with enough privileges, the TCP_REPAIR
3906 socket option is used to block the emission of a TCP reset. On other
3907 systems, the socket's TTL is reduced to 1 so that the TCP reset doesn't
3908 pass the first router, though it's still delivered to local networks. Do
3909 not use it unless you fully understand how it works.
3910
Willy Tarreau20b0de52012-12-24 15:45:22 +01003911 There is no limit to the number of http-request statements per instance.
3912
3913 It is important to know that http-request rules are processed very early in
3914 the HTTP processing, just after "block" rules and before "reqdel" or "reqrep"
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08003915 or "reqadd" rules. That way, headers added by "add-header"/"set-header" are
3916 visible by almost all further ACL rules.
3917
3918 Using "reqadd"/"reqdel"/"reqrep" to manipulate request headers is discouraged
3919 in newer versions (>= 1.5). But if you need to use regular expression to
3920 delete headers, you can still use "reqdel". Also please use
3921 "http-request deny/allow/tarpit" instead of "reqdeny"/"reqpass"/"reqtarpit".
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003922
3923 Example:
Cyril Bonté78caf842010-03-10 22:41:43 +01003924 acl nagios src 192.168.129.3
3925 acl local_net src 192.168.0.0/16
3926 acl auth_ok http_auth(L1)
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003927
Cyril Bonté78caf842010-03-10 22:41:43 +01003928 http-request allow if nagios
3929 http-request allow if local_net auth_ok
3930 http-request auth realm Gimme if local_net auth_ok
3931 http-request deny
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003932
Cyril Bonté78caf842010-03-10 22:41:43 +01003933 Example:
3934 acl auth_ok http_auth_group(L1) G1
Cyril Bonté78caf842010-03-10 22:41:43 +01003935 http-request auth unless auth_ok
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003936
Willy Tarreau20b0de52012-12-24 15:45:22 +01003937 Example:
3938 http-request set-header X-Haproxy-Current-Date %T
3939 http-request set-header X-SSL %[ssl_fc]
Willy Tarreaufca42612015-08-27 17:15:05 +02003940 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id,hex]
Willy Tarreau20b0de52012-12-24 15:45:22 +01003941 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
3942 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
3943 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
3944 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
3945 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
3946 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
3947
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003948 Example:
3949 acl key req.hdr(X-Add-Acl-Key) -m found
3950 acl add path /addacl
3951 acl del path /delacl
3952
3953 acl myhost hdr(Host) -f myhost.lst
3954
3955 http-request add-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key add
3956 http-request del-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key del
3957
3958 Example:
3959 acl value req.hdr(X-Value) -m found
3960 acl setmap path /setmap
3961 acl delmap path /delmap
3962
3963 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
3964
3965 http-request set-map(map.lst) %[src] %[req.hdr(X-Value)] if setmap value
3966 http-request del-map(map.lst) %[src] if delmap
3967
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02003968 See also : "stats http-request", section 3.4 about userlists and section 7
3969 about ACL usage.
Willy Tarreauef781042010-01-27 11:53:01 +01003970
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003971http-response { allow | deny | add-header <name> <fmt> | set-nice <nice> |
Willy Tarreau51d861a2015-05-22 17:30:48 +02003972 capture <sample> id <id> | redirect <rule> |
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003973 set-header <name> <fmt> | del-header <name> |
Sasha Pachev218f0642014-06-16 12:05:59 -06003974 replace-header <name> <regex-match> <replace-fmt> |
3975 replace-value <name> <regex-match> <replace-fmt> |
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02003976 set-status <status> |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003977 set-log-level <level> | set-mark <mark> | set-tos <tos> |
3978 add-acl(<file name>) <key fmt> |
3979 del-acl(<file name>) <key fmt> |
3980 del-map(<file name>) <key fmt> |
Thierry FOURNIER90da1912015-03-05 11:17:06 +01003981 set-map(<file name>) <key fmt> <value fmt> |
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003982 set-var(<var-name>) <expr> |
Ruoshan Huange4edc6b2016-07-14 15:07:45 +08003983 { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] |
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02003984 sc-inc-gpc0(<sc-id>) |
Thierry FOURNIER236657b2015-08-19 08:25:14 +02003985 sc-set-gpt0(<sc-id>) <int> |
Willy Tarreau2d392c22015-08-24 01:43:45 +02003986 silent-drop |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003987 }
Lukas Tribus2dd1d1a2013-06-19 23:34:41 +02003988 [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02003989 Access control for Layer 7 responses
3990
3991 May be used in sections: defaults | frontend | listen | backend
3992 no | yes | yes | yes
3993
3994 The http-response statement defines a set of rules which apply to layer 7
3995 processing. The rules are evaluated in their declaration order when they are
3996 met in a frontend, listen or backend section. Any rule may optionally be
3997 followed by an ACL-based condition, in which case it will only be evaluated
3998 if the condition is true. Since these rules apply on responses, the backend
3999 rules are applied first, followed by the frontend's rules.
4000
4001 The first keyword is the rule's action. Currently supported actions include :
4002 - "allow" : this stops the evaluation of the rules and lets the response
4003 pass the check. No further "http-response" rules are evaluated for the
4004 current section.
4005
4006 - "deny" : this stops the evaluation of the rules and immediately rejects
4007 the response and emits an HTTP 502 error. No further "http-response"
4008 rules are evaluated.
4009
4010 - "add-header" appends an HTTP header field whose name is specified in
4011 <name> and whose value is defined by <fmt> which follows the log-format
4012 rules (see Custom Log Format in section 8.2.4). This may be used to send
4013 a cookie to a client for example, or to pass some internal information.
4014 This rule is not final, so it is possible to add other similar rules.
4015 Note that header addition is performed immediately, so one rule might
4016 reuse the resulting header from a previous rule.
4017
4018 - "set-header" does the same as "add-header" except that the header name
4019 is first removed if it existed. This is useful when passing security
4020 information to the server, where the header must not be manipulated by
4021 external users.
4022
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02004023 - "del-header" removes all HTTP header fields whose name is specified in
4024 <name>.
4025
Sasha Pachev218f0642014-06-16 12:05:59 -06004026 - "replace-header" matches the regular expression in all occurrences of
4027 header field <name> according to <match-regex>, and replaces them with
4028 the <replace-fmt> argument. Format characters are allowed in replace-fmt
4029 and work like in <fmt> arguments in "add-header". The match is only
4030 case-sensitive. It is important to understand that this action only
4031 considers whole header lines, regardless of the number of values they
4032 may contain. This usage is suited to headers naturally containing commas
4033 in their value, such as Set-Cookie, Expires and so on.
4034
4035 Example:
4036
4037 http-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
4038
4039 applied to:
4040
4041 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
4042
4043 outputs:
4044
4045 Set-Cookie: C=1;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT
4046
4047 assuming the backend IP is 192.168.1.20.
4048
4049 - "replace-value" works like "replace-header" except that it matches the
4050 regex against every comma-delimited value of the header field <name>
4051 instead of the entire header. This is suited for all headers which are
4052 allowed to carry more than one value. An example could be the Accept
4053 header.
4054
4055 Example:
4056
4057 http-response replace-value Cache-control ^public$ private
4058
4059 applied to:
4060
4061 Cache-Control: max-age=3600, public
4062
4063 outputs:
4064
4065 Cache-Control: max-age=3600, private
4066
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004067 - "set-status" replaces the response status code with <status> which must
4068 be an integer between 100 and 999. Note that the reason is automatically
4069 adapted to the new code.
4070
4071 Example:
4072
4073 # return "431 Request Header Fields Too Large"
4074 http-response set-status 431
4075
Willy Tarreauf4c43c12013-06-11 17:01:13 +02004076 - "set-nice" sets the "nice" factor of the current request being processed.
4077 It only has effect against the other requests being processed at the same
4078 time. The default value is 0, unless altered by the "nice" setting on the
4079 "bind" line. The accepted range is -1024..1024. The higher the value, the
4080 nicest the request will be. Lower values will make the request more
4081 important than other ones. This can be useful to improve the speed of
4082 some requests, or lower the priority of non-important requests. Using
4083 this setting without prior experimentation can cause some major slowdown.
4084
Willy Tarreau9a355ec2013-06-11 17:45:46 +02004085 - "set-log-level" is used to change the log level of the current request
4086 when a certain condition is met. Valid levels are the 8 syslog levels
4087 (see the "log" keyword) plus the special level "silent" which disables
4088 logging for this request. This rule is not final so the last matching
4089 rule wins. This rule can be useful to disable health checks coming from
4090 another equipment.
4091
Willy Tarreau42cf39e2013-06-11 18:51:32 +02004092 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
4093 the client to the value passed in <tos> on platforms which support this.
4094 This value represents the whole 8 bits of the IP TOS field, and can be
4095 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
4096 that only the 6 higher bits are used in DSCP or TOS, and the two lower
4097 bits are always 0. This can be used to adjust some routing behaviour on
4098 border routers based on some information from the request. See RFC 2474,
4099 2597, 3260 and 4594 for more information.
4100
Willy Tarreau51347ed2013-06-11 19:34:13 +02004101 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
4102 client to the value passed in <mark> on platforms which support it. This
4103 value is an unsigned 32 bit value which can be matched by netfilter and
4104 by the routing table. It can be expressed both in decimal or hexadecimal
4105 format (prefixed by "0x"). This can be useful to force certain packets to
4106 take a different route (for example a cheaper network path for bulk
4107 downloads). This works on Linux kernels 2.6.32 and above and requires
4108 admin privileges.
4109
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004110 - "add-acl" is used to add a new entry into an ACL. The ACL must be loaded
4111 from a file (even a dummy empty file). The file name of the ACL to be
4112 updated is passed between parentheses. It takes one argument: <key fmt>,
4113 which follows log-format rules, to collect content of the new entry. It
4114 performs a lookup in the ACL before insertion, to avoid duplicated (or
4115 more) values. This lookup is done by a linear search and can be expensive
4116 with large lists! It is the equivalent of the "add acl" command from the
4117 stats socket, but can be triggered by an HTTP response.
4118
4119 - "del-acl" is used to delete an entry from an ACL. The ACL must be loaded
4120 from a file (even a dummy empty file). The file name of the ACL to be
4121 updated is passed between parentheses. It takes one argument: <key fmt>,
4122 which follows log-format rules, to collect content of the entry to delete.
4123 It is the equivalent of the "del acl" command from the stats socket, but
4124 can be triggered by an HTTP response.
4125
4126 - "del-map" is used to delete an entry from a MAP. The MAP must be loaded
4127 from a file (even a dummy empty file). The file name of the MAP to be
4128 updated is passed between parentheses. It takes one argument: <key fmt>,
4129 which follows log-format rules, to collect content of the entry to delete.
4130 It takes one argument: "file name" It is the equivalent of the "del map"
4131 command from the stats socket, but can be triggered by an HTTP response.
4132
4133 - "set-map" is used to add a new entry into a MAP. The MAP must be loaded
4134 from a file (even a dummy empty file). The file name of the MAP to be
4135 updated is passed between parentheses. It takes 2 arguments: <key fmt>,
4136 which follows log-format rules, used to collect MAP key, and <value fmt>,
4137 which follows log-format rules, used to collect content for the new entry.
4138 It performs a lookup in the MAP before insertion, to avoid duplicated (or
4139 more) values. This lookup is done by a linear search and can be expensive
4140 with large lists! It is the equivalent of the "set map" command from the
4141 stats socket, but can be triggered by an HTTP response.
4142
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02004143 - capture <sample> id <id> :
4144 captures sample expression <sample> from the response buffer, and converts
4145 it to a string. The resulting string is stored into the next request
4146 "capture" slot, so it will possibly appear next to some captured HTTP
4147 headers. It will then automatically appear in the logs, and it will be
4148 possible to extract it using sample fetch rules to feed it into headers or
4149 anything. Please check section 7.3 (Fetching samples) and "capture
4150 response header" for more information.
4151
4152 The keyword "id" is the id of the capture slot which is used for storing
4153 the string. The capture slot must be defined in an associated frontend.
4154 This is useful to run captures in backends. The slot id can be declared by
4155 a previous directive "http-response capture" or with the "declare capture"
4156 keyword.
Baptiste Assmanne9544932015-11-03 23:31:35 +01004157 If the slot <id> doesn't exist, then HAProxy fails parsing the
4158 configuration to prevent unexpected behavior at run time.
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02004159
Willy Tarreau51d861a2015-05-22 17:30:48 +02004160 - "redirect" : this performs an HTTP redirection based on a redirect rule.
4161 This supports a format string similarly to "http-request redirect" rules,
4162 with the exception that only the "location" type of redirect is possible
4163 on the response. See the "redirect" keyword for the rule's syntax. When
4164 a redirect rule is applied during a response, connections to the server
4165 are closed so that no data can be forwarded from the server to the client.
4166
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004167 - set-var(<var-name>) expr:
4168 Is used to set the contents of a variable. The variable is declared
4169 inline.
4170
Daniel Schneller0b547052016-03-21 20:46:57 +01004171 <var-name> The name of the variable starts with an indication about
4172 its scope. The scopes allowed are:
4173 "sess" : the variable is shared with the whole session
4174 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004175 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +01004176 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004177 processing
Daniel Schneller0b547052016-03-21 20:46:57 +01004178 "res" : the variable is shared only during response
4179 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004180 This prefix is followed by a name. The separator is a '.'.
Daniel Schneller0b547052016-03-21 20:46:57 +01004181 The name may only contain characters 'a-z', 'A-Z', '0-9'
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004182 and '_'.
4183
4184 <expr> Is a standard HAProxy expression formed by a sample-fetch
4185 followed by some converters.
4186
4187 Example:
4188
4189 http-response set-var(sess.last_redir) res.hdr(location)
4190
Ruoshan Huange4edc6b2016-07-14 15:07:45 +08004191 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
4192 enables tracking of sticky counters from current response. Please refer to
4193 "http-request track-sc" for a complete description. The only difference
4194 from "http-request track-sc" is the <key> sample expression can only make
4195 use of samples in response (eg. res.*, status etc.) and samples below
4196 Layer 6 (eg. ssl related samples, see section 7.3.4). If the sample is
4197 not supported, haproxy will fail and warn while parsing the config.
4198
Thierry FOURNIER236657b2015-08-19 08:25:14 +02004199 - sc-set-gpt0(<sc-id>) <int> :
4200 This action sets the GPT0 tag according to the sticky counter designated
4201 by <sc-id> and the value of <int>. The expected result is a boolean. If
4202 an error occurs, this action silently fails and the actions evaluation
4203 continues.
4204
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02004205 - sc-inc-gpc0(<sc-id>):
4206 This action increments the GPC0 counter according with the sticky counter
4207 designated by <sc-id>. If an error occurs, this action silently fails and
4208 the actions evaluation continues.
4209
Willy Tarreau2d392c22015-08-24 01:43:45 +02004210 - "silent-drop" : this stops the evaluation of the rules and makes the
4211 client-facing connection suddenly disappear using a system-dependant way
4212 that tries to prevent the client from being notified. The effect it then
4213 that the client still sees an established connection while there's none
4214 on HAProxy. The purpose is to achieve a comparable effect to "tarpit"
4215 except that it doesn't use any local resource at all on the machine
4216 running HAProxy. It can resist much higher loads than "tarpit", and slow
4217 down stronger attackers. It is important to undestand the impact of using
4218 this mechanism. All stateful equipments placed between the client and
4219 HAProxy (firewalls, proxies, load balancers) will also keep the
4220 established connection for a long time and may suffer from this action.
4221 On modern Linux systems running with enough privileges, the TCP_REPAIR
4222 socket option is used to block the emission of a TCP reset. On other
4223 systems, the socket's TTL is reduced to 1 so that the TCP reset doesn't
4224 pass the first router, though it's still delivered to local networks. Do
4225 not use it unless you fully understand how it works.
4226
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004227 There is no limit to the number of http-response statements per instance.
4228
Godbach09250262013-07-02 01:19:15 +08004229 It is important to know that http-response rules are processed very early in
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08004230 the HTTP processing, before "rspdel" or "rsprep" or "rspadd" rules. That way,
4231 headers added by "add-header"/"set-header" are visible by almost all further ACL
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004232 rules.
4233
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08004234 Using "rspadd"/"rspdel"/"rsprep" to manipulate request headers is discouraged
4235 in newer versions (>= 1.5). But if you need to use regular expression to
4236 delete headers, you can still use "rspdel". Also please use
4237 "http-response deny" instead of "rspdeny".
4238
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004239 Example:
4240 acl key_acl res.hdr(X-Acl-Key) -m found
4241
4242 acl myhost hdr(Host) -f myhost.lst
4243
4244 http-response add-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
4245 http-response del-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
4246
4247 Example:
4248 acl value res.hdr(X-Value) -m found
4249
4250 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
4251
4252 http-response set-map(map.lst) %[src] %[res.hdr(X-Value)] if value
4253 http-response del-map(map.lst) %[src] if ! value
4254
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004255 See also : "http-request", section 3.4 about userlists and section 7 about
4256 ACL usage.
4257
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02004258
Willy Tarreau30631952015-08-06 15:05:24 +02004259http-reuse { never | safe | aggressive | always }
4260 Declare how idle HTTP connections may be shared between requests
4261
4262 May be used in sections: defaults | frontend | listen | backend
4263 yes | no | yes | yes
4264
4265 By default, a connection established between haproxy and the backend server
4266 belongs to the session that initiated it. The downside is that between the
4267 response and the next request, the connection remains idle and is not used.
4268 In many cases for performance reasons it is desirable to make it possible to
4269 reuse these idle connections to serve other requests from different sessions.
4270 This directive allows to tune this behaviour.
4271
4272 The argument indicates the desired connection reuse strategy :
4273
4274 - "never" : idle connections are never shared between sessions. This is
4275 the default choice. It may be enforced to cancel a different
4276 strategy inherited from a defaults section or for
4277 troubleshooting. For example, if an old bogus application
4278 considers that multiple requests over the same connection come
4279 from the same client and it is not possible to fix the
4280 application, it may be desirable to disable connection sharing
4281 in a single backend. An example of such an application could
4282 be an old haproxy using cookie insertion in tunnel mode and
4283 not checking any request past the first one.
4284
4285 - "safe" : this is the recommended strategy. The first request of a
4286 session is always sent over its own connection, and only
4287 subsequent requests may be dispatched over other existing
4288 connections. This ensures that in case the server closes the
4289 connection when the request is being sent, the browser can
4290 decide to silently retry it. Since it is exactly equivalent to
4291 regular keep-alive, there should be no side effects.
4292
4293 - "aggressive" : this mode may be useful in webservices environments where
4294 all servers are not necessarily known and where it would be
4295 appreciable to deliver most first requests over existing
4296 connections. In this case, first requests are only delivered
4297 over existing connections that have been reused at least once,
4298 proving that the server correctly supports connection reuse.
4299 It should only be used when it's sure that the client can
4300 retry a failed request once in a while and where the benefit
4301 of aggressive connection reuse significantly outweights the
4302 downsides of rare connection failures.
4303
4304 - "always" : this mode is only recommended when the path to the server is
4305 known for never breaking existing connections quickly after
4306 releasing them. It allows the first request of a session to be
4307 sent to an existing connection. This can provide a significant
4308 performance increase over the "safe" strategy when the backend
4309 is a cache farm, since such components tend to show a
4310 consistent behaviour and will benefit from the connection
4311 sharing. It is recommended that the "http-keep-alive" timeout
4312 remains low in this mode so that no dead connections remain
4313 usable. In most cases, this will lead to the same performance
4314 gains as "aggressive" but with more risks. It should only be
4315 used when it improves the situation over "aggressive".
4316
4317 When http connection sharing is enabled, a great care is taken to respect the
4318 connection properties and compatiblities. Specifically :
4319 - connections made with "usesrc" followed by a client-dependant value
4320 ("client", "clientip", "hdr_ip") are marked private and never shared ;
4321
4322 - connections sent to a server with a TLS SNI extension are marked private
4323 and are never shared ;
4324
4325 - connections receiving a status code 401 or 407 expect some authentication
4326 to be sent in return. Due to certain bogus authentication schemes (such
4327 as NTLM) relying on the connection, these connections are marked private
4328 and are never shared ;
4329
4330 No connection pool is involved, once a session dies, the last idle connection
4331 it was attached to is deleted at the same time. This ensures that connections
4332 may not last after all sessions are closed.
4333
4334 Note: connection reuse improves the accuracy of the "server maxconn" setting,
4335 because almost no new connection will be established while idle connections
4336 remain available. This is particularly true with the "always" strategy.
4337
4338 See also : "option http-keep-alive", "server maxconn"
4339
4340
Mark Lamourinec2247f02012-01-04 13:02:01 -05004341http-send-name-header [<header>]
4342 Add the server name to a request. Use the header string given by <header>
4343
4344 May be used in sections: defaults | frontend | listen | backend
4345 yes | no | yes | yes
4346
4347 Arguments :
4348
4349 <header> The header string to use to send the server name
4350
4351 The "http-send-name-header" statement causes the name of the target
4352 server to be added to the headers of an HTTP request. The name
4353 is added with the header string proved.
4354
4355 See also : "server"
4356
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01004357id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02004358 Set a persistent ID to a proxy.
4359 May be used in sections : defaults | frontend | listen | backend
4360 no | yes | yes | yes
4361 Arguments : none
4362
4363 Set a persistent ID for the proxy. This ID must be unique and positive.
4364 An unused ID will automatically be assigned if unset. The first assigned
4365 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01004366
4367
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004368ignore-persist { if | unless } <condition>
4369 Declare a condition to ignore persistence
4370 May be used in sections: defaults | frontend | listen | backend
4371 no | yes | yes | yes
4372
4373 By default, when cookie persistence is enabled, every requests containing
4374 the cookie are unconditionally persistent (assuming the target server is up
4375 and running).
4376
4377 The "ignore-persist" statement allows one to declare various ACL-based
4378 conditions which, when met, will cause a request to ignore persistence.
4379 This is sometimes useful to load balance requests for static files, which
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03004380 often don't require persistence. This can also be used to fully disable
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004381 persistence for a specific User-Agent (for example, some web crawler bots).
4382
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004383 The persistence is ignored when an "if" condition is met, or unless an
4384 "unless" condition is met.
4385
4386 See also : "force-persist", "cookie", and section 7 about ACL usage.
4387
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004388load-server-state-from-file { global | local | none }
4389 Allow seamless reload of HAProxy
4390 May be used in sections: defaults | frontend | listen | backend
4391 yes | no | yes | yes
4392
4393 This directive points HAProxy to a file where server state from previous
4394 running process has been saved. That way, when starting up, before handling
4395 traffic, the new process can apply old states to servers exactly has if no
4396 reload occured. The purpose of the "load-server-state-from-file" directive is
4397 to tell haproxy which file to use. For now, only 2 arguments to either prevent
4398 loading state or load states from a file containing all backends and servers.
4399 The state file can be generated by running the command "show servers state"
4400 over the stats socket and redirect output.
4401
4402 The format of the file is versionned and is very specific. To understand it,
4403 please read the documentation of the "show servers state" command (chapter
Kevin Decherf949c7202015-10-13 23:26:44 +02004404 9.2 of Management Guide).
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004405
4406 Arguments:
4407 global load the content of the file pointed by the global directive
4408 named "server-state-file".
4409
4410 local load the content of the file pointed by the directive
4411 "server-state-file-name" if set. If not set, then the backend
4412 name is used as a file name.
4413
4414 none don't load any stat for this backend
4415
4416 Notes:
4417 - server's IP address is not updated unless DNS resolution is enabled on
4418 the server. It means that if a server IP address has been changed using
4419 the stat socket, this information won't be re-applied after reloading.
4420
4421 - server's weight is applied from previous running process unless it has
4422 has changed between previous and new configuration files.
4423
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004424 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004425
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004426 global
4427 stats socket /tmp/socket
4428 server-state-file /tmp/server_state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004429
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004430 defaults
4431 load-server-state-from-file global
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004432
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004433 backend bk
4434 server s1 127.0.0.1:22 check weight 11
4435 server s2 127.0.0.1:22 check weight 12
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004436
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004437
4438 Then one can run :
4439
4440 socat /tmp/socket - <<< "show servers state" > /tmp/server_state
4441
4442 Content of the file /tmp/server_state would be like this:
4443
4444 1
4445 # <field names skipped for the doc example>
4446 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
4447 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
4448
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004449 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004450
4451 global
4452 stats socket /tmp/socket
4453 server-state-base /etc/haproxy/states
4454
4455 defaults
4456 load-server-state-from-file local
4457
4458 backend bk
4459 server s1 127.0.0.1:22 check weight 11
4460 server s2 127.0.0.1:22 check weight 12
4461
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004462
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004463 Then one can run :
4464
4465 socat /tmp/socket - <<< "show servers state bk" > /etc/haproxy/states/bk
4466
4467 Content of the file /etc/haproxy/states/bk would be like this:
4468
4469 1
4470 # <field names skipped for the doc example>
4471 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
4472 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
4473
4474 See also: "server-state-file", "server-state-file-name", and
4475 "show servers state"
4476
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004477
Willy Tarreau2769aa02007-12-27 18:26:09 +01004478log global
Willy Tarreau18324f52014-06-27 18:10:07 +02004479log <address> [len <length>] <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02004480no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01004481 Enable per-instance logging of events and traffic.
4482 May be used in sections : defaults | frontend | listen | backend
4483 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02004484
4485 Prefix :
4486 no should be used when the logger list must be flushed. For example,
4487 if you don't want to inherit from the default logger list. This
4488 prefix does not allow arguments.
4489
Willy Tarreau2769aa02007-12-27 18:26:09 +01004490 Arguments :
4491 global should be used when the instance's logging parameters are the
4492 same as the global ones. This is the most common usage. "global"
4493 replaces <address>, <facility> and <level> with those of the log
4494 entries found in the "global" section. Only one "log global"
4495 statement may be used per instance, and this form takes no other
4496 parameter.
4497
4498 <address> indicates where to send the logs. It takes the same format as
4499 for the "global" section's logs, and can be one of :
4500
4501 - An IPv4 address optionally followed by a colon (':') and a UDP
4502 port. If no port is specified, 514 is used by default (the
4503 standard syslog port).
4504
David du Colombier24bb5f52011-03-17 10:40:23 +01004505 - An IPv6 address followed by a colon (':') and optionally a UDP
4506 port. If no port is specified, 514 is used by default (the
4507 standard syslog port).
4508
Willy Tarreau2769aa02007-12-27 18:26:09 +01004509 - A filesystem path to a UNIX domain socket, keeping in mind
4510 considerations for chroot (be sure the path is accessible
4511 inside the chroot) and uid/gid (be sure the path is
4512 appropriately writeable).
4513
William Lallemandb2f07452015-05-12 14:27:13 +02004514 You may want to reference some environment variables in the
4515 address parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01004516
Willy Tarreau18324f52014-06-27 18:10:07 +02004517 <length> is an optional maximum line length. Log lines larger than this
4518 value will be truncated before being sent. The reason is that
4519 syslog servers act differently on log line length. All servers
4520 support the default value of 1024, but some servers simply drop
4521 larger lines while others do log them. If a server supports long
4522 lines, it may make sense to set this value here in order to avoid
4523 truncating long lines. Similarly, if a server drops long lines,
4524 it is preferable to truncate them before sending them. Accepted
4525 values are 80 to 65535 inclusive. The default value of 1024 is
4526 generally fine for all standard usages. Some specific cases of
4527 long captures or JSON-formated logs may require larger values.
4528
Willy Tarreau2769aa02007-12-27 18:26:09 +01004529 <facility> must be one of the 24 standard syslog facilities :
4530
4531 kern user mail daemon auth syslog lpr news
4532 uucp cron auth2 ftp ntp audit alert cron2
4533 local0 local1 local2 local3 local4 local5 local6 local7
4534
4535 <level> is optional and can be specified to filter outgoing messages. By
4536 default, all messages are sent. If a level is specified, only
4537 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02004538 will be sent. An optional minimum level can be specified. If it
4539 is set, logs emitted with a more severe level than this one will
4540 be capped to this level. This is used to avoid sending "emerg"
4541 messages on all terminals on some default syslog configurations.
4542 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01004543
4544 emerg alert crit err warning notice info debug
4545
William Lallemand0f99e342011-10-12 17:50:54 +02004546 It is important to keep in mind that it is the frontend which decides what to
4547 log from a connection, and that in case of content switching, the log entries
4548 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01004549
4550 However, backend log declaration define how and where servers status changes
4551 will be logged. Level "notice" will be used to indicate a server going up,
4552 "warning" will be used for termination signals and definitive service
4553 termination, and "alert" will be used for when a server goes down.
4554
4555 Note : According to RFC3164, messages are truncated to 1024 bytes before
4556 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004557
4558 Example :
4559 log global
Willy Tarreauf7edefa2009-05-10 17:20:05 +02004560 log 127.0.0.1:514 local0 notice # only send important events
4561 log 127.0.0.1:514 local0 notice notice # same but limit output level
William Lallemandb2f07452015-05-12 14:27:13 +02004562 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
Willy Tarreaudad36a32013-03-11 01:20:04 +01004563
Willy Tarreau2769aa02007-12-27 18:26:09 +01004564
William Lallemand48940402012-01-30 16:47:22 +01004565log-format <string>
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01004566 Specifies the log format string to use for traffic logs
4567 May be used in sections: defaults | frontend | listen | backend
4568 yes | yes | yes | no
William Lallemand48940402012-01-30 16:47:22 +01004569
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01004570 This directive specifies the log format string that will be used for all logs
4571 resulting from traffic passing through the frontend using this line. If the
4572 directive is used in a defaults section, all subsequent frontends will use
4573 the same log format. Please see section 8.2.4 which covers the log format
4574 string in depth.
William Lallemand48940402012-01-30 16:47:22 +01004575
Dragan Dosen7ad31542015-09-28 17:16:47 +02004576log-format-sd <string>
4577 Specifies the RFC5424 structured-data log format string
4578 May be used in sections: defaults | frontend | listen | backend
4579 yes | yes | yes | no
4580
4581 This directive specifies the RFC5424 structured-data log format string that
4582 will be used for all logs resulting from traffic passing through the frontend
4583 using this line. If the directive is used in a defaults section, all
4584 subsequent frontends will use the same log format. Please see section 8.2.4
4585 which covers the log format string in depth.
4586
4587 See https://tools.ietf.org/html/rfc5424#section-6.3 for more information
4588 about the RFC5424 structured-data part.
4589
4590 Note : This log format string will be used only for loggers that have set
4591 log format to "rfc5424".
4592
4593 Example :
4594 log-format-sd [exampleSDID@1234\ bytes=\"%B\"\ status=\"%ST\"]
4595
4596
Willy Tarreau094af4e2015-01-07 15:03:42 +01004597log-tag <string>
4598 Specifies the log tag to use for all outgoing logs
4599 May be used in sections: defaults | frontend | listen | backend
4600 yes | yes | yes | yes
4601
4602 Sets the tag field in the syslog header to this string. It defaults to the
4603 log-tag set in the global section, otherwise the program name as launched
4604 from the command line, which usually is "haproxy". Sometimes it can be useful
4605 to differentiate between multiple processes running on the same host, or to
4606 differentiate customer instances running in the same process. In the backend,
4607 logs about servers up/down will use this tag. As a hint, it can be convenient
4608 to set a log-tag related to a hosted customer in a defaults section then put
4609 all the frontends and backends for that customer, then start another customer
4610 in a new defaults section. See also the global "log-tag" directive.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004611
Willy Tarreauc35362a2014-04-25 13:58:37 +02004612max-keep-alive-queue <value>
4613 Set the maximum server queue size for maintaining keep-alive connections
4614 May be used in sections: defaults | frontend | listen | backend
4615 yes | no | yes | yes
4616
4617 HTTP keep-alive tries to reuse the same server connection whenever possible,
4618 but sometimes it can be counter-productive, for example if a server has a lot
4619 of connections while other ones are idle. This is especially true for static
4620 servers.
4621
4622 The purpose of this setting is to set a threshold on the number of queued
4623 connections at which haproxy stops trying to reuse the same server and prefers
4624 to find another one. The default value, -1, means there is no limit. A value
4625 of zero means that keep-alive requests will never be queued. For very close
4626 servers which can be reached with a low latency and which are not sensible to
4627 breaking keep-alive, a low value is recommended (eg: local static server can
4628 use a value of 10 or less). For remote servers suffering from a high latency,
4629 higher values might be needed to cover for the latency and/or the cost of
4630 picking a different server.
4631
4632 Note that this has no impact on responses which are maintained to the same
4633 server consecutively to a 401 response. They will still go to the same server
4634 even if they have to be queued.
4635
4636 See also : "option http-server-close", "option prefer-last-server", server
4637 "maxconn" and cookie persistence.
4638
4639
Willy Tarreau2769aa02007-12-27 18:26:09 +01004640maxconn <conns>
4641 Fix the maximum number of concurrent connections on a frontend
4642 May be used in sections : defaults | frontend | listen | backend
4643 yes | yes | yes | no
4644 Arguments :
4645 <conns> is the maximum number of concurrent connections the frontend will
4646 accept to serve. Excess connections will be queued by the system
4647 in the socket's listen queue and will be served once a connection
4648 closes.
4649
4650 If the system supports it, it can be useful on big sites to raise this limit
4651 very high so that haproxy manages connection queues, instead of leaving the
4652 clients with unanswered connection attempts. This value should not exceed the
4653 global maxconn. Also, keep in mind that a connection contains two buffers
Baptiste Assmann79fb45d2016-03-06 23:34:31 +01004654 of tune.bufsize (16kB by default) each, as well as some other data resulting
4655 in about 33 kB of RAM being consumed per established connection. That means
4656 that a medium system equipped with 1GB of RAM can withstand around
4657 20000-25000 concurrent connections if properly tuned.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004658
4659 Also, when <conns> is set to large values, it is possible that the servers
4660 are not sized to accept such loads, and for this reason it is generally wise
4661 to assign them some reasonable connection limits.
4662
Vincent Bernat6341be52012-06-27 17:18:30 +02004663 By default, this value is set to 2000.
4664
Willy Tarreau2769aa02007-12-27 18:26:09 +01004665 See also : "server", global section's "maxconn", "fullconn"
4666
4667
4668mode { tcp|http|health }
4669 Set the running mode or protocol of the instance
4670 May be used in sections : defaults | frontend | listen | backend
4671 yes | yes | yes | yes
4672 Arguments :
4673 tcp The instance will work in pure TCP mode. A full-duplex connection
4674 will be established between clients and servers, and no layer 7
4675 examination will be performed. This is the default mode. It
4676 should be used for SSL, SSH, SMTP, ...
4677
4678 http The instance will work in HTTP mode. The client request will be
4679 analyzed in depth before connecting to any server. Any request
4680 which is not RFC-compliant will be rejected. Layer 7 filtering,
4681 processing and switching will be possible. This is the mode which
4682 brings HAProxy most of its value.
4683
4684 health The instance will work in "health" mode. It will just reply "OK"
Willy Tarreau82569f92012-09-27 23:48:56 +02004685 to incoming connections and close the connection. Alternatively,
4686 If the "httpchk" option is set, "HTTP/1.0 200 OK" will be sent
4687 instead. Nothing will be logged in either case. This mode is used
4688 to reply to external components health checks. This mode is
4689 deprecated and should not be used anymore as it is possible to do
4690 the same and even better by combining TCP or HTTP modes with the
4691 "monitor" keyword.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004692
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004693 When doing content switching, it is mandatory that the frontend and the
4694 backend are in the same mode (generally HTTP), otherwise the configuration
4695 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004696
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004697 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01004698 defaults http_instances
4699 mode http
4700
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004701 See also : "monitor", "monitor-net"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004702
Willy Tarreau0ba27502007-12-24 16:55:16 +01004703
Cyril Bontéf0c60612010-02-06 14:44:47 +01004704monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01004705 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004706 May be used in sections : defaults | frontend | listen | backend
4707 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01004708 Arguments :
4709 if <cond> the monitor request will fail if the condition is satisfied,
4710 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004711 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01004712 are met, for instance a low number of servers both in a
4713 backend and its backup.
4714
4715 unless <cond> the monitor request will succeed only if the condition is
4716 satisfied, and will fail otherwise. Such a condition may be
4717 based on a test on the presence of a minimum number of active
4718 servers in a list of backends.
4719
4720 This statement adds a condition which can force the response to a monitor
4721 request to report a failure. By default, when an external component queries
4722 the URI dedicated to monitoring, a 200 response is returned. When one of the
4723 conditions above is met, haproxy will return 503 instead of 200. This is
4724 very useful to report a site failure to an external component which may base
4725 routing advertisements between multiple sites on the availability reported by
4726 haproxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02004727 criterion. Note that "monitor fail" only works in HTTP mode. Both status
4728 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004729
4730 Example:
4731 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01004732 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01004733 acl site_dead nbsrv(dynamic) lt 2
4734 acl site_dead nbsrv(static) lt 2
4735 monitor-uri /site_alive
4736 monitor fail if site_dead
4737
Willy Tarreauae94d4d2011-05-11 16:28:49 +02004738 See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004739
4740
4741monitor-net <source>
4742 Declare a source network which is limited to monitor requests
4743 May be used in sections : defaults | frontend | listen | backend
4744 yes | yes | yes | no
4745 Arguments :
4746 <source> is the source IPv4 address or network which will only be able to
4747 get monitor responses to any request. It can be either an IPv4
4748 address, a host name, or an address followed by a slash ('/')
4749 followed by a mask.
4750
4751 In TCP mode, any connection coming from a source matching <source> will cause
4752 the connection to be immediately closed without any log. This allows another
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004753 equipment to probe the port and verify that it is still listening, without
Willy Tarreau2769aa02007-12-27 18:26:09 +01004754 forwarding the connection to a remote server.
4755
4756 In HTTP mode, a connection coming from a source matching <source> will be
4757 accepted, the following response will be sent without waiting for a request,
4758 then the connection will be closed : "HTTP/1.0 200 OK". This is normally
4759 enough for any front-end HTTP probe to detect that the service is UP and
Willy Tarreau82569f92012-09-27 23:48:56 +02004760 running without forwarding the request to a backend server. Note that this
4761 response is sent in raw format, without any transformation. This is important
4762 as it means that it will not be SSL-encrypted on SSL listeners.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004763
Willy Tarreau82569f92012-09-27 23:48:56 +02004764 Monitor requests are processed very early, just after tcp-request connection
4765 ACLs which are the only ones able to block them. These connections are short
4766 lived and never wait for any data from the client. They cannot be logged, and
4767 it is the intended purpose. They are only used to report HAProxy's health to
4768 an upper component, nothing more. Please note that "monitor fail" rules do
4769 not apply to connections intercepted by "monitor-net".
Willy Tarreau2769aa02007-12-27 18:26:09 +01004770
Willy Tarreau95cd2832010-03-04 23:36:33 +01004771 Last, please note that only one "monitor-net" statement can be specified in
4772 a frontend. If more than one is found, only the last one will be considered.
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004773
Willy Tarreau2769aa02007-12-27 18:26:09 +01004774 Example :
4775 # addresses .252 and .253 are just probing us.
4776 frontend www
4777 monitor-net 192.168.0.252/31
4778
4779 See also : "monitor fail", "monitor-uri"
4780
4781
4782monitor-uri <uri>
4783 Intercept a URI used by external components' monitor requests
4784 May be used in sections : defaults | frontend | listen | backend
4785 yes | yes | yes | no
4786 Arguments :
4787 <uri> is the exact URI which we want to intercept to return HAProxy's
4788 health status instead of forwarding the request.
4789
4790 When an HTTP request referencing <uri> will be received on a frontend,
4791 HAProxy will not forward it nor log it, but instead will return either
4792 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
4793 conditions defined with "monitor fail". This is normally enough for any
4794 front-end HTTP probe to detect that the service is UP and running without
4795 forwarding the request to a backend server. Note that the HTTP method, the
4796 version and all headers are ignored, but the request must at least be valid
4797 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
4798
4799 Monitor requests are processed very early. It is not possible to block nor
4800 divert them using ACLs. They cannot be logged either, and it is the intended
4801 purpose. They are only used to report HAProxy's health to an upper component,
4802 nothing more. However, it is possible to add any number of conditions using
4803 "monitor fail" and ACLs so that the result can be adjusted to whatever check
4804 can be imagined (most often the number of available servers in a backend).
4805
4806 Example :
4807 # Use /haproxy_test to report haproxy's status
4808 frontend www
4809 mode http
4810 monitor-uri /haproxy_test
4811
4812 See also : "monitor fail", "monitor-net"
4813
Willy Tarreau0ba27502007-12-24 16:55:16 +01004814
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004815option abortonclose
4816no option abortonclose
4817 Enable or disable early dropping of aborted requests pending in queues.
4818 May be used in sections : defaults | frontend | listen | backend
4819 yes | no | yes | yes
4820 Arguments : none
4821
4822 In presence of very high loads, the servers will take some time to respond.
4823 The per-instance connection queue will inflate, and the response time will
4824 increase respective to the size of the queue times the average per-session
4825 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01004826 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004827 the queue, and slowing down other users, and the servers as well, because the
4828 request will eventually be served, then aborted at the first error
4829 encountered while delivering the response.
4830
4831 As there is no way to distinguish between a full STOP and a simple output
4832 close on the client side, HTTP agents should be conservative and consider
4833 that the client might only have closed its output channel while waiting for
4834 the response. However, this introduces risks of congestion when lots of users
4835 do the same, and is completely useless nowadays because probably no client at
4836 all will close the session while waiting for the response. Some HTTP agents
Willy Tarreaud72758d2010-01-12 10:42:19 +01004837 support this behaviour (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004838 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01004839 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004840 of being the single component to break rare but valid traffic is extremely
4841 low, which adds to the temptation to be able to abort a session early while
4842 still not served and not pollute the servers.
4843
4844 In HAProxy, the user can choose the desired behaviour using the option
4845 "abortonclose". By default (without the option) the behaviour is HTTP
4846 compliant and aborted requests will be served. But when the option is
4847 specified, a session with an incoming channel closed will be aborted while
4848 it is still possible, either pending in the queue for a connection slot, or
4849 during the connection establishment if the server has not yet acknowledged
4850 the connection request. This considerably reduces the queue size and the load
4851 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01004852 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004853
4854 If this option has been enabled in a "defaults" section, it can be disabled
4855 in a specific instance by prepending the "no" keyword before it.
4856
4857 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
4858
4859
Willy Tarreau4076a152009-04-02 15:18:36 +02004860option accept-invalid-http-request
4861no option accept-invalid-http-request
4862 Enable or disable relaxing of HTTP request parsing
4863 May be used in sections : defaults | frontend | listen | backend
4864 yes | yes | yes | no
4865 Arguments : none
4866
Willy Tarreau91852eb2015-05-01 13:26:00 +02004867 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02004868 means that invalid characters in header names are not permitted and cause an
4869 error to be returned to the client. This is the desired behaviour as such
4870 forbidden characters are essentially used to build attacks exploiting server
4871 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
4872 server will emit invalid header names for whatever reason (configuration,
4873 implementation) and the issue will not be immediately fixed. In such a case,
4874 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01004875 even if that does not make sense, by specifying this option. Similarly, the
4876 list of characters allowed to appear in a URI is well defined by RFC3986, and
4877 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
4878 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
4879 not allowed at all. Haproxy always blocks a number of them (0..32, 127). The
Willy Tarreau91852eb2015-05-01 13:26:00 +02004880 remaining ones are blocked by default unless this option is enabled. This
Willy Tarreau13317662015-05-01 13:47:08 +02004881 option also relaxes the test on the HTTP version, it allows HTTP/0.9 requests
4882 to pass through (no version specified) and multiple digits for both the major
4883 and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02004884
4885 This option should never be enabled by default as it hides application bugs
4886 and open security breaches. It should only be deployed after a problem has
4887 been confirmed.
4888
4889 When this option is enabled, erroneous header names will still be accepted in
4890 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01004891 analysis using the "show errors" request on the UNIX stats socket. Similarly,
4892 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02004893 also helps confirming that the issue has been solved.
4894
4895 If this option has been enabled in a "defaults" section, it can be disabled
4896 in a specific instance by prepending the "no" keyword before it.
4897
4898 See also : "option accept-invalid-http-response" and "show errors" on the
4899 stats socket.
4900
4901
4902option accept-invalid-http-response
4903no option accept-invalid-http-response
4904 Enable or disable relaxing of HTTP response parsing
4905 May be used in sections : defaults | frontend | listen | backend
4906 yes | no | yes | yes
4907 Arguments : none
4908
Willy Tarreau91852eb2015-05-01 13:26:00 +02004909 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02004910 means that invalid characters in header names are not permitted and cause an
4911 error to be returned to the client. This is the desired behaviour as such
4912 forbidden characters are essentially used to build attacks exploiting server
4913 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
4914 server will emit invalid header names for whatever reason (configuration,
4915 implementation) and the issue will not be immediately fixed. In such a case,
4916 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau91852eb2015-05-01 13:26:00 +02004917 even if that does not make sense, by specifying this option. This option also
4918 relaxes the test on the HTTP version format, it allows multiple digits for
4919 both the major and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02004920
4921 This option should never be enabled by default as it hides application bugs
4922 and open security breaches. It should only be deployed after a problem has
4923 been confirmed.
4924
4925 When this option is enabled, erroneous header names will still be accepted in
4926 responses, but the complete response will be captured in order to permit
4927 later analysis using the "show errors" request on the UNIX stats socket.
4928 Doing this also helps confirming that the issue has been solved.
4929
4930 If this option has been enabled in a "defaults" section, it can be disabled
4931 in a specific instance by prepending the "no" keyword before it.
4932
4933 See also : "option accept-invalid-http-request" and "show errors" on the
4934 stats socket.
4935
4936
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004937option allbackups
4938no option allbackups
4939 Use either all backup servers at a time or only the first one
4940 May be used in sections : defaults | frontend | listen | backend
4941 yes | no | yes | yes
4942 Arguments : none
4943
4944 By default, the first operational backup server gets all traffic when normal
4945 servers are all down. Sometimes, it may be preferred to use multiple backups
4946 at once, because one will not be enough. When "option allbackups" is enabled,
4947 the load balancing will be performed among all backup servers when all normal
4948 ones are unavailable. The same load balancing algorithm will be used and the
4949 servers' weights will be respected. Thus, there will not be any priority
4950 order between the backup servers anymore.
4951
4952 This option is mostly used with static server farms dedicated to return a
4953 "sorry" page when an application is completely offline.
4954
4955 If this option has been enabled in a "defaults" section, it can be disabled
4956 in a specific instance by prepending the "no" keyword before it.
4957
4958
4959option checkcache
4960no option checkcache
Godbach7056a352013-12-11 20:01:07 +08004961 Analyze all server responses and block responses with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004962 May be used in sections : defaults | frontend | listen | backend
4963 yes | no | yes | yes
4964 Arguments : none
4965
4966 Some high-level frameworks set application cookies everywhere and do not
4967 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004968 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004969 high risk of session crossing or stealing between users traversing the same
4970 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02004971 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004972
4973 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004974 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01004975 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004976 response to check if there's a risk of caching a cookie on a client-side
4977 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01004978 to the client are :
4979 - all those without "Set-Cookie" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004980 - all those with a return code other than 200, 203, 206, 300, 301, 410,
Willy Tarreau198a7442008-01-17 12:05:32 +01004981 provided that the server has not set a "Cache-control: public" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004982 - all those that come from a POST request, provided that the server has not
4983 set a 'Cache-Control: public' header ;
4984 - those with a 'Pragma: no-cache' header
4985 - those with a 'Cache-control: private' header
4986 - those with a 'Cache-control: no-store' header
4987 - those with a 'Cache-control: max-age=0' header
4988 - those with a 'Cache-control: s-maxage=0' header
4989 - those with a 'Cache-control: no-cache' header
4990 - those with a 'Cache-control: no-cache="set-cookie"' header
4991 - those with a 'Cache-control: no-cache="set-cookie,' header
4992 (allowing other fields after set-cookie)
4993
4994 If a response doesn't respect these requirements, then it will be blocked
Willy Tarreau198a7442008-01-17 12:05:32 +01004995 just as if it was from an "rspdeny" filter, with an "HTTP 502 bad gateway".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004996 The session state shows "PH--" meaning that the proxy blocked the response
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004997 during headers processing. Additionally, an alert will be sent in the logs so
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004998 that admins are informed that there's something to be fixed.
4999
5000 Due to the high impact on the application, the application should be tested
5001 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01005002 good practice to always activate it during tests, even if it is not used in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005003 production, as it will report potentially dangerous application behaviours.
5004
5005 If this option has been enabled in a "defaults" section, it can be disabled
5006 in a specific instance by prepending the "no" keyword before it.
5007
5008
5009option clitcpka
5010no option clitcpka
5011 Enable or disable the sending of TCP keepalive packets on the client side
5012 May be used in sections : defaults | frontend | listen | backend
5013 yes | yes | yes | no
5014 Arguments : none
5015
5016 When there is a firewall or any session-aware component between a client and
5017 a server, and when the protocol involves very long sessions with long idle
5018 periods (eg: remote desktops), there is a risk that one of the intermediate
5019 components decides to expire a session which has remained idle for too long.
5020
5021 Enabling socket-level TCP keep-alives makes the system regularly send packets
5022 to the other end of the connection, leaving it active. The delay between
5023 keep-alive probes is controlled by the system only and depends both on the
5024 operating system and its tuning parameters.
5025
5026 It is important to understand that keep-alive packets are neither emitted nor
5027 received at the application level. It is only the network stacks which sees
5028 them. For this reason, even if one side of the proxy already uses keep-alives
5029 to maintain its connection alive, those keep-alive packets will not be
5030 forwarded to the other side of the proxy.
5031
5032 Please note that this has nothing to do with HTTP keep-alive.
5033
5034 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
5035 client side of a connection, which should help when session expirations are
5036 noticed between HAProxy and a client.
5037
5038 If this option has been enabled in a "defaults" section, it can be disabled
5039 in a specific instance by prepending the "no" keyword before it.
5040
5041 See also : "option srvtcpka", "option tcpka"
5042
5043
Willy Tarreau0ba27502007-12-24 16:55:16 +01005044option contstats
5045 Enable continuous traffic statistics updates
5046 May be used in sections : defaults | frontend | listen | backend
5047 yes | yes | yes | no
5048 Arguments : none
5049
5050 By default, counters used for statistics calculation are incremented
5051 only when a session finishes. It works quite well when serving small
5052 objects, but with big ones (for example large images or archives) or
5053 with A/V streaming, a graph generated from haproxy counters looks like
5054 a hedgehog. With this option enabled counters get incremented continuously,
5055 during a whole session. Recounting touches a hotpath directly so
5056 it is not enabled by default, as it has small performance impact (~0.5%).
5057
5058
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005059option dontlog-normal
5060no option dontlog-normal
5061 Enable or disable logging of normal, successful connections
5062 May be used in sections : defaults | frontend | listen | backend
5063 yes | yes | yes | no
5064 Arguments : none
5065
5066 There are large sites dealing with several thousand connections per second
5067 and for which logging is a major pain. Some of them are even forced to turn
5068 logs off and cannot debug production issues. Setting this option ensures that
5069 normal connections, those which experience no error, no timeout, no retry nor
5070 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
5071 mode, the response status code is checked and return codes 5xx will still be
5072 logged.
5073
5074 It is strongly discouraged to use this option as most of the time, the key to
5075 complex issues is in the normal logs which will not be logged here. If you
5076 need to separate logs, see the "log-separate-errors" option instead.
5077
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005078 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005079 logging.
5080
5081
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005082option dontlognull
5083no option dontlognull
5084 Enable or disable logging of null connections
5085 May be used in sections : defaults | frontend | listen | backend
5086 yes | yes | yes | no
5087 Arguments : none
5088
5089 In certain environments, there are components which will regularly connect to
5090 various systems to ensure that they are still alive. It can be the case from
5091 another load balancer as well as from monitoring systems. By default, even a
5092 simple port probe or scan will produce a log. If those connections pollute
5093 the logs too much, it is possible to enable option "dontlognull" to indicate
5094 that a connection on which no data has been transferred will not be logged,
Willy Tarreau0f228a02015-05-01 15:37:53 +02005095 which typically corresponds to those probes. Note that errors will still be
5096 returned to the client and accounted for in the stats. If this is not what is
5097 desired, option http-ignore-probes can be used instead.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005098
5099 It is generally recommended not to use this option in uncontrolled
5100 environments (eg: internet), otherwise scans and other malicious activities
5101 would not be logged.
5102
5103 If this option has been enabled in a "defaults" section, it can be disabled
5104 in a specific instance by prepending the "no" keyword before it.
5105
Willy Tarreau0f228a02015-05-01 15:37:53 +02005106 See also : "log", "http-ignore-probes", "monitor-net", "monitor-uri", and
5107 section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005108
5109
5110option forceclose
5111no option forceclose
5112 Enable or disable active connection closing after response is transferred.
5113 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaua31e5df2009-12-30 01:10:35 +01005114 yes | yes | yes | yes
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005115 Arguments : none
5116
5117 Some HTTP servers do not necessarily close the connections when they receive
5118 the "Connection: close" set by "option httpclose", and if the client does not
5119 close either, then the connection remains open till the timeout expires. This
5120 causes high number of simultaneous connections on the servers and shows high
5121 global session times in the logs.
5122
5123 When this happens, it is possible to use "option forceclose". It will
Willy Tarreau82eeaf22009-12-29 12:09:05 +01005124 actively close the outgoing server channel as soon as the server has finished
Cyril Bonté653dcd62014-02-20 00:13:15 +01005125 to respond and release some resources earlier than with "option httpclose".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005126
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005127 This option may also be combined with "option http-pretend-keepalive", which
5128 will disable sending of the "Connection: close" header, but will still cause
5129 the connection to be closed once the whole response is received.
5130
Cyril Bonté653dcd62014-02-20 00:13:15 +01005131 This option disables and replaces any previous "option httpclose", "option
5132 http-server-close", "option http-keep-alive", or "option http-tunnel".
Willy Tarreau02bce8b2014-01-30 00:15:28 +01005133
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005134 If this option has been enabled in a "defaults" section, it can be disabled
5135 in a specific instance by prepending the "no" keyword before it.
5136
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005137 See also : "option httpclose" and "option http-pretend-keepalive"
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005138
5139
Willy Tarreau87cf5142011-08-19 22:57:24 +02005140option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01005141 Enable insertion of the X-Forwarded-For header to requests sent to servers
5142 May be used in sections : defaults | frontend | listen | backend
5143 yes | yes | yes | yes
5144 Arguments :
5145 <network> is an optional argument used to disable this option for sources
5146 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02005147 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01005148 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005149
5150 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
5151 their client address. This is sometimes annoying when the client's IP address
5152 is expected in server logs. To solve this problem, the well-known HTTP header
5153 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
5154 This header contains a value representing the client's IP address. Since this
5155 header is always appended at the end of the existing header list, the server
5156 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02005157 the server's manual to find how to enable use of this standard header. Note
5158 that only the last occurrence of the header must be used, since it is really
5159 possible that the client has already brought one.
5160
Willy Tarreaud72758d2010-01-12 10:42:19 +01005161 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02005162 the default "X-Forwarded-For". This can be useful where you might already
Willy Tarreaud72758d2010-01-12 10:42:19 +01005163 have a "X-Forwarded-For" header from a different application (eg: stunnel),
5164 and you need preserve it. Also if your backend server doesn't use the
Ross Westaf72a1d2008-08-03 10:51:45 +02005165 "X-Forwarded-For" header and requires different one (eg: Zeus Web Servers
5166 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01005167
5168 Sometimes, a same HAProxy instance may be shared between a direct client
5169 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
5170 used to decrypt HTTPS traffic). It is possible to disable the addition of the
5171 header for a known source address or network by adding the "except" keyword
5172 followed by the network address. In this case, any source IP matching the
5173 network will not cause an addition of this header. Most common uses are with
5174 private networks or 127.0.0.1.
5175
Willy Tarreau87cf5142011-08-19 22:57:24 +02005176 Alternatively, the keyword "if-none" states that the header will only be
5177 added if it is not present. This should only be used in perfectly trusted
5178 environment, as this might cause a security issue if headers reaching haproxy
5179 are under the control of the end-user.
5180
Willy Tarreauc27debf2008-01-06 08:57:02 +01005181 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02005182 least one of them uses it, the header will be added. Note that the backend's
5183 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02005184 both are defined. In the case of the "if-none" argument, if at least one of
5185 the frontend or the backend does not specify it, it wants the addition to be
5186 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005187
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005188 Example :
Willy Tarreauc27debf2008-01-06 08:57:02 +01005189 # Public HTTP address also used by stunnel on the same machine
5190 frontend www
5191 mode http
5192 option forwardfor except 127.0.0.1 # stunnel already adds the header
5193
Ross Westaf72a1d2008-08-03 10:51:45 +02005194 # Those servers want the IP Address in X-Client
5195 backend www
5196 mode http
5197 option forwardfor header X-Client
5198
Willy Tarreau87cf5142011-08-19 22:57:24 +02005199 See also : "option httpclose", "option http-server-close",
Willy Tarreau70dffda2014-01-30 03:07:23 +01005200 "option forceclose", "option http-keep-alive"
Willy Tarreauc27debf2008-01-06 08:57:02 +01005201
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005202
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005203option http-buffer-request
5204no option http-buffer-request
5205 Enable or disable waiting for whole HTTP request body before proceeding
5206 May be used in sections : defaults | frontend | listen | backend
5207 yes | yes | yes | yes
5208 Arguments : none
5209
5210 It is sometimes desirable to wait for the body of an HTTP request before
5211 taking a decision. This is what is being done by "balance url_param" for
5212 example. The first use case is to buffer requests from slow clients before
5213 connecting to the server. Another use case consists in taking the routing
5214 decision based on the request body's contents. This option placed in a
5215 frontend or backend forces the HTTP processing to wait until either the whole
5216 body is received, or the request buffer is full, or the first chunk is
5217 complete in case of chunked encoding. It can have undesired side effects with
5218 some applications abusing HTTP by expecting unbufferred transmissions between
5219 the frontend and the backend, so this should definitely not be used by
5220 default.
5221
Baptiste Assmanneccdf432015-10-28 13:49:01 +01005222 See also : "option http-no-delay", "timeout http-request"
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005223
5224
Willy Tarreau0f228a02015-05-01 15:37:53 +02005225option http-ignore-probes
5226no option http-ignore-probes
5227 Enable or disable logging of null connections and request timeouts
5228 May be used in sections : defaults | frontend | listen | backend
5229 yes | yes | yes | no
5230 Arguments : none
5231
5232 Recently some browsers started to implement a "pre-connect" feature
5233 consisting in speculatively connecting to some recently visited web sites
5234 just in case the user would like to visit them. This results in many
5235 connections being established to web sites, which end up in 408 Request
5236 Timeout if the timeout strikes first, or 400 Bad Request when the browser
5237 decides to close them first. These ones pollute the log and feed the error
5238 counters. There was already "option dontlognull" but it's insufficient in
5239 this case. Instead, this option does the following things :
5240 - prevent any 400/408 message from being sent to the client if nothing
5241 was received over a connection before it was closed ;
5242 - prevent any log from being emitted in this situation ;
5243 - prevent any error counter from being incremented
5244
5245 That way the empty connection is silently ignored. Note that it is better
5246 not to use this unless it is clear that it is needed, because it will hide
5247 real problems. The most common reason for not receiving a request and seeing
5248 a 408 is due to an MTU inconsistency between the client and an intermediary
5249 element such as a VPN, which blocks too large packets. These issues are
5250 generally seen with POST requests as well as GET with large cookies. The logs
5251 are often the only way to detect them.
5252
5253 If this option has been enabled in a "defaults" section, it can be disabled
5254 in a specific instance by prepending the "no" keyword before it.
5255
5256 See also : "log", "dontlognull", "errorfile", and section 8 about logging.
5257
5258
Willy Tarreau16bfb022010-01-16 19:48:41 +01005259option http-keep-alive
5260no option http-keep-alive
5261 Enable or disable HTTP keep-alive from client to server
5262 May be used in sections : defaults | frontend | listen | backend
5263 yes | yes | yes | yes
5264 Arguments : none
5265
Willy Tarreau70dffda2014-01-30 03:07:23 +01005266 By default HAProxy operates in keep-alive mode with regards to persistent
5267 connections: for each connection it processes each request and response, and
5268 leaves the connection idle on both sides between the end of a response and the
5269 start of a new request. This mode may be changed by several options such as
5270 "option http-server-close", "option forceclose", "option httpclose" or
5271 "option http-tunnel". This option allows to set back the keep-alive mode,
5272 which can be useful when another mode was used in a defaults section.
5273
5274 Setting "option http-keep-alive" enables HTTP keep-alive mode on the client-
5275 and server- sides. This provides the lowest latency on the client side (slow
Willy Tarreau16bfb022010-01-16 19:48:41 +01005276 network) and the fastest session reuse on the server side at the expense
5277 of maintaining idle connections to the servers. In general, it is possible
5278 with this option to achieve approximately twice the request rate that the
5279 "http-server-close" option achieves on small objects. There are mainly two
5280 situations where this option may be useful :
5281
5282 - when the server is non-HTTP compliant and authenticates the connection
5283 instead of requests (eg: NTLM authentication)
5284
5285 - when the cost of establishing the connection to the server is significant
5286 compared to the cost of retrieving the associated object from the server.
5287
5288 This last case can happen when the server is a fast static server of cache.
5289 In this case, the server will need to be properly tuned to support high enough
5290 connection counts because connections will last until the client sends another
5291 request.
5292
5293 If the client request has to go to another backend or another server due to
5294 content switching or the load balancing algorithm, the idle connection will
Willy Tarreau9420b122013-12-15 18:58:25 +01005295 immediately be closed and a new one re-opened. Option "prefer-last-server" is
5296 available to try optimize server selection so that if the server currently
5297 attached to an idle connection is usable, it will be used.
Willy Tarreau16bfb022010-01-16 19:48:41 +01005298
5299 In general it is preferred to use "option http-server-close" with application
5300 servers, and some static servers might benefit from "option http-keep-alive".
5301
5302 At the moment, logs will not indicate whether requests came from the same
5303 session or not. The accept date reported in the logs corresponds to the end
5304 of the previous request, and the request time corresponds to the time spent
5305 waiting for a new request. The keep-alive request time is still bound to the
5306 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
5307 not set.
5308
Cyril Bonté653dcd62014-02-20 00:13:15 +01005309 This option disables and replaces any previous "option httpclose", "option
5310 http-server-close", "option forceclose" or "option http-tunnel". When backend
Willy Tarreau70dffda2014-01-30 03:07:23 +01005311 and frontend options differ, all of these 4 options have precedence over
Cyril Bonté653dcd62014-02-20 00:13:15 +01005312 "option http-keep-alive".
Willy Tarreau16bfb022010-01-16 19:48:41 +01005313
5314 See also : "option forceclose", "option http-server-close",
Willy Tarreau9420b122013-12-15 18:58:25 +01005315 "option prefer-last-server", "option http-pretend-keepalive",
5316 "option httpclose", and "1.1. The HTTP transaction model".
Willy Tarreau16bfb022010-01-16 19:48:41 +01005317
5318
Willy Tarreau96e31212011-05-30 18:10:30 +02005319option http-no-delay
5320no option http-no-delay
5321 Instruct the system to favor low interactive delays over performance in HTTP
5322 May be used in sections : defaults | frontend | listen | backend
5323 yes | yes | yes | yes
5324 Arguments : none
5325
5326 In HTTP, each payload is unidirectional and has no notion of interactivity.
5327 Any agent is expected to queue data somewhat for a reasonably low delay.
5328 There are some very rare server-to-server applications that abuse the HTTP
5329 protocol and expect the payload phase to be highly interactive, with many
5330 interleaved data chunks in both directions within a single request. This is
5331 absolutely not supported by the HTTP specification and will not work across
5332 most proxies or servers. When such applications attempt to do this through
5333 haproxy, it works but they will experience high delays due to the network
5334 optimizations which favor performance by instructing the system to wait for
5335 enough data to be available in order to only send full packets. Typical
5336 delays are around 200 ms per round trip. Note that this only happens with
5337 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
5338 affected.
5339
5340 When "option http-no-delay" is present in either the frontend or the backend
5341 used by a connection, all such optimizations will be disabled in order to
5342 make the exchanges as fast as possible. Of course this offers no guarantee on
5343 the functionality, as it may break at any other place. But if it works via
5344 HAProxy, it will work as fast as possible. This option should never be used
5345 by default, and should never be used at all unless such a buggy application
5346 is discovered. The impact of using this option is an increase of bandwidth
5347 usage and CPU usage, which may significantly lower performance in high
5348 latency environments.
5349
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005350 See also : "option http-buffer-request"
5351
Willy Tarreau96e31212011-05-30 18:10:30 +02005352
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005353option http-pretend-keepalive
5354no option http-pretend-keepalive
5355 Define whether haproxy will announce keepalive to the server or not
5356 May be used in sections : defaults | frontend | listen | backend
5357 yes | yes | yes | yes
5358 Arguments : none
5359
5360 When running with "option http-server-close" or "option forceclose", haproxy
5361 adds a "Connection: close" header to the request forwarded to the server.
5362 Unfortunately, when some servers see this header, they automatically refrain
5363 from using the chunked encoding for responses of unknown length, while this
5364 is totally unrelated. The immediate effect is that this prevents haproxy from
5365 maintaining the client connection alive. A second effect is that a client or
5366 a cache could receive an incomplete response without being aware of it, and
5367 consider the response complete.
5368
5369 By setting "option http-pretend-keepalive", haproxy will make the server
5370 believe it will keep the connection alive. The server will then not fall back
5371 to the abnormal undesired above. When haproxy gets the whole response, it
5372 will close the connection with the server just as it would do with the
5373 "forceclose" option. That way the client gets a normal response and the
5374 connection is correctly closed on the server side.
5375
5376 It is recommended not to enable this option by default, because most servers
5377 will more efficiently close the connection themselves after the last packet,
5378 and release its buffers slightly earlier. Also, the added packet on the
5379 network could slightly reduce the overall peak performance. However it is
5380 worth noting that when this option is enabled, haproxy will have slightly
5381 less work to do. So if haproxy is the bottleneck on the whole architecture,
5382 enabling this option might save a few CPU cycles.
5383
5384 This option may be set both in a frontend and in a backend. It is enabled if
5385 at least one of the frontend or backend holding a connection has it enabled.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005386 This option may be combined with "option httpclose", which will cause
Willy Tarreau22a95342010-09-29 14:31:41 +02005387 keepalive to be announced to the server and close to be announced to the
5388 client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005389
5390 If this option has been enabled in a "defaults" section, it can be disabled
5391 in a specific instance by prepending the "no" keyword before it.
5392
Willy Tarreau16bfb022010-01-16 19:48:41 +01005393 See also : "option forceclose", "option http-server-close", and
5394 "option http-keep-alive"
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005395
Willy Tarreauc27debf2008-01-06 08:57:02 +01005396
Willy Tarreaub608feb2010-01-02 22:47:18 +01005397option http-server-close
5398no option http-server-close
5399 Enable or disable HTTP connection closing on the server side
5400 May be used in sections : defaults | frontend | listen | backend
5401 yes | yes | yes | yes
5402 Arguments : none
5403
Willy Tarreau70dffda2014-01-30 03:07:23 +01005404 By default HAProxy operates in keep-alive mode with regards to persistent
5405 connections: for each connection it processes each request and response, and
5406 leaves the connection idle on both sides between the end of a response and
5407 the start of a new request. This mode may be changed by several options such
5408 as "option http-server-close", "option forceclose", "option httpclose" or
5409 "option http-tunnel". Setting "option http-server-close" enables HTTP
5410 connection-close mode on the server side while keeping the ability to support
5411 HTTP keep-alive and pipelining on the client side. This provides the lowest
5412 latency on the client side (slow network) and the fastest session reuse on
5413 the server side to save server resources, similarly to "option forceclose".
5414 It also permits non-keepalive capable servers to be served in keep-alive mode
5415 to the clients if they conform to the requirements of RFC2616. Please note
5416 that some servers do not always conform to those requirements when they see
5417 "Connection: close" in the request. The effect will be that keep-alive will
5418 never be used. A workaround consists in enabling "option
5419 http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01005420
5421 At the moment, logs will not indicate whether requests came from the same
5422 session or not. The accept date reported in the logs corresponds to the end
5423 of the previous request, and the request time corresponds to the time spent
5424 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01005425 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
5426 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01005427
5428 This option may be set both in a frontend and in a backend. It is enabled if
5429 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01005430 It disables and replaces any previous "option httpclose", "option forceclose",
5431 "option http-tunnel" or "option http-keep-alive". Please check section 4
Willy Tarreau70dffda2014-01-30 03:07:23 +01005432 ("Proxies") to see how this option combines with others when frontend and
5433 backend options differ.
Willy Tarreaub608feb2010-01-02 22:47:18 +01005434
5435 If this option has been enabled in a "defaults" section, it can be disabled
5436 in a specific instance by prepending the "no" keyword before it.
5437
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02005438 See also : "option forceclose", "option http-pretend-keepalive",
Willy Tarreau16bfb022010-01-16 19:48:41 +01005439 "option httpclose", "option http-keep-alive", and
5440 "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01005441
5442
Willy Tarreau02bce8b2014-01-30 00:15:28 +01005443option http-tunnel
5444no option http-tunnel
5445 Disable or enable HTTP connection processing after first transaction
5446 May be used in sections : defaults | frontend | listen | backend
5447 yes | yes | yes | yes
5448 Arguments : none
5449
Willy Tarreau70dffda2014-01-30 03:07:23 +01005450 By default HAProxy operates in keep-alive mode with regards to persistent
5451 connections: for each connection it processes each request and response, and
5452 leaves the connection idle on both sides between the end of a response and
5453 the start of a new request. This mode may be changed by several options such
5454 as "option http-server-close", "option forceclose", "option httpclose" or
5455 "option http-tunnel".
5456
5457 Option "http-tunnel" disables any HTTP processing past the first request and
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005458 the first response. This is the mode which was used by default in versions
Willy Tarreau70dffda2014-01-30 03:07:23 +01005459 1.0 to 1.5-dev21. It is the mode with the lowest processing overhead, which
5460 is normally not needed anymore unless in very specific cases such as when
5461 using an in-house protocol that looks like HTTP but is not compatible, or
5462 just to log one request per client in order to reduce log size. Note that
5463 everything which works at the HTTP level, including header parsing/addition,
5464 cookie processing or content switching will only work for the first request
5465 and will be ignored after the first response.
Willy Tarreau02bce8b2014-01-30 00:15:28 +01005466
5467 If this option has been enabled in a "defaults" section, it can be disabled
5468 in a specific instance by prepending the "no" keyword before it.
5469
5470 See also : "option forceclose", "option http-server-close",
5471 "option httpclose", "option http-keep-alive", and
5472 "1.1. The HTTP transaction model".
5473
5474
Willy Tarreau88d349d2010-01-25 12:15:43 +01005475option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01005476no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01005477 Make use of non-standard Proxy-Connection header instead of Connection
5478 May be used in sections : defaults | frontend | listen | backend
5479 yes | yes | yes | no
5480 Arguments : none
5481
5482 While RFC2616 explicitly states that HTTP/1.1 agents must use the
5483 Connection header to indicate their wish of persistent or non-persistent
5484 connections, both browsers and proxies ignore this header for proxied
5485 connections and make use of the undocumented, non-standard Proxy-Connection
5486 header instead. The issue begins when trying to put a load balancer between
5487 browsers and such proxies, because there will be a difference between what
5488 haproxy understands and what the client and the proxy agree on.
5489
5490 By setting this option in a frontend, haproxy can automatically switch to use
5491 that non-standard header if it sees proxied requests. A proxied request is
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01005492 defined here as one where the URI begins with neither a '/' nor a '*'. This
5493 is incompatible with the HTTP tunnel mode. Note that this option can only be
5494 specified in a frontend and will affect the request along its whole life.
Willy Tarreau88d349d2010-01-25 12:15:43 +01005495
Willy Tarreau844a7e72010-01-31 21:46:18 +01005496 Also, when this option is set, a request which requires authentication will
5497 automatically switch to use proxy authentication headers if it is itself a
5498 proxied request. That makes it possible to check or enforce authentication in
5499 front of an existing proxy.
5500
Willy Tarreau88d349d2010-01-25 12:15:43 +01005501 This option should normally never be used, except in front of a proxy.
5502
5503 See also : "option httpclose", "option forceclose" and "option
5504 http-server-close".
5505
5506
Willy Tarreaud63335a2010-02-26 12:56:52 +01005507option httpchk
5508option httpchk <uri>
5509option httpchk <method> <uri>
5510option httpchk <method> <uri> <version>
5511 Enable HTTP protocol to check on the servers health
5512 May be used in sections : defaults | frontend | listen | backend
5513 yes | no | yes | yes
5514 Arguments :
5515 <method> is the optional HTTP method used with the requests. When not set,
5516 the "OPTIONS" method is used, as it generally requires low server
5517 processing and is easy to filter out from the logs. Any method
5518 may be used, though it is not recommended to invent non-standard
5519 ones.
5520
5521 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
5522 which is accessible by default on almost any server, but may be
5523 changed to any other URI. Query strings are permitted.
5524
5525 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
5526 but some servers might behave incorrectly in HTTP 1.0, so turning
5527 it to HTTP/1.1 may sometimes help. Note that the Host field is
5528 mandatory in HTTP/1.1, and as a trick, it is possible to pass it
5529 after "\r\n" following the version string.
5530
5531 By default, server health checks only consist in trying to establish a TCP
5532 connection. When "option httpchk" is specified, a complete HTTP request is
5533 sent once the TCP connection is established, and responses 2xx and 3xx are
5534 considered valid, while all other ones indicate a server failure, including
5535 the lack of any response.
5536
5537 The port and interval are specified in the server configuration.
5538
5539 This option does not necessarily require an HTTP backend, it also works with
5540 plain TCP backends. This is particularly useful to check simple scripts bound
5541 to some dedicated ports using the inetd daemon.
5542
5543 Examples :
5544 # Relay HTTPS traffic to Apache instance and check service availability
5545 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
5546 backend https_relay
5547 mode tcp
5548 option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www
5549 server apache1 192.168.1.1:443 check port 80
5550
Simon Hormanafc47ee2013-11-25 10:46:35 +09005551 See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",
5552 "option pgsql-check", "http-check" and the "check", "port" and
5553 "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01005554
5555
Willy Tarreauc27debf2008-01-06 08:57:02 +01005556option httpclose
5557no option httpclose
5558 Enable or disable passive HTTP connection closing
5559 May be used in sections : defaults | frontend | listen | backend
5560 yes | yes | yes | yes
5561 Arguments : none
5562
Willy Tarreau70dffda2014-01-30 03:07:23 +01005563 By default HAProxy operates in keep-alive mode with regards to persistent
5564 connections: for each connection it processes each request and response, and
5565 leaves the connection idle on both sides between the end of a response and
5566 the start of a new request. This mode may be changed by several options such
Cyril Bonté653dcd62014-02-20 00:13:15 +01005567 as "option http-server-close", "option forceclose", "option httpclose" or
Willy Tarreau70dffda2014-01-30 03:07:23 +01005568 "option http-tunnel".
5569
5570 If "option httpclose" is set, HAProxy will work in HTTP tunnel mode and check
5571 if a "Connection: close" header is already set in each direction, and will
5572 add one if missing. Each end should react to this by actively closing the TCP
5573 connection after each transfer, thus resulting in a switch to the HTTP close
5574 mode. Any "Connection" header different from "close" will also be removed.
5575 Note that this option is deprecated since what it does is very cheap but not
5576 reliable. Using "option http-server-close" or "option forceclose" is strongly
5577 recommended instead.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005578
5579 It seldom happens that some servers incorrectly ignore this header and do not
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005580 close the connection even though they reply "Connection: close". For this
Willy Tarreau0dfdf192010-01-05 11:33:11 +01005581 reason, they are not compatible with older HTTP 1.0 browsers. If this happens
5582 it is possible to use the "option forceclose" which actively closes the
5583 request connection once the server responds. Option "forceclose" also
5584 releases the server connection earlier because it does not have to wait for
5585 the client to acknowledge it.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005586
5587 This option may be set both in a frontend and in a backend. It is enabled if
5588 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01005589 It disables and replaces any previous "option http-server-close",
5590 "option forceclose", "option http-keep-alive" or "option http-tunnel". Please
Willy Tarreau70dffda2014-01-30 03:07:23 +01005591 check section 4 ("Proxies") to see how this option combines with others when
5592 frontend and backend options differ.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005593
5594 If this option has been enabled in a "defaults" section, it can be disabled
5595 in a specific instance by prepending the "no" keyword before it.
5596
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02005597 See also : "option forceclose", "option http-server-close" and
5598 "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01005599
5600
Emeric Brun3a058f32009-06-30 18:26:00 +02005601option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01005602 Enable logging of HTTP request, session state and timers
5603 May be used in sections : defaults | frontend | listen | backend
5604 yes | yes | yes | yes
Emeric Brun3a058f32009-06-30 18:26:00 +02005605 Arguments :
5606 clf if the "clf" argument is added, then the output format will be
5607 the CLF format instead of HAProxy's default HTTP format. You can
5608 use this when you need to feed HAProxy's logs through a specific
5609 log analyser which only support the CLF format and which is not
5610 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005611
5612 By default, the log output format is very poor, as it only contains the
5613 source and destination addresses, and the instance name. By specifying
5614 "option httplog", each log line turns into a much richer format including,
5615 but not limited to, the HTTP request, the connection timers, the session
5616 status, the connections numbers, the captured headers and cookies, the
5617 frontend, backend and server name, and of course the source address and
5618 ports.
5619
5620 This option may be set either in the frontend or the backend.
5621
PiBa-NLbd556bf2014-12-11 21:31:54 +01005622 Specifying only "option httplog" will automatically clear the 'clf' mode
5623 if it was set by default.
Emeric Brun3a058f32009-06-30 18:26:00 +02005624
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005625 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005626
Willy Tarreau55165fe2009-05-10 12:02:55 +02005627
5628option http_proxy
5629no option http_proxy
5630 Enable or disable plain HTTP proxy mode
5631 May be used in sections : defaults | frontend | listen | backend
5632 yes | yes | yes | yes
5633 Arguments : none
5634
5635 It sometimes happens that people need a pure HTTP proxy which understands
5636 basic proxy requests without caching nor any fancy feature. In this case,
5637 it may be worth setting up an HAProxy instance with the "option http_proxy"
5638 set. In this mode, no server is declared, and the connection is forwarded to
5639 the IP address and port found in the URL after the "http://" scheme.
5640
5641 No host address resolution is performed, so this only works when pure IP
5642 addresses are passed. Since this option's usage perimeter is rather limited,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01005643 it will probably be used only by experts who know they need exactly it. This
5644 is incompatible with the HTTP tunnel mode.
Willy Tarreau55165fe2009-05-10 12:02:55 +02005645
5646 If this option has been enabled in a "defaults" section, it can be disabled
5647 in a specific instance by prepending the "no" keyword before it.
5648
5649 Example :
5650 # this backend understands HTTP proxy requests and forwards them directly.
5651 backend direct_forward
5652 option httpclose
5653 option http_proxy
5654
5655 See also : "option httpclose"
5656
Willy Tarreau211ad242009-10-03 21:45:07 +02005657
Jamie Gloudon801a0a32012-08-25 00:18:33 -04005658option independent-streams
5659no option independent-streams
5660 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02005661 May be used in sections : defaults | frontend | listen | backend
5662 yes | yes | yes | yes
5663 Arguments : none
5664
5665 By default, when data is sent over a socket, both the write timeout and the
5666 read timeout for that socket are refreshed, because we consider that there is
5667 activity on that socket, and we have no other means of guessing if we should
5668 receive data or not.
5669
5670 While this default behaviour is desirable for almost all applications, there
5671 exists a situation where it is desirable to disable it, and only refresh the
5672 read timeout if there are incoming data. This happens on sessions with large
5673 timeouts and low amounts of exchanged data such as telnet session. If the
5674 server suddenly disappears, the output data accumulates in the system's
5675 socket buffers, both timeouts are correctly refreshed, and there is no way
5676 to know the server does not receive them, so we don't timeout. However, when
5677 the underlying protocol always echoes sent data, it would be enough by itself
5678 to detect the issue using the read timeout. Note that this problem does not
5679 happen with more verbose protocols because data won't accumulate long in the
5680 socket buffers.
5681
5682 When this option is set on the frontend, it will disable read timeout updates
5683 on data sent to the client. There probably is little use of this case. When
5684 the option is set on the backend, it will disable read timeout updates on
5685 data sent to the server. Doing so will typically break large HTTP posts from
5686 slow lines, so use it with caution.
5687
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005688 Note: older versions used to call this setting "option independent-streams"
Jamie Gloudon801a0a32012-08-25 00:18:33 -04005689 with a spelling mistake. This spelling is still supported but
5690 deprecated.
5691
Willy Tarreauce887fd2012-05-12 12:50:00 +02005692 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02005693
5694
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02005695option ldap-check
5696 Use LDAPv3 health checks for server testing
5697 May be used in sections : defaults | frontend | listen | backend
5698 yes | no | yes | yes
5699 Arguments : none
5700
5701 It is possible to test that the server correctly talks LDAPv3 instead of just
5702 testing that it accepts the TCP connection. When this option is set, an
5703 LDAPv3 anonymous simple bind message is sent to the server, and the response
5704 is analyzed to find an LDAPv3 bind response message.
5705
5706 The server is considered valid only when the LDAP response contains success
5707 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
5708
5709 Logging of bind requests is server dependent see your documentation how to
5710 configure it.
5711
5712 Example :
5713 option ldap-check
5714
5715 See also : "option httpchk"
5716
5717
Simon Horman98637e52014-06-20 12:30:16 +09005718option external-check
5719 Use external processes for server health checks
5720 May be used in sections : defaults | frontend | listen | backend
5721 yes | no | yes | yes
5722
5723 It is possible to test the health of a server using an external command.
5724 This is achieved by running the executable set using "external-check
5725 command".
5726
5727 Requires the "external-check" global to be set.
5728
5729 See also : "external-check", "external-check command", "external-check path"
5730
5731
Willy Tarreau211ad242009-10-03 21:45:07 +02005732option log-health-checks
5733no option log-health-checks
Willy Tarreaubef1b322014-05-13 21:01:39 +02005734 Enable or disable logging of health checks status updates
Willy Tarreau211ad242009-10-03 21:45:07 +02005735 May be used in sections : defaults | frontend | listen | backend
5736 yes | no | yes | yes
5737 Arguments : none
5738
Willy Tarreaubef1b322014-05-13 21:01:39 +02005739 By default, failed health check are logged if server is UP and successful
5740 health checks are logged if server is DOWN, so the amount of additional
5741 information is limited.
Willy Tarreau211ad242009-10-03 21:45:07 +02005742
Willy Tarreaubef1b322014-05-13 21:01:39 +02005743 When this option is enabled, any change of the health check status or to
5744 the server's health will be logged, so that it becomes possible to know
5745 that a server was failing occasional checks before crashing, or exactly when
5746 it failed to respond a valid HTTP status, then when the port started to
5747 reject connections, then when the server stopped responding at all.
5748
5749 Note that status changes not caused by health checks (eg: enable/disable on
5750 the CLI) are intentionally not logged by this option.
Willy Tarreau211ad242009-10-03 21:45:07 +02005751
Willy Tarreaubef1b322014-05-13 21:01:39 +02005752 See also: "option httpchk", "option ldap-check", "option mysql-check",
5753 "option pgsql-check", "option redis-check", "option smtpchk",
5754 "option tcp-check", "log" and section 8 about logging.
Willy Tarreau211ad242009-10-03 21:45:07 +02005755
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005756
5757option log-separate-errors
5758no option log-separate-errors
5759 Change log level for non-completely successful connections
5760 May be used in sections : defaults | frontend | listen | backend
5761 yes | yes | yes | no
5762 Arguments : none
5763
5764 Sometimes looking for errors in logs is not easy. This option makes haproxy
5765 raise the level of logs containing potentially interesting information such
5766 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
5767 level changes from "info" to "err". This makes it possible to log them
5768 separately to a different file with most syslog daemons. Be careful not to
5769 remove them from the original file, otherwise you would lose ordering which
5770 provides very important information.
5771
5772 Using this option, large sites dealing with several thousand connections per
5773 second may log normal traffic to a rotating buffer and only archive smaller
5774 error logs.
5775
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005776 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005777 logging.
5778
Willy Tarreauc27debf2008-01-06 08:57:02 +01005779
5780option logasap
5781no option logasap
5782 Enable or disable early logging of HTTP requests
5783 May be used in sections : defaults | frontend | listen | backend
5784 yes | yes | yes | no
5785 Arguments : none
5786
5787 By default, HTTP requests are logged upon termination so that the total
5788 transfer time and the number of bytes appear in the logs. When large objects
5789 are being transferred, it may take a while before the request appears in the
5790 logs. Using "option logasap", the request gets logged as soon as the server
5791 sends the complete headers. The only missing information in the logs will be
5792 the total number of bytes which will indicate everything except the amount
5793 of data transferred, and the total time which will not take the transfer
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01005794 time into account. In such a situation, it's a good practice to capture the
Willy Tarreauc27debf2008-01-06 08:57:02 +01005795 "Content-Length" response header so that the logs at least indicate how many
5796 bytes are expected to be transferred.
5797
Willy Tarreaucc6c8912009-02-22 10:53:55 +01005798 Examples :
5799 listen http_proxy 0.0.0.0:80
5800 mode http
5801 option httplog
5802 option logasap
5803 log 192.168.2.200 local3
5804
5805 >>> Feb 6 12:14:14 localhost \
5806 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
5807 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
5808 "GET /image.iso HTTP/1.0"
5809
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005810 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01005811 logging.
5812
5813
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02005814option mysql-check [ user <username> [ post-41 ] ]
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005815 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005816 May be used in sections : defaults | frontend | listen | backend
5817 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005818 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005819 <username> This is the username which will be used when connecting to MySQL
5820 server.
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02005821 post-41 Send post v4.1 client compatible checks
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005822
5823 If you specify a username, the check consists of sending two MySQL packet,
5824 one Client Authentication packet, and one QUIT packet, to correctly close
5825 MySQL session. We then parse the MySQL Handshake Initialisation packet and/or
5826 Error packet. It is a basic but useful test which does not produce error nor
5827 aborted connect on the server. However, it requires adding an authorization
5828 in the MySQL table, like this :
5829
5830 USE mysql;
5831 INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');
5832 FLUSH PRIVILEGES;
5833
5834 If you don't specify a username (it is deprecated and not recommended), the
5835 check only consists in parsing the Mysql Handshake Initialisation packet or
5836 Error packet, we don't send anything in this mode. It was reported that it
5837 can generate lockout if check is too frequent and/or if there is not enough
5838 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
5839 value as if a connection is established successfully within fewer than MySQL
5840 "max_connect_errors" attempts after a previous connection was interrupted,
5841 the error count for the host is cleared to zero. If HAProxy's server get
5842 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
5843
5844 Remember that this does not check database presence nor database consistency.
5845 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005846
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02005847 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005848
5849 Most often, an incoming MySQL server needs to see the client's IP address for
5850 various purposes, including IP privilege matching and connection logging.
5851 When possible, it is often wise to masquerade the client's IP address when
5852 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02005853 which requires the transparent proxy feature to be compiled in, and the MySQL
5854 server to route the client via the machine hosting haproxy.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005855
5856 See also: "option httpchk"
5857
5858
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005859option nolinger
5860no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005861 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005862 May be used in sections: defaults | frontend | listen | backend
5863 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005864 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005865
5866 When clients or servers abort connections in a dirty way (eg: they are
5867 physically disconnected), the session timeouts triggers and the session is
5868 closed. But it will remain in FIN_WAIT1 state for some time in the system,
5869 using some resources and possibly limiting the ability to establish newer
5870 connections.
5871
5872 When this happens, it is possible to activate "option nolinger" which forces
5873 the system to immediately remove any socket's pending data on close. Thus,
5874 the session is instantly purged from the system's tables. This usually has
5875 side effects such as increased number of TCP resets due to old retransmits
5876 getting immediately rejected. Some firewalls may sometimes complain about
5877 this too.
5878
5879 For this reason, it is not recommended to use this option when not absolutely
5880 needed. You know that you need it when you have thousands of FIN_WAIT1
5881 sessions on your system (TIME_WAIT ones do not count).
5882
5883 This option may be used both on frontends and backends, depending on the side
5884 where it is required. Use it on the frontend for clients, and on the backend
5885 for servers.
5886
5887 If this option has been enabled in a "defaults" section, it can be disabled
5888 in a specific instance by prepending the "no" keyword before it.
5889
5890
Willy Tarreau55165fe2009-05-10 12:02:55 +02005891option originalto [ except <network> ] [ header <name> ]
5892 Enable insertion of the X-Original-To header to requests sent to servers
5893 May be used in sections : defaults | frontend | listen | backend
5894 yes | yes | yes | yes
5895 Arguments :
5896 <network> is an optional argument used to disable this option for sources
5897 matching <network>
5898 <name> an optional argument to specify a different "X-Original-To"
5899 header name.
5900
5901 Since HAProxy can work in transparent mode, every request from a client can
5902 be redirected to the proxy and HAProxy itself can proxy every request to a
5903 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
5904 be lost. This is annoying when you want access rules based on destination ip
5905 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
5906 added by HAProxy to all requests sent to the server. This header contains a
5907 value representing the original destination IP address. Since this must be
5908 configured to always use the last occurrence of this header only. Note that
5909 only the last occurrence of the header must be used, since it is really
5910 possible that the client has already brought one.
5911
5912 The keyword "header" may be used to supply a different header name to replace
5913 the default "X-Original-To". This can be useful where you might already
5914 have a "X-Original-To" header from a different application, and you need
5915 preserve it. Also if your backend server doesn't use the "X-Original-To"
5916 header and requires different one.
5917
5918 Sometimes, a same HAProxy instance may be shared between a direct client
5919 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
5920 used to decrypt HTTPS traffic). It is possible to disable the addition of the
5921 header for a known source address or network by adding the "except" keyword
5922 followed by the network address. In this case, any source IP matching the
5923 network will not cause an addition of this header. Most common uses are with
5924 private networks or 127.0.0.1.
5925
5926 This option may be specified either in the frontend or in the backend. If at
5927 least one of them uses it, the header will be added. Note that the backend's
5928 setting of the header subargument takes precedence over the frontend's if
5929 both are defined.
5930
Willy Tarreau55165fe2009-05-10 12:02:55 +02005931 Examples :
5932 # Original Destination address
5933 frontend www
5934 mode http
5935 option originalto except 127.0.0.1
5936
5937 # Those servers want the IP Address in X-Client-Dst
5938 backend www
5939 mode http
5940 option originalto header X-Client-Dst
5941
Willy Tarreau87cf5142011-08-19 22:57:24 +02005942 See also : "option httpclose", "option http-server-close",
5943 "option forceclose"
Willy Tarreau55165fe2009-05-10 12:02:55 +02005944
5945
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005946option persist
5947no option persist
5948 Enable or disable forced persistence on down servers
5949 May be used in sections: defaults | frontend | listen | backend
5950 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005951 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005952
5953 When an HTTP request reaches a backend with a cookie which references a dead
5954 server, by default it is redispatched to another server. It is possible to
5955 force the request to be sent to the dead server first using "option persist"
5956 if absolutely needed. A common use case is when servers are under extreme
5957 load and spend their time flapping. In this case, the users would still be
5958 directed to the server they opened the session on, in the hope they would be
5959 correctly served. It is recommended to use "option redispatch" in conjunction
5960 with this option so that in the event it would not be possible to connect to
5961 the server at all (server definitely dead), the client would finally be
5962 redirected to another valid server.
5963
5964 If this option has been enabled in a "defaults" section, it can be disabled
5965 in a specific instance by prepending the "no" keyword before it.
5966
Willy Tarreau4de91492010-01-22 19:10:05 +01005967 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005968
5969
Willy Tarreau0c122822013-12-15 18:49:01 +01005970option pgsql-check [ user <username> ]
5971 Use PostgreSQL health checks for server testing
5972 May be used in sections : defaults | frontend | listen | backend
5973 yes | no | yes | yes
5974 Arguments :
5975 <username> This is the username which will be used when connecting to
5976 PostgreSQL server.
5977
5978 The check sends a PostgreSQL StartupMessage and waits for either
5979 Authentication request or ErrorResponse message. It is a basic but useful
5980 test which does not produce error nor aborted connect on the server.
5981 This check is identical with the "mysql-check".
5982
5983 See also: "option httpchk"
5984
5985
Willy Tarreau9420b122013-12-15 18:58:25 +01005986option prefer-last-server
5987no option prefer-last-server
5988 Allow multiple load balanced requests to remain on the same server
5989 May be used in sections: defaults | frontend | listen | backend
5990 yes | no | yes | yes
5991 Arguments : none
5992
5993 When the load balancing algorithm in use is not deterministic, and a previous
5994 request was sent to a server to which haproxy still holds a connection, it is
5995 sometimes desirable that subsequent requests on a same session go to the same
5996 server as much as possible. Note that this is different from persistence, as
5997 we only indicate a preference which haproxy tries to apply without any form
5998 of warranty. The real use is for keep-alive connections sent to servers. When
5999 this option is used, haproxy will try to reuse the same connection that is
6000 attached to the server instead of rebalancing to another server, causing a
6001 close of the connection. This can make sense for static file servers. It does
Willy Tarreau068621e2013-12-23 15:11:25 +01006002 not make much sense to use this in combination with hashing algorithms. Note,
6003 haproxy already automatically tries to stick to a server which sends a 401 or
6004 to a proxy which sends a 407 (authentication required). This is mandatory for
6005 use with the broken NTLM authentication challenge, and significantly helps in
6006 troubleshooting some faulty applications. Option prefer-last-server might be
6007 desirable in these environments as well, to avoid redistributing the traffic
6008 after every other response.
Willy Tarreau9420b122013-12-15 18:58:25 +01006009
6010 If this option has been enabled in a "defaults" section, it can be disabled
6011 in a specific instance by prepending the "no" keyword before it.
6012
6013 See also: "option http-keep-alive"
6014
6015
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006016option redispatch
Joseph Lynch726ab712015-05-11 23:25:34 -07006017option redispatch <interval>
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006018no option redispatch
6019 Enable or disable session redistribution in case of connection failure
6020 May be used in sections: defaults | frontend | listen | backend
6021 yes | no | yes | yes
Joseph Lynch726ab712015-05-11 23:25:34 -07006022 Arguments :
6023 <interval> The optional integer value that controls how often redispatches
6024 occur when retrying connections. Positive value P indicates a
6025 redispatch is desired on every Pth retry, and negative value
6026 N indicate a redispath is desired on the Nth retry prior to the
6027 last retry. For example, the default of -1 preserves the
6028 historical behaviour of redispatching on the last retry, a
6029 positive value of 1 would indicate a redispatch on every retry,
6030 and a positive value of 3 would indicate a redispatch on every
6031 third retry. You can disable redispatches with a value of 0.
6032
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006033
6034 In HTTP mode, if a server designated by a cookie is down, clients may
6035 definitely stick to it because they cannot flush the cookie, so they will not
6036 be able to access the service anymore.
6037
6038 Specifying "option redispatch" will allow the proxy to break their
6039 persistence and redistribute them to a working server.
6040
Joseph Lynch726ab712015-05-11 23:25:34 -07006041 It also allows to retry connections to another server in case of multiple
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006042 connection failures. Of course, it requires having "retries" set to a nonzero
6043 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01006044
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006045 This form is the preferred form, which replaces both the "redispatch" and
6046 "redisp" keywords.
6047
6048 If this option has been enabled in a "defaults" section, it can be disabled
6049 in a specific instance by prepending the "no" keyword before it.
6050
Willy Tarreau4de91492010-01-22 19:10:05 +01006051 See also : "redispatch", "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006052
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006053
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02006054option redis-check
6055 Use redis health checks for server testing
6056 May be used in sections : defaults | frontend | listen | backend
6057 yes | no | yes | yes
6058 Arguments : none
6059
6060 It is possible to test that the server correctly talks REDIS protocol instead
6061 of just testing that it accepts the TCP connection. When this option is set,
6062 a PING redis command is sent to the server, and the response is analyzed to
6063 find the "+PONG" response message.
6064
6065 Example :
6066 option redis-check
6067
6068 See also : "option httpchk"
6069
6070
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006071option smtpchk
6072option smtpchk <hello> <domain>
6073 Use SMTP health checks for server testing
6074 May be used in sections : defaults | frontend | listen | backend
6075 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01006076 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006077 <hello> is an optional argument. It is the "hello" command to use. It can
6078 be either "HELO" (for SMTP) or "EHLO" (for ESTMP). All other
6079 values will be turned into the default command ("HELO").
6080
6081 <domain> is the domain name to present to the server. It may only be
6082 specified (and is mandatory) if the hello command has been
6083 specified. By default, "localhost" is used.
6084
6085 When "option smtpchk" is set, the health checks will consist in TCP
6086 connections followed by an SMTP command. By default, this command is
6087 "HELO localhost". The server's return code is analyzed and only return codes
6088 starting with a "2" will be considered as valid. All other responses,
6089 including a lack of response will constitute an error and will indicate a
6090 dead server.
6091
6092 This test is meant to be used with SMTP servers or relays. Depending on the
6093 request, it is possible that some servers do not log each connection attempt,
6094 so you may want to experiment to improve the behaviour. Using telnet on port
6095 25 is often easier than adjusting the configuration.
6096
6097 Most often, an incoming SMTP server needs to see the client's IP address for
6098 various purposes, including spam filtering, anti-spoofing and logging. When
6099 possible, it is often wise to masquerade the client's IP address when
6100 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02006101 which requires the transparent proxy feature to be compiled in.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006102
6103 Example :
6104 option smtpchk HELO mydomain.org
6105
6106 See also : "option httpchk", "source"
6107
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006108
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02006109option socket-stats
6110no option socket-stats
6111
6112 Enable or disable collecting & providing separate statistics for each socket.
6113 May be used in sections : defaults | frontend | listen | backend
6114 yes | yes | yes | no
6115
6116 Arguments : none
6117
6118
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006119option splice-auto
6120no option splice-auto
6121 Enable or disable automatic kernel acceleration on sockets in both directions
6122 May be used in sections : defaults | frontend | listen | backend
6123 yes | yes | yes | yes
6124 Arguments : none
6125
6126 When this option is enabled either on a frontend or on a backend, haproxy
6127 will automatically evaluate the opportunity to use kernel tcp splicing to
6128 forward data between the client and the server, in either direction. Haproxy
6129 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006130 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006131 are not much aggressive in order to limit excessive use of splicing. This
6132 option requires splicing to be enabled at compile time, and may be globally
6133 disabled with the global option "nosplice". Since splice uses pipes, using it
6134 requires that there are enough spare pipes.
6135
6136 Important note: kernel-based TCP splicing is a Linux-specific feature which
6137 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
6138 transfer data between sockets without copying these data to user-space, thus
6139 providing noticeable performance gains and CPU cycles savings. Since many
6140 early implementations are buggy, corrupt data and/or are inefficient, this
6141 feature is not enabled by default, and it should be used with extreme care.
6142 While it is not possible to detect the correctness of an implementation,
6143 2.6.29 is the first version offering a properly working implementation. In
6144 case of doubt, splicing may be globally disabled using the global "nosplice"
6145 keyword.
6146
6147 Example :
6148 option splice-auto
6149
6150 If this option has been enabled in a "defaults" section, it can be disabled
6151 in a specific instance by prepending the "no" keyword before it.
6152
6153 See also : "option splice-request", "option splice-response", and global
6154 options "nosplice" and "maxpipes"
6155
6156
6157option splice-request
6158no option splice-request
6159 Enable or disable automatic kernel acceleration on sockets for requests
6160 May be used in sections : defaults | frontend | listen | backend
6161 yes | yes | yes | yes
6162 Arguments : none
6163
6164 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04006165 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006166 the client to the server. It might still use the recv/send scheme if there
6167 are no spare pipes left. This option requires splicing to be enabled at
6168 compile time, and may be globally disabled with the global option "nosplice".
6169 Since splice uses pipes, using it requires that there are enough spare pipes.
6170
6171 Important note: see "option splice-auto" for usage limitations.
6172
6173 Example :
6174 option splice-request
6175
6176 If this option has been enabled in a "defaults" section, it can be disabled
6177 in a specific instance by prepending the "no" keyword before it.
6178
6179 See also : "option splice-auto", "option splice-response", and global options
6180 "nosplice" and "maxpipes"
6181
6182
6183option splice-response
6184no option splice-response
6185 Enable or disable automatic kernel acceleration on sockets for responses
6186 May be used in sections : defaults | frontend | listen | backend
6187 yes | yes | yes | yes
6188 Arguments : none
6189
6190 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04006191 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006192 the server to the client. It might still use the recv/send scheme if there
6193 are no spare pipes left. This option requires splicing to be enabled at
6194 compile time, and may be globally disabled with the global option "nosplice".
6195 Since splice uses pipes, using it requires that there are enough spare pipes.
6196
6197 Important note: see "option splice-auto" for usage limitations.
6198
6199 Example :
6200 option splice-response
6201
6202 If this option has been enabled in a "defaults" section, it can be disabled
6203 in a specific instance by prepending the "no" keyword before it.
6204
6205 See also : "option splice-auto", "option splice-request", and global options
6206 "nosplice" and "maxpipes"
6207
6208
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006209option srvtcpka
6210no option srvtcpka
6211 Enable or disable the sending of TCP keepalive packets on the server side
6212 May be used in sections : defaults | frontend | listen | backend
6213 yes | no | yes | yes
6214 Arguments : none
6215
6216 When there is a firewall or any session-aware component between a client and
6217 a server, and when the protocol involves very long sessions with long idle
6218 periods (eg: remote desktops), there is a risk that one of the intermediate
6219 components decides to expire a session which has remained idle for too long.
6220
6221 Enabling socket-level TCP keep-alives makes the system regularly send packets
6222 to the other end of the connection, leaving it active. The delay between
6223 keep-alive probes is controlled by the system only and depends both on the
6224 operating system and its tuning parameters.
6225
6226 It is important to understand that keep-alive packets are neither emitted nor
6227 received at the application level. It is only the network stacks which sees
6228 them. For this reason, even if one side of the proxy already uses keep-alives
6229 to maintain its connection alive, those keep-alive packets will not be
6230 forwarded to the other side of the proxy.
6231
6232 Please note that this has nothing to do with HTTP keep-alive.
6233
6234 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
6235 server side of a connection, which should help when session expirations are
6236 noticed between HAProxy and a server.
6237
6238 If this option has been enabled in a "defaults" section, it can be disabled
6239 in a specific instance by prepending the "no" keyword before it.
6240
6241 See also : "option clitcpka", "option tcpka"
6242
6243
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006244option ssl-hello-chk
6245 Use SSLv3 client hello health checks for server testing
6246 May be used in sections : defaults | frontend | listen | backend
6247 yes | no | yes | yes
6248 Arguments : none
6249
6250 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
6251 possible to test that the server correctly talks SSL instead of just testing
6252 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
6253 SSLv3 client hello messages are sent once the connection is established to
6254 the server, and the response is analyzed to find an SSL server hello message.
6255 The server is considered valid only when the response contains this server
6256 hello message.
6257
6258 All servers tested till there correctly reply to SSLv3 client hello messages,
6259 and most servers tested do not even log the requests containing only hello
6260 messages, which is appreciable.
6261
Willy Tarreau763a95b2012-10-04 23:15:39 +02006262 Note that this check works even when SSL support was not built into haproxy
6263 because it forges the SSL message. When SSL support is available, it is best
6264 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006265
Willy Tarreau763a95b2012-10-04 23:15:39 +02006266 See also: "option httpchk", "check-ssl"
6267
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006268
Willy Tarreaued179852013-12-16 01:07:00 +01006269option tcp-check
6270 Perform health checks using tcp-check send/expect sequences
6271 May be used in sections: defaults | frontend | listen | backend
6272 yes | no | yes | yes
6273
6274 This health check method is intended to be combined with "tcp-check" command
6275 lists in order to support send/expect types of health check sequences.
6276
6277 TCP checks currently support 4 modes of operations :
6278 - no "tcp-check" directive : the health check only consists in a connection
6279 attempt, which remains the default mode.
6280
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006281 - "tcp-check send" or "tcp-check send-binary" only is mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01006282 used to send a string along with a connection opening. With some
6283 protocols, it helps sending a "QUIT" message for example that prevents
6284 the server from logging a connection error for each health check. The
6285 check result will still be based on the ability to open the connection
6286 only.
6287
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006288 - "tcp-check expect" only is mentioned : this is used to test a banner.
Willy Tarreaued179852013-12-16 01:07:00 +01006289 The connection is opened and haproxy waits for the server to present some
6290 contents which must validate some rules. The check result will be based
6291 on the matching between the contents and the rules. This is suited for
6292 POP, IMAP, SMTP, FTP, SSH, TELNET.
6293
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006294 - both "tcp-check send" and "tcp-check expect" are mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01006295 used to test a hello-type protocol. Haproxy sends a message, the server
6296 responds and its response is analysed. the check result will be based on
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006297 the matching between the response contents and the rules. This is often
Willy Tarreaued179852013-12-16 01:07:00 +01006298 suited for protocols which require a binding or a request/response model.
6299 LDAP, MySQL, Redis and SSL are example of such protocols, though they
6300 already all have their dedicated checks with a deeper understanding of
6301 the respective protocols.
6302 In this mode, many questions may be sent and many answers may be
6303 analysed.
6304
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006305 A fifth mode can be used to insert comments in different steps of the
6306 script.
6307
6308 For each tcp-check rule you create, you can add a "comment" directive,
6309 followed by a string. This string will be reported in the log and stderr
6310 in debug mode. It is useful to make user-friendly error reporting.
6311 The "comment" is of course optional.
6312
6313
Willy Tarreaued179852013-12-16 01:07:00 +01006314 Examples :
6315 # perform a POP check (analyse only server's banner)
6316 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006317 tcp-check expect string +OK\ POP3\ ready comment POP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01006318
6319 # perform an IMAP check (analyse only server's banner)
6320 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006321 tcp-check expect string *\ OK\ IMAP4\ ready comment IMAP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01006322
6323 # look for the redis master server after ensuring it speaks well
6324 # redis protocol, then it exits properly.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006325 # (send a command then analyse the response 3 times)
Willy Tarreaued179852013-12-16 01:07:00 +01006326 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006327 tcp-check comment PING\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01006328 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02006329 tcp-check expect string +PONG
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006330 tcp-check comment role\ check
Willy Tarreaued179852013-12-16 01:07:00 +01006331 tcp-check send info\ replication\r\n
6332 tcp-check expect string role:master
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006333 tcp-check comment QUIT\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01006334 tcp-check send QUIT\r\n
6335 tcp-check expect string +OK
6336
6337 forge a HTTP request, then analyse the response
6338 (send many headers before analyzing)
6339 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006340 tcp-check comment forge\ and\ send\ HTTP\ request
Willy Tarreaued179852013-12-16 01:07:00 +01006341 tcp-check send HEAD\ /\ HTTP/1.1\r\n
6342 tcp-check send Host:\ www.mydomain.com\r\n
6343 tcp-check send User-Agent:\ HAProxy\ tcpcheck\r\n
6344 tcp-check send \r\n
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006345 tcp-check expect rstring HTTP/1\..\ (2..|3..) comment check\ HTTP\ response
Willy Tarreaued179852013-12-16 01:07:00 +01006346
6347
6348 See also : "tcp-check expect", "tcp-check send"
6349
6350
Willy Tarreau9ea05a72009-06-14 12:07:01 +02006351option tcp-smart-accept
6352no option tcp-smart-accept
6353 Enable or disable the saving of one ACK packet during the accept sequence
6354 May be used in sections : defaults | frontend | listen | backend
6355 yes | yes | yes | no
6356 Arguments : none
6357
6358 When an HTTP connection request comes in, the system acknowledges it on
6359 behalf of HAProxy, then the client immediately sends its request, and the
6360 system acknowledges it too while it is notifying HAProxy about the new
6361 connection. HAProxy then reads the request and responds. This means that we
6362 have one TCP ACK sent by the system for nothing, because the request could
6363 very well be acknowledged by HAProxy when it sends its response.
6364
6365 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
6366 sending this useless ACK on platforms which support it (currently at least
6367 Linux). It must not cause any problem, because the system will send it anyway
6368 after 40 ms if the response takes more time than expected to come.
6369
6370 During complex network debugging sessions, it may be desirable to disable
6371 this optimization because delayed ACKs can make troubleshooting more complex
6372 when trying to identify where packets are delayed. It is then possible to
6373 fall back to normal behaviour by specifying "no option tcp-smart-accept".
6374
6375 It is also possible to force it for non-HTTP proxies by simply specifying
6376 "option tcp-smart-accept". For instance, it can make sense with some services
6377 such as SMTP where the server speaks first.
6378
6379 It is recommended to avoid forcing this option in a defaults section. In case
6380 of doubt, consider setting it back to automatic values by prepending the
6381 "default" keyword before it, or disabling it using the "no" keyword.
6382
Willy Tarreaud88edf22009-06-14 15:48:17 +02006383 See also : "option tcp-smart-connect"
6384
6385
6386option tcp-smart-connect
6387no option tcp-smart-connect
6388 Enable or disable the saving of one ACK packet during the connect sequence
6389 May be used in sections : defaults | frontend | listen | backend
6390 yes | no | yes | yes
6391 Arguments : none
6392
6393 On certain systems (at least Linux), HAProxy can ask the kernel not to
6394 immediately send an empty ACK upon a connection request, but to directly
6395 send the buffer request instead. This saves one packet on the network and
6396 thus boosts performance. It can also be useful for some servers, because they
6397 immediately get the request along with the incoming connection.
6398
6399 This feature is enabled when "option tcp-smart-connect" is set in a backend.
6400 It is not enabled by default because it makes network troubleshooting more
6401 complex.
6402
6403 It only makes sense to enable it with protocols where the client speaks first
6404 such as HTTP. In other situations, if there is no data to send in place of
6405 the ACK, a normal ACK is sent.
6406
6407 If this option has been enabled in a "defaults" section, it can be disabled
6408 in a specific instance by prepending the "no" keyword before it.
6409
6410 See also : "option tcp-smart-accept"
6411
Willy Tarreau9ea05a72009-06-14 12:07:01 +02006412
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006413option tcpka
6414 Enable or disable the sending of TCP keepalive packets on both sides
6415 May be used in sections : defaults | frontend | listen | backend
6416 yes | yes | yes | yes
6417 Arguments : none
6418
6419 When there is a firewall or any session-aware component between a client and
6420 a server, and when the protocol involves very long sessions with long idle
6421 periods (eg: remote desktops), there is a risk that one of the intermediate
6422 components decides to expire a session which has remained idle for too long.
6423
6424 Enabling socket-level TCP keep-alives makes the system regularly send packets
6425 to the other end of the connection, leaving it active. The delay between
6426 keep-alive probes is controlled by the system only and depends both on the
6427 operating system and its tuning parameters.
6428
6429 It is important to understand that keep-alive packets are neither emitted nor
6430 received at the application level. It is only the network stacks which sees
6431 them. For this reason, even if one side of the proxy already uses keep-alives
6432 to maintain its connection alive, those keep-alive packets will not be
6433 forwarded to the other side of the proxy.
6434
6435 Please note that this has nothing to do with HTTP keep-alive.
6436
6437 Using option "tcpka" enables the emission of TCP keep-alive probes on both
6438 the client and server sides of a connection. Note that this is meaningful
6439 only in "defaults" or "listen" sections. If this option is used in a
6440 frontend, only the client side will get keep-alives, and if this option is
6441 used in a backend, only the server side will get keep-alives. For this
6442 reason, it is strongly recommended to explicitly use "option clitcpka" and
6443 "option srvtcpka" when the configuration is split between frontends and
6444 backends.
6445
6446 See also : "option clitcpka", "option srvtcpka"
6447
Willy Tarreau844e3c52008-01-11 16:28:18 +01006448
6449option tcplog
6450 Enable advanced logging of TCP connections with session state and timers
6451 May be used in sections : defaults | frontend | listen | backend
6452 yes | yes | yes | yes
6453 Arguments : none
6454
6455 By default, the log output format is very poor, as it only contains the
6456 source and destination addresses, and the instance name. By specifying
6457 "option tcplog", each log line turns into a much richer format including, but
6458 not limited to, the connection timers, the session status, the connections
6459 numbers, the frontend, backend and server name, and of course the source
6460 address and ports. This option is useful for pure TCP proxies in order to
6461 find which of the client or server disconnects or times out. For normal HTTP
6462 proxies, it's better to use "option httplog" which is even more complete.
6463
6464 This option may be set either in the frontend or the backend.
6465
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006466 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01006467
6468
Willy Tarreau844e3c52008-01-11 16:28:18 +01006469option transparent
6470no option transparent
6471 Enable client-side transparent proxying
6472 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01006473 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01006474 Arguments : none
6475
6476 This option was introduced in order to provide layer 7 persistence to layer 3
6477 load balancers. The idea is to use the OS's ability to redirect an incoming
6478 connection for a remote address to a local process (here HAProxy), and let
6479 this process know what address was initially requested. When this option is
6480 used, sessions without cookies will be forwarded to the original destination
6481 IP address of the incoming request (which should match that of another
6482 equipment), while requests with cookies will still be forwarded to the
6483 appropriate server.
6484
6485 Note that contrary to a common belief, this option does NOT make HAProxy
6486 present the client's IP to the server when establishing the connection.
6487
Willy Tarreaua1146052011-03-01 09:51:54 +01006488 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006489 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01006490
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006491
Simon Horman98637e52014-06-20 12:30:16 +09006492external-check command <command>
6493 Executable to run when performing an external-check
6494 May be used in sections : defaults | frontend | listen | backend
6495 yes | no | yes | yes
6496
6497 Arguments :
6498 <command> is the external command to run
6499
Simon Horman98637e52014-06-20 12:30:16 +09006500 The arguments passed to the to the command are:
6501
Cyril Bonté777be862014-12-02 21:21:35 +01006502 <proxy_address> <proxy_port> <server_address> <server_port>
Simon Horman98637e52014-06-20 12:30:16 +09006503
Cyril Bonté777be862014-12-02 21:21:35 +01006504 The <proxy_address> and <proxy_port> are derived from the first listener
6505 that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
6506 listener the proxy_address will be the path of the socket and the
6507 <proxy_port> will be the string "NOT_USED". In a backend section, it's not
6508 possible to determine a listener, and both <proxy_address> and <proxy_port>
6509 will have the string value "NOT_USED".
Simon Horman98637e52014-06-20 12:30:16 +09006510
Cyril Bonté72cda2a2014-12-27 22:28:39 +01006511 Some values are also provided through environment variables.
6512
6513 Environment variables :
6514 HAPROXY_PROXY_ADDR The first bind address if available (or empty if not
6515 applicable, for example in a "backend" section).
6516
6517 HAPROXY_PROXY_ID The backend id.
6518
6519 HAPROXY_PROXY_NAME The backend name.
6520
6521 HAPROXY_PROXY_PORT The first bind port if available (or empty if not
6522 applicable, for example in a "backend" section or
6523 for a UNIX socket).
6524
6525 HAPROXY_SERVER_ADDR The server address.
6526
6527 HAPROXY_SERVER_CURCONN The current number of connections on the server.
6528
6529 HAPROXY_SERVER_ID The server id.
6530
6531 HAPROXY_SERVER_MAXCONN The server max connections.
6532
6533 HAPROXY_SERVER_NAME The server name.
6534
6535 HAPROXY_SERVER_PORT The server port if available (or empty for a UNIX
6536 socket).
6537
6538 PATH The PATH environment variable used when executing
6539 the command may be set using "external-check path".
6540
Simon Horman98637e52014-06-20 12:30:16 +09006541 If the command executed and exits with a zero status then the check is
6542 considered to have passed, otherwise the check is considered to have
6543 failed.
6544
6545 Example :
6546 external-check command /bin/true
6547
6548 See also : "external-check", "option external-check", "external-check path"
6549
6550
6551external-check path <path>
6552 The value of the PATH environment variable used when running an external-check
6553 May be used in sections : defaults | frontend | listen | backend
6554 yes | no | yes | yes
6555
6556 Arguments :
6557 <path> is the path used when executing external command to run
6558
6559 The default path is "".
6560
6561 Example :
6562 external-check path "/usr/bin:/bin"
6563
6564 See also : "external-check", "option external-check",
6565 "external-check command"
6566
6567
Emeric Brun647caf12009-06-30 17:57:00 +02006568persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02006569persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02006570 Enable RDP cookie-based persistence
6571 May be used in sections : defaults | frontend | listen | backend
6572 yes | no | yes | yes
6573 Arguments :
6574 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02006575 default cookie name "msts" will be used. There currently is no
6576 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02006577
6578 This statement enables persistence based on an RDP cookie. The RDP cookie
6579 contains all information required to find the server in the list of known
6580 servers. So when this option is set in the backend, the request is analysed
6581 and if an RDP cookie is found, it is decoded. If it matches a known server
6582 which is still UP (or if "option persist" is set), then the connection is
6583 forwarded to this server.
6584
6585 Note that this only makes sense in a TCP backend, but for this to work, the
6586 frontend must have waited long enough to ensure that an RDP cookie is present
6587 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006588 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02006589 a single "listen" section.
6590
Willy Tarreau61e28f22010-05-16 22:31:05 +02006591 Also, it is important to understand that the terminal server will emit this
6592 RDP cookie only if it is configured for "token redirection mode", which means
6593 that the "IP address redirection" option is disabled.
6594
Emeric Brun647caf12009-06-30 17:57:00 +02006595 Example :
6596 listen tse-farm
6597 bind :3389
6598 # wait up to 5s for an RDP cookie in the request
6599 tcp-request inspect-delay 5s
6600 tcp-request content accept if RDP_COOKIE
6601 # apply RDP cookie persistence
6602 persist rdp-cookie
6603 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02006604 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02006605 balance rdp-cookie
6606 server srv1 1.1.1.1:3389
6607 server srv2 1.1.1.2:3389
6608
Simon Hormanab814e02011-06-24 14:50:20 +09006609 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
6610 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02006611
6612
Willy Tarreau3a7d2072009-03-05 23:48:25 +01006613rate-limit sessions <rate>
6614 Set a limit on the number of new sessions accepted per second on a frontend
6615 May be used in sections : defaults | frontend | listen | backend
6616 yes | yes | yes | no
6617 Arguments :
6618 <rate> The <rate> parameter is an integer designating the maximum number
6619 of new sessions per second to accept on the frontend.
6620
6621 When the frontend reaches the specified number of new sessions per second, it
6622 stops accepting new connections until the rate drops below the limit again.
6623 During this time, the pending sessions will be kept in the socket's backlog
6624 (in system buffers) and haproxy will not even be aware that sessions are
6625 pending. When applying very low limit on a highly loaded service, it may make
6626 sense to increase the socket's backlog using the "backlog" keyword.
6627
6628 This feature is particularly efficient at blocking connection-based attacks
6629 or service abuse on fragile servers. Since the session rate is measured every
6630 millisecond, it is extremely accurate. Also, the limit applies immediately,
6631 no delay is needed at all to detect the threshold.
6632
6633 Example : limit the connection rate on SMTP to 10 per second max
6634 listen smtp
6635 mode tcp
6636 bind :25
6637 rate-limit sessions 10
Panagiotis Panagiotopoulos7282d8e2016-02-11 16:37:15 +02006638 server smtp1 127.0.0.1:1025
Willy Tarreau3a7d2072009-03-05 23:48:25 +01006639
Willy Tarreaua17c2d92011-07-25 08:16:20 +02006640 Note : when the maximum rate is reached, the frontend's status is not changed
6641 but its sockets appear as "WAITING" in the statistics if the
6642 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +01006643
6644 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
6645
6646
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006647redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
6648redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
6649redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006650 Return an HTTP redirection if/unless a condition is matched
6651 May be used in sections : defaults | frontend | listen | backend
6652 no | yes | yes | yes
6653
6654 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +01006655 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006656
Willy Tarreau0140f252008-11-19 21:07:09 +01006657 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006658 <loc> With "redirect location", the exact value in <loc> is placed into
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006659 the HTTP "Location" header. When used in an "http-request" rule,
6660 <loc> value follows the log-format rules and can include some
6661 dynamic values (see Custom Log Format in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006662
6663 <pfx> With "redirect prefix", the "Location" header is built from the
6664 concatenation of <pfx> and the complete URI path, including the
6665 query string, unless the "drop-query" option is specified (see
6666 below). As a special case, if <pfx> equals exactly "/", then
6667 nothing is inserted before the original URI. It allows one to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006668 redirect to the same URL (for instance, to insert a cookie). When
6669 used in an "http-request" rule, <pfx> value follows the log-format
6670 rules and can include some dynamic values (see Custom Log Format
6671 in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006672
6673 <sch> With "redirect scheme", then the "Location" header is built by
6674 concatenating <sch> with "://" then the first occurrence of the
6675 "Host" header, and then the URI path, including the query string
6676 unless the "drop-query" option is specified (see below). If no
6677 path is found or if the path is "*", then "/" is used instead. If
6678 no "Host" header is found, then an empty host component will be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006679 returned, which most recent browsers interpret as redirecting to
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006680 the same host. This directive is mostly used to redirect HTTP to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006681 HTTPS. When used in an "http-request" rule, <sch> value follows
6682 the log-format rules and can include some dynamic values (see
6683 Custom Log Format in section 8.2.4).
Willy Tarreau0140f252008-11-19 21:07:09 +01006684
6685 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +01006686 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
6687 with 302 used by default if no code is specified. 301 means
6688 "Moved permanently", and a browser may cache the Location. 302
Baptiste Assmannea849c02015-08-03 11:42:50 +02006689 means "Moved temporarily" and means that the browser should not
Willy Tarreaub67fdc42013-03-29 19:28:11 +01006690 cache the redirection. 303 is equivalent to 302 except that the
6691 browser will fetch the location with a GET method. 307 is just
6692 like 302 but makes it clear that the same method must be reused.
6693 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +01006694
6695 <option> There are several options which can be specified to adjust the
6696 expected behaviour of a redirection :
6697
6698 - "drop-query"
6699 When this keyword is used in a prefix-based redirection, then the
6700 location will be set without any possible query-string, which is useful
6701 for directing users to a non-secure page for instance. It has no effect
6702 with a location-type redirect.
6703
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01006704 - "append-slash"
6705 This keyword may be used in conjunction with "drop-query" to redirect
6706 users who use a URL not ending with a '/' to the same one with the '/'.
6707 It can be useful to ensure that search engines will only see one URL.
6708 For this, a return code 301 is preferred.
6709
Willy Tarreau0140f252008-11-19 21:07:09 +01006710 - "set-cookie NAME[=value]"
6711 A "Set-Cookie" header will be added with NAME (and optionally "=value")
6712 to the response. This is sometimes used to indicate that a user has
6713 been seen, for instance to protect against some types of DoS. No other
6714 cookie option is added, so the cookie will be a session cookie. Note
6715 that for a browser, a sole cookie name without an equal sign is
6716 different from a cookie with an equal sign.
6717
6718 - "clear-cookie NAME[=]"
6719 A "Set-Cookie" header will be added with NAME (and optionally "="), but
6720 with the "Max-Age" attribute set to zero. This will tell the browser to
6721 delete this cookie. It is useful for instance on logout pages. It is
6722 important to note that clearing the cookie "NAME" will not remove a
6723 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
6724 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006725
6726 Example: move the login URL only to HTTPS.
6727 acl clear dst_port 80
6728 acl secure dst_port 8080
6729 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +01006730 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +01006731 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +01006732 acl cookie_set hdr_sub(cookie) SEEN=1
6733
6734 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +01006735 redirect prefix https://mysite.com if login_page !secure
6736 redirect prefix http://mysite.com drop-query if login_page !uid_given
6737 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +01006738 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006739
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01006740 Example: send redirects for request for articles without a '/'.
6741 acl missing_slash path_reg ^/article/[^/]*$
6742 redirect code 301 prefix / drop-query append-slash if missing_slash
6743
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006744 Example: redirect all HTTP traffic to HTTPS when SSL is handled by haproxy.
David BERARDe7153042012-11-03 00:11:31 +01006745 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006746
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006747 Example: append 'www.' prefix in front of all hosts not having it
Coen Rosdorff596659b2016-04-11 11:33:49 +02006748 http-request redirect code 301 location \
6749 http://www.%[hdr(host)]%[capture.req.uri] \
6750 unless { hdr_beg(host) -i www }
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006751
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006752 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006753
6754
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006755redisp (deprecated)
6756redispatch (deprecated)
6757 Enable or disable session redistribution in case of connection failure
6758 May be used in sections: defaults | frontend | listen | backend
6759 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006760 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006761
6762 In HTTP mode, if a server designated by a cookie is down, clients may
6763 definitely stick to it because they cannot flush the cookie, so they will not
6764 be able to access the service anymore.
6765
6766 Specifying "redispatch" will allow the proxy to break their persistence and
6767 redistribute them to a working server.
6768
6769 It also allows to retry last connection to another server in case of multiple
6770 connection failures. Of course, it requires having "retries" set to a nonzero
6771 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01006772
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006773 This form is deprecated, do not use it in any new configuration, use the new
6774 "option redispatch" instead.
6775
6776 See also : "option redispatch"
6777
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006778
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006779reqadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01006780 Add a header at the end of the HTTP request
6781 May be used in sections : defaults | frontend | listen | backend
6782 no | yes | yes | yes
6783 Arguments :
6784 <string> is the complete line to be added. Any space or known delimiter
6785 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006786 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006787
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006788 <cond> is an optional matching condition built from ACLs. It makes it
6789 possible to ignore this rule when other conditions are not met.
6790
Willy Tarreau303c0352008-01-17 19:01:39 +01006791 A new line consisting in <string> followed by a line feed will be added after
6792 the last header of an HTTP request.
6793
6794 Header transformations only apply to traffic which passes through HAProxy,
6795 and not to traffic generated by HAProxy, such as health-checks or error
6796 responses.
6797
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006798 Example : add "X-Proto: SSL" to requests coming via port 81
6799 acl is-ssl dst_port 81
6800 reqadd X-Proto:\ SSL if is-ssl
6801
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006802 See also: "rspadd", "http-request", section 6 about HTTP header manipulation,
6803 and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006804
6805
Willy Tarreau5321c422010-01-28 20:35:13 +01006806reqallow <search> [{if | unless} <cond>]
6807reqiallow <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006808 Definitely allow an HTTP request if a line matches a regular expression
6809 May be used in sections : defaults | frontend | listen | backend
6810 no | yes | yes | yes
6811 Arguments :
6812 <search> is the regular expression applied to HTTP headers and to the
6813 request line. This is an extended regular expression. Parenthesis
6814 grouping is supported and no preliminary backslash is required.
6815 Any space or known delimiter must be escaped using a backslash
6816 ('\'). The pattern applies to a full line at a time. The
6817 "reqallow" keyword strictly matches case while "reqiallow"
6818 ignores case.
6819
Willy Tarreau5321c422010-01-28 20:35:13 +01006820 <cond> is an optional matching condition built from ACLs. It makes it
6821 possible to ignore this rule when other conditions are not met.
6822
Willy Tarreau303c0352008-01-17 19:01:39 +01006823 A request containing any line which matches extended regular expression
6824 <search> will mark the request as allowed, even if any later test would
6825 result in a deny. The test applies both to the request line and to request
6826 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01006827 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01006828
6829 It is easier, faster and more powerful to use ACLs to write access policies.
6830 Reqdeny, reqallow and reqpass should be avoided in new designs.
6831
6832 Example :
6833 # allow www.* but refuse *.local
6834 reqiallow ^Host:\ www\.
6835 reqideny ^Host:\ .*\.local
6836
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006837 See also: "reqdeny", "block", "http-request", section 6 about HTTP header
6838 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006839
6840
Willy Tarreau5321c422010-01-28 20:35:13 +01006841reqdel <search> [{if | unless} <cond>]
6842reqidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006843 Delete all headers matching a regular expression in an HTTP request
6844 May be used in sections : defaults | frontend | listen | backend
6845 no | yes | yes | yes
6846 Arguments :
6847 <search> is the regular expression applied to HTTP headers and to the
6848 request line. This is an extended regular expression. Parenthesis
6849 grouping is supported and no preliminary backslash is required.
6850 Any space or known delimiter must be escaped using a backslash
6851 ('\'). The pattern applies to a full line at a time. The "reqdel"
6852 keyword strictly matches case while "reqidel" ignores case.
6853
Willy Tarreau5321c422010-01-28 20:35:13 +01006854 <cond> is an optional matching condition built from ACLs. It makes it
6855 possible to ignore this rule when other conditions are not met.
6856
Willy Tarreau303c0352008-01-17 19:01:39 +01006857 Any header line matching extended regular expression <search> in the request
6858 will be completely deleted. Most common use of this is to remove unwanted
6859 and/or dangerous headers or cookies from a request before passing it to the
6860 next servers.
6861
6862 Header transformations only apply to traffic which passes through HAProxy,
6863 and not to traffic generated by HAProxy, such as health-checks or error
6864 responses. Keep in mind that header names are not case-sensitive.
6865
6866 Example :
6867 # remove X-Forwarded-For header and SERVER cookie
6868 reqidel ^X-Forwarded-For:.*
6869 reqidel ^Cookie:.*SERVER=
6870
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006871 See also: "reqadd", "reqrep", "rspdel", "http-request", section 6 about
6872 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006873
6874
Willy Tarreau5321c422010-01-28 20:35:13 +01006875reqdeny <search> [{if | unless} <cond>]
6876reqideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006877 Deny an HTTP request if a line matches a regular expression
6878 May be used in sections : defaults | frontend | listen | backend
6879 no | yes | yes | yes
6880 Arguments :
6881 <search> is the regular expression applied to HTTP headers and to the
6882 request line. This is an extended regular expression. Parenthesis
6883 grouping is supported and no preliminary backslash is required.
6884 Any space or known delimiter must be escaped using a backslash
6885 ('\'). The pattern applies to a full line at a time. The
6886 "reqdeny" keyword strictly matches case while "reqideny" ignores
6887 case.
6888
Willy Tarreau5321c422010-01-28 20:35:13 +01006889 <cond> is an optional matching condition built from ACLs. It makes it
6890 possible to ignore this rule when other conditions are not met.
6891
Willy Tarreau303c0352008-01-17 19:01:39 +01006892 A request containing any line which matches extended regular expression
6893 <search> will mark the request as denied, even if any later test would
6894 result in an allow. The test applies both to the request line and to request
6895 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01006896 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01006897
Willy Tarreauced27012008-01-17 20:35:34 +01006898 A denied request will generate an "HTTP 403 forbidden" response once the
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006899 complete request has been parsed. This is consistent with what is practiced
Willy Tarreaud72758d2010-01-12 10:42:19 +01006900 using ACLs.
Willy Tarreauced27012008-01-17 20:35:34 +01006901
Willy Tarreau303c0352008-01-17 19:01:39 +01006902 It is easier, faster and more powerful to use ACLs to write access policies.
6903 Reqdeny, reqallow and reqpass should be avoided in new designs.
6904
6905 Example :
6906 # refuse *.local, then allow www.*
6907 reqideny ^Host:\ .*\.local
6908 reqiallow ^Host:\ www\.
6909
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006910 See also: "reqallow", "rspdeny", "block", "http-request", section 6 about
6911 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006912
6913
Willy Tarreau5321c422010-01-28 20:35:13 +01006914reqpass <search> [{if | unless} <cond>]
6915reqipass <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006916 Ignore any HTTP request line matching a regular expression in next rules
6917 May be used in sections : defaults | frontend | listen | backend
6918 no | yes | yes | yes
6919 Arguments :
6920 <search> is the regular expression applied to HTTP headers and to the
6921 request line. This is an extended regular expression. Parenthesis
6922 grouping is supported and no preliminary backslash is required.
6923 Any space or known delimiter must be escaped using a backslash
6924 ('\'). The pattern applies to a full line at a time. The
6925 "reqpass" keyword strictly matches case while "reqipass" ignores
6926 case.
6927
Willy Tarreau5321c422010-01-28 20:35:13 +01006928 <cond> is an optional matching condition built from ACLs. It makes it
6929 possible to ignore this rule when other conditions are not met.
6930
Willy Tarreau303c0352008-01-17 19:01:39 +01006931 A request containing any line which matches extended regular expression
6932 <search> will skip next rules, without assigning any deny or allow verdict.
6933 The test applies both to the request line and to request headers. Keep in
6934 mind that URLs in request line are case-sensitive while header names are not.
6935
6936 It is easier, faster and more powerful to use ACLs to write access policies.
6937 Reqdeny, reqallow and reqpass should be avoided in new designs.
6938
6939 Example :
6940 # refuse *.local, then allow www.*, but ignore "www.private.local"
6941 reqipass ^Host:\ www.private\.local
6942 reqideny ^Host:\ .*\.local
6943 reqiallow ^Host:\ www\.
6944
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006945 See also: "reqallow", "reqdeny", "block", "http-request", section 6 about
6946 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006947
6948
Willy Tarreau5321c422010-01-28 20:35:13 +01006949reqrep <search> <string> [{if | unless} <cond>]
6950reqirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006951 Replace a regular expression with a string in an HTTP request line
6952 May be used in sections : defaults | frontend | listen | backend
6953 no | yes | yes | yes
6954 Arguments :
6955 <search> is the regular expression applied to HTTP headers and to the
6956 request line. This is an extended regular expression. Parenthesis
6957 grouping is supported and no preliminary backslash is required.
6958 Any space or known delimiter must be escaped using a backslash
6959 ('\'). The pattern applies to a full line at a time. The "reqrep"
6960 keyword strictly matches case while "reqirep" ignores case.
6961
6962 <string> is the complete line to be added. Any space or known delimiter
6963 must be escaped using a backslash ('\'). References to matched
6964 pattern groups are possible using the common \N form, with N
6965 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006966 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006967
Willy Tarreau5321c422010-01-28 20:35:13 +01006968 <cond> is an optional matching condition built from ACLs. It makes it
6969 possible to ignore this rule when other conditions are not met.
6970
Willy Tarreau303c0352008-01-17 19:01:39 +01006971 Any line matching extended regular expression <search> in the request (both
6972 the request line and header lines) will be completely replaced with <string>.
6973 Most common use of this is to rewrite URLs or domain names in "Host" headers.
6974
6975 Header transformations only apply to traffic which passes through HAProxy,
6976 and not to traffic generated by HAProxy, such as health-checks or error
6977 responses. Note that for increased readability, it is suggested to add enough
6978 spaces between the request and the response. Keep in mind that URLs in
6979 request line are case-sensitive while header names are not.
6980
6981 Example :
6982 # replace "/static/" with "/" at the beginning of any request path.
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04006983 reqrep ^([^\ :]*)\ /static/(.*) \1\ /\2
Willy Tarreau303c0352008-01-17 19:01:39 +01006984 # replace "www.mydomain.com" with "www" in the host name.
6985 reqirep ^Host:\ www.mydomain.com Host:\ www
6986
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006987 See also: "reqadd", "reqdel", "rsprep", "tune.bufsize", "http-request",
6988 section 6 about HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006989
6990
Willy Tarreau5321c422010-01-28 20:35:13 +01006991reqtarpit <search> [{if | unless} <cond>]
6992reqitarpit <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006993 Tarpit an HTTP request containing a line matching a regular expression
6994 May be used in sections : defaults | frontend | listen | backend
6995 no | yes | yes | yes
6996 Arguments :
6997 <search> is the regular expression applied to HTTP headers and to the
6998 request line. This is an extended regular expression. Parenthesis
6999 grouping is supported and no preliminary backslash is required.
7000 Any space or known delimiter must be escaped using a backslash
7001 ('\'). The pattern applies to a full line at a time. The
7002 "reqtarpit" keyword strictly matches case while "reqitarpit"
7003 ignores case.
7004
Willy Tarreau5321c422010-01-28 20:35:13 +01007005 <cond> is an optional matching condition built from ACLs. It makes it
7006 possible to ignore this rule when other conditions are not met.
7007
Willy Tarreau303c0352008-01-17 19:01:39 +01007008 A request containing any line which matches extended regular expression
7009 <search> will be tarpitted, which means that it will connect to nowhere, will
Willy Tarreauced27012008-01-17 20:35:34 +01007010 be kept open for a pre-defined time, then will return an HTTP error 500 so
7011 that the attacker does not suspect it has been tarpitted. The status 500 will
7012 be reported in the logs, but the completion flags will indicate "PT". The
Willy Tarreau303c0352008-01-17 19:01:39 +01007013 delay is defined by "timeout tarpit", or "timeout connect" if the former is
7014 not set.
7015
7016 The goal of the tarpit is to slow down robots attacking servers with
7017 identifiable requests. Many robots limit their outgoing number of connections
7018 and stay connected waiting for a reply which can take several minutes to
7019 come. Depending on the environment and attack, it may be particularly
7020 efficient at reducing the load on the network and firewalls.
7021
Willy Tarreau5321c422010-01-28 20:35:13 +01007022 Examples :
Willy Tarreau303c0352008-01-17 19:01:39 +01007023 # ignore user-agents reporting any flavour of "Mozilla" or "MSIE", but
7024 # block all others.
7025 reqipass ^User-Agent:\.*(Mozilla|MSIE)
7026 reqitarpit ^User-Agent:
7027
Willy Tarreau5321c422010-01-28 20:35:13 +01007028 # block bad guys
7029 acl badguys src 10.1.0.3 172.16.13.20/28
7030 reqitarpit . if badguys
7031
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007032 See also: "reqallow", "reqdeny", "reqpass", "http-request", section 6
7033 about HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007034
7035
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02007036retries <value>
7037 Set the number of retries to perform on a server after a connection failure
7038 May be used in sections: defaults | frontend | listen | backend
7039 yes | no | yes | yes
7040 Arguments :
7041 <value> is the number of times a connection attempt should be retried on
7042 a server when a connection either is refused or times out. The
7043 default value is 3.
7044
7045 It is important to understand that this value applies to the number of
7046 connection attempts, not full requests. When a connection has effectively
7047 been established to a server, there will be no more retry.
7048
7049 In order to avoid immediate reconnections to a server which is restarting,
Joseph Lynch726ab712015-05-11 23:25:34 -07007050 a turn-around timer of min("timeout connect", one second) is applied before
7051 a retry occurs.
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02007052
7053 When "option redispatch" is set, the last retry may be performed on another
7054 server even if a cookie references a different server.
7055
7056 See also : "option redispatch"
7057
7058
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007059rspadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01007060 Add a header at the end of the HTTP response
7061 May be used in sections : defaults | frontend | listen | backend
7062 no | yes | yes | yes
7063 Arguments :
7064 <string> is the complete line to be added. Any space or known delimiter
7065 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007066 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01007067
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007068 <cond> is an optional matching condition built from ACLs. It makes it
7069 possible to ignore this rule when other conditions are not met.
7070
Willy Tarreau303c0352008-01-17 19:01:39 +01007071 A new line consisting in <string> followed by a line feed will be added after
7072 the last header of an HTTP response.
7073
7074 Header transformations only apply to traffic which passes through HAProxy,
7075 and not to traffic generated by HAProxy, such as health-checks or error
7076 responses.
7077
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007078 See also: "rspdel" "reqadd", "http-response", section 6 about HTTP header
7079 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007080
7081
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007082rspdel <search> [{if | unless} <cond>]
7083rspidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007084 Delete all headers matching a regular expression in an HTTP response
7085 May be used in sections : defaults | frontend | listen | backend
7086 no | yes | yes | yes
7087 Arguments :
7088 <search> is the regular expression applied to HTTP headers and to the
7089 response line. This is an extended regular expression, so
7090 parenthesis grouping is supported and no preliminary backslash
7091 is required. Any space or known delimiter must be escaped using
7092 a backslash ('\'). The pattern applies to a full line at a time.
7093 The "rspdel" keyword strictly matches case while "rspidel"
7094 ignores case.
7095
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007096 <cond> is an optional matching condition built from ACLs. It makes it
7097 possible to ignore this rule when other conditions are not met.
7098
Willy Tarreau303c0352008-01-17 19:01:39 +01007099 Any header line matching extended regular expression <search> in the response
7100 will be completely deleted. Most common use of this is to remove unwanted
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02007101 and/or sensitive headers or cookies from a response before passing it to the
Willy Tarreau303c0352008-01-17 19:01:39 +01007102 client.
7103
7104 Header transformations only apply to traffic which passes through HAProxy,
7105 and not to traffic generated by HAProxy, such as health-checks or error
7106 responses. Keep in mind that header names are not case-sensitive.
7107
7108 Example :
7109 # remove the Server header from responses
Willy Tarreau5e80e022013-05-25 08:31:25 +02007110 rspidel ^Server:.*
Willy Tarreau303c0352008-01-17 19:01:39 +01007111
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007112 See also: "rspadd", "rsprep", "reqdel", "http-response", section 6 about
7113 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007114
7115
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007116rspdeny <search> [{if | unless} <cond>]
7117rspideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007118 Block an HTTP response if a line matches a regular expression
7119 May be used in sections : defaults | frontend | listen | backend
7120 no | yes | yes | yes
7121 Arguments :
7122 <search> is the regular expression applied to HTTP headers and to the
7123 response line. This is an extended regular expression, so
7124 parenthesis grouping is supported and no preliminary backslash
7125 is required. Any space or known delimiter must be escaped using
7126 a backslash ('\'). The pattern applies to a full line at a time.
7127 The "rspdeny" keyword strictly matches case while "rspideny"
7128 ignores case.
7129
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007130 <cond> is an optional matching condition built from ACLs. It makes it
7131 possible to ignore this rule when other conditions are not met.
7132
Willy Tarreau303c0352008-01-17 19:01:39 +01007133 A response containing any line which matches extended regular expression
7134 <search> will mark the request as denied. The test applies both to the
7135 response line and to response headers. Keep in mind that header names are not
7136 case-sensitive.
7137
7138 Main use of this keyword is to prevent sensitive information leak and to
Willy Tarreauced27012008-01-17 20:35:34 +01007139 block the response before it reaches the client. If a response is denied, it
7140 will be replaced with an HTTP 502 error so that the client never retrieves
7141 any sensitive data.
Willy Tarreau303c0352008-01-17 19:01:39 +01007142
7143 It is easier, faster and more powerful to use ACLs to write access policies.
7144 Rspdeny should be avoided in new designs.
7145
7146 Example :
7147 # Ensure that no content type matching ms-word will leak
7148 rspideny ^Content-type:\.*/ms-word
7149
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007150 See also: "reqdeny", "acl", "block", "http-response", section 6 about
7151 HTTP header manipulation and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007152
7153
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007154rsprep <search> <string> [{if | unless} <cond>]
7155rspirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007156 Replace a regular expression with a string in an HTTP response line
7157 May be used in sections : defaults | frontend | listen | backend
7158 no | yes | yes | yes
7159 Arguments :
7160 <search> is the regular expression applied to HTTP headers and to the
7161 response line. This is an extended regular expression, so
7162 parenthesis grouping is supported and no preliminary backslash
7163 is required. Any space or known delimiter must be escaped using
7164 a backslash ('\'). The pattern applies to a full line at a time.
7165 The "rsprep" keyword strictly matches case while "rspirep"
7166 ignores case.
7167
7168 <string> is the complete line to be added. Any space or known delimiter
7169 must be escaped using a backslash ('\'). References to matched
7170 pattern groups are possible using the common \N form, with N
7171 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007172 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01007173
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007174 <cond> is an optional matching condition built from ACLs. It makes it
7175 possible to ignore this rule when other conditions are not met.
7176
Willy Tarreau303c0352008-01-17 19:01:39 +01007177 Any line matching extended regular expression <search> in the response (both
7178 the response line and header lines) will be completely replaced with
7179 <string>. Most common use of this is to rewrite Location headers.
7180
7181 Header transformations only apply to traffic which passes through HAProxy,
7182 and not to traffic generated by HAProxy, such as health-checks or error
7183 responses. Note that for increased readability, it is suggested to add enough
7184 spaces between the request and the response. Keep in mind that header names
7185 are not case-sensitive.
7186
7187 Example :
7188 # replace "Location: 127.0.0.1:8080" with "Location: www.mydomain.com"
7189 rspirep ^Location:\ 127.0.0.1:8080 Location:\ www.mydomain.com
7190
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007191 See also: "rspadd", "rspdel", "reqrep", "http-response", section 6 about
7192 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007193
7194
David du Colombier486df472011-03-17 10:40:26 +01007195server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007196 Declare a server in a backend
7197 May be used in sections : defaults | frontend | listen | backend
7198 no | no | yes | yes
7199 Arguments :
7200 <name> is the internal name assigned to this server. This name will
Cyril Bonté941a0c62012-10-15 19:44:24 +02007201 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -05007202 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007203
David du Colombier486df472011-03-17 10:40:26 +01007204 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
7205 resolvable hostname is supported, but this name will be resolved
7206 during start-up. Address "0.0.0.0" or "*" has a special meaning.
7207 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +02007208 address as the one from the client connection. This is useful in
7209 transparent proxy architectures where the client's connection is
7210 intercepted and haproxy must forward to the original destination
7211 address. This is more or less what the "transparent" keyword does
7212 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +01007213 to report statistics. Optionally, an address family prefix may be
7214 used before the address to force the family regardless of the
7215 address format, which can be useful to specify a path to a unix
7216 socket with no slash ('/'). Currently supported prefixes are :
7217 - 'ipv4@' -> address is always IPv4
7218 - 'ipv6@' -> address is always IPv6
7219 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02007220 - 'abns@' -> address is in abstract namespace (Linux only)
William Lallemandb2f07452015-05-12 14:27:13 +02007221 You may want to reference some environment variables in the
7222 address parameter, see section 2.3 about environment
7223 variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007224
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007225 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007226 be sent to this port. If unset, the same port the client
7227 connected to will be used. The port may also be prefixed by a "+"
7228 or a "-". In this case, the server's port will be determined by
7229 adding this value to the client's port.
7230
7231 <param*> is a list of parameters for this server. The "server" keywords
7232 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007233 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007234
7235 Examples :
7236 server first 10.1.1.1:1080 cookie first check inter 1000
7237 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +01007238 server transp ipv4@
William Lallemandb2f07452015-05-12 14:27:13 +02007239 server backup "${SRV_BACKUP}:1080" backup
7240 server www1_dc1 "${LAN_DC1}.101:80"
7241 server www1_dc2 "${LAN_DC2}.101:80"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007242
Willy Tarreau55dcaf62015-09-27 15:03:15 +02007243 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
7244 sun_path length is used for the address length. Some other programs
7245 such as socat use the string length only by default. Pass the option
7246 ",unix-tightsocklen=0" to any abstract socket definition in socat to
7247 make it compatible with HAProxy's.
7248
Mark Lamourinec2247f02012-01-04 13:02:01 -05007249 See also: "default-server", "http-send-name-header" and section 5 about
7250 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007251
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007252server-state-file-name [<file>]
7253 Set the server state file to read, load and apply to servers available in
7254 this backend. It only applies when the directive "load-server-state-from-file"
7255 is set to "local". When <file> is not provided or if this directive is not
7256 set, then backend name is used. If <file> starts with a slash '/', then it is
7257 considered as an absolute path. Otherwise, <file> is concatenated to the
7258 global directive "server-state-file-base".
7259
7260 Example: the minimal configuration below would make HAProxy look for the
7261 state server file '/etc/haproxy/states/bk':
7262
7263 global
7264 server-state-file-base /etc/haproxy/states
7265
7266 backend bk
7267 load-server-state-from-file
7268
7269 See also: "server-state-file-base", "load-server-state-from-file", and
7270 "show servers state"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007271
7272source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02007273source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +01007274source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007275 Set the source address for outgoing connections
7276 May be used in sections : defaults | frontend | listen | backend
7277 yes | no | yes | yes
7278 Arguments :
7279 <addr> is the IPv4 address HAProxy will bind to before connecting to a
7280 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +01007281
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007282 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +01007283 the most appropriate address to reach its destination. Optionally
7284 an address family prefix may be used before the address to force
7285 the family regardless of the address format, which can be useful
7286 to specify a path to a unix socket with no slash ('/'). Currently
7287 supported prefixes are :
7288 - 'ipv4@' -> address is always IPv4
7289 - 'ipv6@' -> address is always IPv6
7290 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02007291 - 'abns@' -> address is in abstract namespace (Linux only)
Cyril Bonté307ee1e2015-09-28 23:16:06 +02007292 You may want to reference some environment variables in the
7293 address parameter, see section 2.3 about environment variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007294
7295 <port> is an optional port. It is normally not needed but may be useful
7296 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02007297 the system will select a free port. Note that port ranges are not
7298 supported in the backend. If you want to force port ranges, you
7299 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007300
7301 <addr2> is the IP address to present to the server when connections are
7302 forwarded in full transparent proxy mode. This is currently only
7303 supported on some patched Linux kernels. When this address is
7304 specified, clients connecting to the server will be presented
7305 with this address, while health checks will still use the address
7306 <addr>.
7307
7308 <port2> is the optional port to present to the server when connections
7309 are forwarded in full transparent proxy mode (see <addr2> above).
7310 The default value of zero means the system will select a free
7311 port.
7312
Willy Tarreaubce70882009-09-07 11:51:47 +02007313 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
7314 This is the name of a comma-separated header list which can
7315 contain multiple IP addresses. By default, the last occurrence is
7316 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +01007317 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +02007318 by previous proxy, typically Stunnel. In order to use another
7319 occurrence from the last one, please see the <occ> parameter
7320 below. When the header (or occurrence) is not found, no binding
7321 is performed so that the proxy's default IP address is used. Also
7322 keep in mind that the header name is case insensitive, as for any
7323 HTTP header.
7324
7325 <occ> is the occurrence number of a value to be used in a multi-value
7326 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007327 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +02007328 address. Positive values indicate a position from the first
7329 occurrence, 1 being the first one. Negative values indicate
7330 positions relative to the last one, -1 being the last one. This
7331 is helpful for situations where an X-Forwarded-For header is set
7332 at the entry point of an infrastructure and must be used several
7333 proxy layers away. When this value is not specified, -1 is
7334 assumed. Passing a zero here disables the feature.
7335
Willy Tarreaud53f96b2009-02-04 18:46:54 +01007336 <name> is an optional interface name to which to bind to for outgoing
7337 traffic. On systems supporting this features (currently, only
7338 Linux), this allows one to bind all traffic to the server to
7339 this interface even if it is not the one the system would select
7340 based on routing tables. This should be used with extreme care.
7341 Note that using this option requires root privileges.
7342
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007343 The "source" keyword is useful in complex environments where a specific
7344 address only is allowed to connect to the servers. It may be needed when a
7345 private address must be used through a public gateway for instance, and it is
7346 known that the system cannot determine the adequate source address by itself.
7347
7348 An extension which is available on certain patched Linux kernels may be used
7349 through the "usesrc" optional keyword. It makes it possible to connect to the
7350 servers with an IP address which does not belong to the system itself. This
7351 is called "full transparent proxy mode". For this to work, the destination
7352 servers have to route their traffic back to this address through the machine
7353 running HAProxy, and IP forwarding must generally be enabled on this machine.
7354
7355 In this "full transparent proxy" mode, it is possible to force a specific IP
7356 address to be presented to the servers. This is not much used in fact. A more
7357 common use is to tell HAProxy to present the client's IP address. For this,
7358 there are two methods :
7359
7360 - present the client's IP and port addresses. This is the most transparent
7361 mode, but it can cause problems when IP connection tracking is enabled on
7362 the machine, because a same connection may be seen twice with different
7363 states. However, this solution presents the huge advantage of not
7364 limiting the system to the 64k outgoing address+port couples, because all
7365 of the client ranges may be used.
7366
7367 - present only the client's IP address and select a spare port. This
7368 solution is still quite elegant but slightly less transparent (downstream
7369 firewalls logs will not match upstream's). It also presents the downside
7370 of limiting the number of concurrent connections to the usual 64k ports.
7371 However, since the upstream and downstream ports are different, local IP
7372 connection tracking on the machine will not be upset by the reuse of the
7373 same session.
7374
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007375 This option sets the default source for all servers in the backend. It may
7376 also be specified in a "defaults" section. Finer source address specification
7377 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007378 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007379
Baptiste Assmann91bd3372015-07-17 21:59:42 +02007380 In order to work, "usesrc" requires root privileges.
7381
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007382 Examples :
7383 backend private
7384 # Connect to the servers using our 192.168.1.200 source address
7385 source 192.168.1.200
7386
7387 backend transparent_ssl1
7388 # Connect to the SSL farm from the client's source address
7389 source 192.168.1.200 usesrc clientip
7390
7391 backend transparent_ssl2
7392 # Connect to the SSL farm from the client's source address and port
7393 # not recommended if IP conntrack is present on the local machine.
7394 source 192.168.1.200 usesrc client
7395
7396 backend transparent_ssl3
7397 # Connect to the SSL farm from the client's source address. It
7398 # is more conntrack-friendly.
7399 source 192.168.1.200 usesrc clientip
7400
7401 backend transparent_smtp
7402 # Connect to the SMTP farm from the client's source address/port
7403 # with Tproxy version 4.
7404 source 0.0.0.0 usesrc clientip
7405
Willy Tarreaubce70882009-09-07 11:51:47 +02007406 backend transparent_http
7407 # Connect to the servers using the client's IP as seen by previous
7408 # proxy.
7409 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
7410
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007411 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007412 the Linux kernel on www.balabit.com, the "bind" keyword.
7413
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007414
Willy Tarreau844e3c52008-01-11 16:28:18 +01007415srvtimeout <timeout> (deprecated)
7416 Set the maximum inactivity time on the server side.
7417 May be used in sections : defaults | frontend | listen | backend
7418 yes | no | yes | yes
7419 Arguments :
7420 <timeout> is the timeout value specified in milliseconds by default, but
7421 can be in any other unit if the number is suffixed by the unit,
7422 as explained at the top of this document.
7423
7424 The inactivity timeout applies when the server is expected to acknowledge or
7425 send data. In HTTP mode, this timeout is particularly important to consider
7426 during the first phase of the server's response, when it has to send the
7427 headers, as it directly represents the server's processing time for the
7428 request. To find out what value to put there, it's often good to start with
7429 what would be considered as unacceptable response times, then check the logs
7430 to observe the response time distribution, and adjust the value accordingly.
7431
7432 The value is specified in milliseconds by default, but can be in any other
7433 unit if the number is suffixed by the unit, as specified at the top of this
7434 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
7435 recommended that the client timeout remains equal to the server timeout in
7436 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007437 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01007438 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreaud72758d2010-01-12 10:42:19 +01007439 seconds (eg: 4 or 5 seconds minimum).
Willy Tarreau844e3c52008-01-11 16:28:18 +01007440
7441 This parameter is specific to backends, but can be specified once for all in
7442 "defaults" sections. This is in fact one of the easiest solutions not to
7443 forget about it. An unspecified timeout results in an infinite timeout, which
7444 is not recommended. Such a usage is accepted and works but reports a warning
7445 during startup because it may results in accumulation of expired sessions in
7446 the system if the system's timeouts are not configured either.
7447
7448 This parameter is provided for compatibility but is currently deprecated.
7449 Please use "timeout server" instead.
7450
Willy Tarreauce887fd2012-05-12 12:50:00 +02007451 See also : "timeout server", "timeout tunnel", "timeout client" and
7452 "clitimeout".
Willy Tarreau844e3c52008-01-11 16:28:18 +01007453
7454
Cyril Bonté66c327d2010-10-12 00:14:37 +02007455stats admin { if | unless } <cond>
7456 Enable statistics admin level if/unless a condition is matched
7457 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007458 no | yes | yes | yes
Cyril Bonté66c327d2010-10-12 00:14:37 +02007459
7460 This statement enables the statistics admin level if/unless a condition is
7461 matched.
7462
7463 The admin level allows to enable/disable servers from the web interface. By
7464 default, statistics page is read-only for security reasons.
7465
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007466 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7467 unless you know what you do : memory is not shared between the
7468 processes, which can result in random behaviours.
7469
Cyril Bonté23b39d92011-02-10 22:54:44 +01007470 Currently, the POST request is limited to the buffer size minus the reserved
7471 buffer space, which means that if the list of servers is too long, the
7472 request won't be processed. It is recommended to alter few servers at a
7473 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +02007474
7475 Example :
7476 # statistics admin level only for localhost
7477 backend stats_localhost
7478 stats enable
7479 stats admin if LOCALHOST
7480
7481 Example :
7482 # statistics admin level always enabled because of the authentication
7483 backend stats_auth
7484 stats enable
7485 stats auth admin:AdMiN123
7486 stats admin if TRUE
7487
7488 Example :
7489 # statistics admin level depends on the authenticated user
7490 userlist stats-auth
7491 group admin users admin
7492 user admin insecure-password AdMiN123
7493 group readonly users haproxy
7494 user haproxy insecure-password haproxy
7495
7496 backend stats_auth
7497 stats enable
7498 acl AUTH http_auth(stats-auth)
7499 acl AUTH_ADMIN http_auth_group(stats-auth) admin
7500 stats http-request auth unless AUTH
7501 stats admin if AUTH_ADMIN
7502
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007503 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
7504 "bind-process", section 3.4 about userlists and section 7 about
7505 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +02007506
7507
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007508stats auth <user>:<passwd>
7509 Enable statistics with authentication and grant access to an account
7510 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007511 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007512 Arguments :
7513 <user> is a user name to grant access to
7514
7515 <passwd> is the cleartext password associated to this user
7516
7517 This statement enables statistics with default settings, and restricts access
7518 to declared users only. It may be repeated as many times as necessary to
7519 allow as many users as desired. When a user tries to access the statistics
7520 without a valid account, a "401 Forbidden" response will be returned so that
7521 the browser asks the user to provide a valid user and password. The real
7522 which will be returned to the browser is configurable using "stats realm".
7523
7524 Since the authentication method is HTTP Basic Authentication, the passwords
7525 circulate in cleartext on the network. Thus, it was decided that the
7526 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02007527 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007528
7529 It is also possible to reduce the scope of the proxies which appear in the
7530 report using "stats scope".
7531
7532 Though this statement alone is enough to enable statistics reporting, it is
7533 recommended to set all other settings in order to avoid relying on default
7534 unobvious parameters.
7535
7536 Example :
7537 # public access (limited to this backend only)
7538 backend public_www
7539 server srv1 192.168.0.1:80
7540 stats enable
7541 stats hide-version
7542 stats scope .
7543 stats uri /admin?stats
7544 stats realm Haproxy\ Statistics
7545 stats auth admin1:AdMiN123
7546 stats auth admin2:AdMiN321
7547
7548 # internal monitoring access (unlimited)
7549 backend private_monitoring
7550 stats enable
7551 stats uri /admin?stats
7552 stats refresh 5s
7553
7554 See also : "stats enable", "stats realm", "stats scope", "stats uri"
7555
7556
7557stats enable
7558 Enable statistics reporting with default settings
7559 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007560 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007561 Arguments : none
7562
7563 This statement enables statistics reporting with default settings defined
7564 at build time. Unless stated otherwise, these settings are used :
7565 - stats uri : /haproxy?stats
7566 - stats realm : "HAProxy Statistics"
7567 - stats auth : no authentication
7568 - stats scope : no restriction
7569
7570 Though this statement alone is enough to enable statistics reporting, it is
7571 recommended to set all other settings in order to avoid relying on default
7572 unobvious parameters.
7573
7574 Example :
7575 # public access (limited to this backend only)
7576 backend public_www
7577 server srv1 192.168.0.1:80
7578 stats enable
7579 stats hide-version
7580 stats scope .
7581 stats uri /admin?stats
7582 stats realm Haproxy\ Statistics
7583 stats auth admin1:AdMiN123
7584 stats auth admin2:AdMiN321
7585
7586 # internal monitoring access (unlimited)
7587 backend private_monitoring
7588 stats enable
7589 stats uri /admin?stats
7590 stats refresh 5s
7591
7592 See also : "stats auth", "stats realm", "stats uri"
7593
7594
Willy Tarreaud63335a2010-02-26 12:56:52 +01007595stats hide-version
7596 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007597 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007598 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01007599 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007600
Willy Tarreaud63335a2010-02-26 12:56:52 +01007601 By default, the stats page reports some useful status information along with
7602 the statistics. Among them is HAProxy's version. However, it is generally
7603 considered dangerous to report precise version to anyone, as it can help them
7604 target known weaknesses with specific attacks. The "stats hide-version"
7605 statement removes the version from the statistics report. This is recommended
7606 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007607
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02007608 Though this statement alone is enough to enable statistics reporting, it is
7609 recommended to set all other settings in order to avoid relying on default
7610 unobvious parameters.
7611
Willy Tarreaud63335a2010-02-26 12:56:52 +01007612 Example :
7613 # public access (limited to this backend only)
7614 backend public_www
7615 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02007616 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01007617 stats hide-version
7618 stats scope .
7619 stats uri /admin?stats
7620 stats realm Haproxy\ Statistics
7621 stats auth admin1:AdMiN123
7622 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007623
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007624 # internal monitoring access (unlimited)
7625 backend private_monitoring
7626 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01007627 stats uri /admin?stats
7628 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +01007629
Willy Tarreaud63335a2010-02-26 12:56:52 +01007630 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007631
Willy Tarreau983e01e2010-01-11 18:42:06 +01007632
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02007633stats http-request { allow | deny | auth [realm <realm>] }
7634 [ { if | unless } <condition> ]
7635 Access control for statistics
7636
7637 May be used in sections: defaults | frontend | listen | backend
7638 no | no | yes | yes
7639
7640 As "http-request", these set of options allow to fine control access to
7641 statistics. Each option may be followed by if/unless and acl.
7642 First option with matched condition (or option without condition) is final.
7643 For "deny" a 403 error will be returned, for "allow" normal processing is
7644 performed, for "auth" a 401/407 error code is returned so the client
7645 should be asked to enter a username and password.
7646
7647 There is no fixed limit to the number of http-request statements per
7648 instance.
7649
7650 See also : "http-request", section 3.4 about userlists and section 7
7651 about ACL usage.
7652
7653
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007654stats realm <realm>
7655 Enable statistics and set authentication realm
7656 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007657 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007658 Arguments :
7659 <realm> is the name of the HTTP Basic Authentication realm reported to
7660 the browser. The browser uses it to display it in the pop-up
7661 inviting the user to enter a valid username and password.
7662
7663 The realm is read as a single word, so any spaces in it should be escaped
7664 using a backslash ('\').
7665
7666 This statement is useful only in conjunction with "stats auth" since it is
7667 only related to authentication.
7668
7669 Though this statement alone is enough to enable statistics reporting, it is
7670 recommended to set all other settings in order to avoid relying on default
7671 unobvious parameters.
7672
7673 Example :
7674 # public access (limited to this backend only)
7675 backend public_www
7676 server srv1 192.168.0.1:80
7677 stats enable
7678 stats hide-version
7679 stats scope .
7680 stats uri /admin?stats
7681 stats realm Haproxy\ Statistics
7682 stats auth admin1:AdMiN123
7683 stats auth admin2:AdMiN321
7684
7685 # internal monitoring access (unlimited)
7686 backend private_monitoring
7687 stats enable
7688 stats uri /admin?stats
7689 stats refresh 5s
7690
7691 See also : "stats auth", "stats enable", "stats uri"
7692
7693
7694stats refresh <delay>
7695 Enable statistics with automatic refresh
7696 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007697 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007698 Arguments :
7699 <delay> is the suggested refresh delay, specified in seconds, which will
7700 be returned to the browser consulting the report page. While the
7701 browser is free to apply any delay, it will generally respect it
7702 and refresh the page this every seconds. The refresh interval may
7703 be specified in any other non-default time unit, by suffixing the
7704 unit after the value, as explained at the top of this document.
7705
7706 This statement is useful on monitoring displays with a permanent page
7707 reporting the load balancer's activity. When set, the HTML report page will
7708 include a link "refresh"/"stop refresh" so that the user can select whether
7709 he wants automatic refresh of the page or not.
7710
7711 Though this statement alone is enough to enable statistics reporting, it is
7712 recommended to set all other settings in order to avoid relying on default
7713 unobvious parameters.
7714
7715 Example :
7716 # public access (limited to this backend only)
7717 backend public_www
7718 server srv1 192.168.0.1:80
7719 stats enable
7720 stats hide-version
7721 stats scope .
7722 stats uri /admin?stats
7723 stats realm Haproxy\ Statistics
7724 stats auth admin1:AdMiN123
7725 stats auth admin2:AdMiN321
7726
7727 # internal monitoring access (unlimited)
7728 backend private_monitoring
7729 stats enable
7730 stats uri /admin?stats
7731 stats refresh 5s
7732
7733 See also : "stats auth", "stats enable", "stats realm", "stats uri"
7734
7735
7736stats scope { <name> | "." }
7737 Enable statistics and limit access scope
7738 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007739 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007740 Arguments :
7741 <name> is the name of a listen, frontend or backend section to be
7742 reported. The special name "." (a single dot) designates the
7743 section in which the statement appears.
7744
7745 When this statement is specified, only the sections enumerated with this
7746 statement will appear in the report. All other ones will be hidden. This
7747 statement may appear as many times as needed if multiple sections need to be
7748 reported. Please note that the name checking is performed as simple string
7749 comparisons, and that it is never checked that a give section name really
7750 exists.
7751
7752 Though this statement alone is enough to enable statistics reporting, it is
7753 recommended to set all other settings in order to avoid relying on default
7754 unobvious parameters.
7755
7756 Example :
7757 # public access (limited to this backend only)
7758 backend public_www
7759 server srv1 192.168.0.1:80
7760 stats enable
7761 stats hide-version
7762 stats scope .
7763 stats uri /admin?stats
7764 stats realm Haproxy\ Statistics
7765 stats auth admin1:AdMiN123
7766 stats auth admin2:AdMiN321
7767
7768 # internal monitoring access (unlimited)
7769 backend private_monitoring
7770 stats enable
7771 stats uri /admin?stats
7772 stats refresh 5s
7773
7774 See also : "stats auth", "stats enable", "stats realm", "stats uri"
7775
Willy Tarreaud63335a2010-02-26 12:56:52 +01007776
Willy Tarreauc9705a12010-07-27 20:05:50 +02007777stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +01007778 Enable reporting of a description on the statistics page.
7779 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007780 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01007781
Willy Tarreauc9705a12010-07-27 20:05:50 +02007782 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +01007783 description from global section is automatically used instead.
7784
7785 This statement is useful for users that offer shared services to their
7786 customers, where node or description should be different for each customer.
7787
7788 Though this statement alone is enough to enable statistics reporting, it is
7789 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007790 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +01007791
7792 Example :
7793 # internal monitoring access (unlimited)
7794 backend private_monitoring
7795 stats enable
7796 stats show-desc Master node for Europe, Asia, Africa
7797 stats uri /admin?stats
7798 stats refresh 5s
7799
7800 See also: "show-node", "stats enable", "stats uri" and "description" in
7801 global section.
7802
7803
7804stats show-legends
Willy Tarreaued2119c2014-04-24 22:10:39 +02007805 Enable reporting additional information on the statistics page
7806 May be used in sections : defaults | frontend | listen | backend
7807 yes | yes | yes | yes
7808 Arguments : none
7809
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007810 Enable reporting additional information on the statistics page :
Willy Tarreaud63335a2010-02-26 12:56:52 +01007811 - cap: capabilities (proxy)
7812 - mode: one of tcp, http or health (proxy)
7813 - id: SNMP ID (proxy, socket, server)
7814 - IP (socket, server)
7815 - cookie (backend, server)
7816
7817 Though this statement alone is enough to enable statistics reporting, it is
7818 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007819 unobvious parameters. Default behaviour is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +01007820
7821 See also: "stats enable", "stats uri".
7822
7823
7824stats show-node [ <name> ]
7825 Enable reporting of a host name on the statistics page.
7826 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007827 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01007828 Arguments:
7829 <name> is an optional name to be reported. If unspecified, the
7830 node name from global section is automatically used instead.
7831
7832 This statement is useful for users that offer shared services to their
7833 customers, where node or description might be different on a stats page
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007834 provided for each customer. Default behaviour is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +01007835
7836 Though this statement alone is enough to enable statistics reporting, it is
7837 recommended to set all other settings in order to avoid relying on default
7838 unobvious parameters.
7839
7840 Example:
7841 # internal monitoring access (unlimited)
7842 backend private_monitoring
7843 stats enable
7844 stats show-node Europe-1
7845 stats uri /admin?stats
7846 stats refresh 5s
7847
7848 See also: "show-desc", "stats enable", "stats uri", and "node" in global
7849 section.
7850
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007851
7852stats uri <prefix>
7853 Enable statistics and define the URI prefix to access them
7854 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007855 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007856 Arguments :
7857 <prefix> is the prefix of any URI which will be redirected to stats. This
7858 prefix may contain a question mark ('?') to indicate part of a
7859 query string.
7860
7861 The statistics URI is intercepted on the relayed traffic, so it appears as a
7862 page within the normal application. It is strongly advised to ensure that the
7863 selected URI will never appear in the application, otherwise it will never be
7864 possible to reach it in the application.
7865
7866 The default URI compiled in haproxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007867 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007868 It is generally a good idea to include a question mark in the URI so that
7869 intermediate proxies refrain from caching the results. Also, since any string
7870 beginning with the prefix will be accepted as a stats request, the question
7871 mark helps ensuring that no valid URI will begin with the same words.
7872
7873 It is sometimes very convenient to use "/" as the URI prefix, and put that
7874 statement in a "listen" instance of its own. That makes it easy to dedicate
7875 an address or a port to statistics only.
7876
7877 Though this statement alone is enough to enable statistics reporting, it is
7878 recommended to set all other settings in order to avoid relying on default
7879 unobvious parameters.
7880
7881 Example :
7882 # public access (limited to this backend only)
7883 backend public_www
7884 server srv1 192.168.0.1:80
7885 stats enable
7886 stats hide-version
7887 stats scope .
7888 stats uri /admin?stats
7889 stats realm Haproxy\ Statistics
7890 stats auth admin1:AdMiN123
7891 stats auth admin2:AdMiN321
7892
7893 # internal monitoring access (unlimited)
7894 backend private_monitoring
7895 stats enable
7896 stats uri /admin?stats
7897 stats refresh 5s
7898
7899 See also : "stats auth", "stats enable", "stats realm"
7900
7901
Willy Tarreaud63335a2010-02-26 12:56:52 +01007902stick match <pattern> [table <table>] [{if | unless} <cond>]
7903 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007904 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +01007905 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007906
7907 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02007908 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007909 describes what elements of the incoming request or connection
7910 will be analysed in the hope to find a matching entry in a
7911 stickiness table. This rule is mandatory.
7912
7913 <table> is an optional stickiness table name. If unspecified, the same
7914 backend's table is used. A stickiness table is declared using
7915 the "stick-table" statement.
7916
7917 <cond> is an optional matching condition. It makes it possible to match
7918 on a certain criterion only when other conditions are met (or
7919 not met). For instance, it could be used to match on a source IP
7920 address except when a request passes through a known proxy, in
7921 which case we'd match on a header containing that IP address.
7922
7923 Some protocols or applications require complex stickiness rules and cannot
7924 always simply rely on cookies nor hashing. The "stick match" statement
7925 describes a rule to extract the stickiness criterion from an incoming request
7926 or connection. See section 7 for a complete list of possible patterns and
7927 transformation rules.
7928
7929 The table has to be declared using the "stick-table" statement. It must be of
7930 a type compatible with the pattern. By default it is the one which is present
7931 in the same backend. It is possible to share a table with other backends by
7932 referencing it using the "table" keyword. If another table is referenced,
7933 the server's ID inside the backends are used. By default, all server IDs
7934 start at 1 in each backend, so the server ordering is enough. But in case of
7935 doubt, it is highly recommended to force server IDs using their "id" setting.
7936
7937 It is possible to restrict the conditions where a "stick match" statement
7938 will apply, using "if" or "unless" followed by a condition. See section 7 for
7939 ACL based conditions.
7940
7941 There is no limit on the number of "stick match" statements. The first that
7942 applies and matches will cause the request to be directed to the same server
7943 as was used for the request which created the entry. That way, multiple
7944 matches can be used as fallbacks.
7945
7946 The stick rules are checked after the persistence cookies, so they will not
7947 affect stickiness if a cookie has already been used to select a server. That
7948 way, it becomes very easy to insert cookies and match on IP addresses in
7949 order to maintain stickiness between HTTP and HTTPS.
7950
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007951 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7952 unless you know what you do : memory is not shared between the
7953 processes, which can result in random behaviours.
7954
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007955 Example :
7956 # forward SMTP users to the same server they just used for POP in the
7957 # last 30 minutes
7958 backend pop
7959 mode tcp
7960 balance roundrobin
7961 stick store-request src
7962 stick-table type ip size 200k expire 30m
7963 server s1 192.168.1.1:110
7964 server s2 192.168.1.1:110
7965
7966 backend smtp
7967 mode tcp
7968 balance roundrobin
7969 stick match src table pop
7970 server s1 192.168.1.1:25
7971 server s2 192.168.1.1:25
7972
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007973 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02007974 about ACLs and samples fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007975
7976
7977stick on <pattern> [table <table>] [{if | unless} <condition>]
7978 Define a request pattern to associate a user to a server
7979 May be used in sections : defaults | frontend | listen | backend
7980 no | no | yes | yes
7981
7982 Note : This form is exactly equivalent to "stick match" followed by
7983 "stick store-request", all with the same arguments. Please refer
7984 to both keywords for details. It is only provided as a convenience
7985 for writing more maintainable configurations.
7986
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007987 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7988 unless you know what you do : memory is not shared between the
7989 processes, which can result in random behaviours.
7990
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007991 Examples :
7992 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +01007993 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007994
7995 # ...is strictly equivalent to this one :
7996 stick match src table pop if !localhost
7997 stick store-request src table pop if !localhost
7998
7999
8000 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
8001 # well as HTTP without cookie. Share the same table between both accesses.
8002 backend http
8003 mode http
8004 balance roundrobin
8005 stick on src table https
8006 cookie SRV insert indirect nocache
8007 server s1 192.168.1.1:80 cookie s1
8008 server s2 192.168.1.1:80 cookie s2
8009
8010 backend https
8011 mode tcp
8012 balance roundrobin
8013 stick-table type ip size 200k expire 30m
8014 stick on src
8015 server s1 192.168.1.1:443
8016 server s2 192.168.1.1:443
8017
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008018 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008019
8020
8021stick store-request <pattern> [table <table>] [{if | unless} <condition>]
8022 Define a request pattern used to create an entry in a stickiness table
8023 May be used in sections : defaults | frontend | listen | backend
8024 no | no | yes | yes
8025
8026 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008027 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008028 describes what elements of the incoming request or connection
8029 will be analysed, extracted and stored in the table once a
8030 server is selected.
8031
8032 <table> is an optional stickiness table name. If unspecified, the same
8033 backend's table is used. A stickiness table is declared using
8034 the "stick-table" statement.
8035
8036 <cond> is an optional storage condition. It makes it possible to store
8037 certain criteria only when some conditions are met (or not met).
8038 For instance, it could be used to store the source IP address
8039 except when the request passes through a known proxy, in which
8040 case we'd store a converted form of a header containing that IP
8041 address.
8042
8043 Some protocols or applications require complex stickiness rules and cannot
8044 always simply rely on cookies nor hashing. The "stick store-request" statement
8045 describes a rule to decide what to extract from the request and when to do
8046 it, in order to store it into a stickiness table for further requests to
8047 match it using the "stick match" statement. Obviously the extracted part must
8048 make sense and have a chance to be matched in a further request. Storing a
8049 client's IP address for instance often makes sense. Storing an ID found in a
8050 URL parameter also makes sense. Storing a source port will almost never make
8051 any sense because it will be randomly matched. See section 7 for a complete
8052 list of possible patterns and transformation rules.
8053
8054 The table has to be declared using the "stick-table" statement. It must be of
8055 a type compatible with the pattern. By default it is the one which is present
8056 in the same backend. It is possible to share a table with other backends by
8057 referencing it using the "table" keyword. If another table is referenced,
8058 the server's ID inside the backends are used. By default, all server IDs
8059 start at 1 in each backend, so the server ordering is enough. But in case of
8060 doubt, it is highly recommended to force server IDs using their "id" setting.
8061
8062 It is possible to restrict the conditions where a "stick store-request"
8063 statement will apply, using "if" or "unless" followed by a condition. This
8064 condition will be evaluated while parsing the request, so any criteria can be
8065 used. See section 7 for ACL based conditions.
8066
8067 There is no limit on the number of "stick store-request" statements, but
8068 there is a limit of 8 simultaneous stores per request or response. This
8069 makes it possible to store up to 8 criteria, all extracted from either the
8070 request or the response, regardless of the number of rules. Only the 8 first
8071 ones which match will be kept. Using this, it is possible to feed multiple
8072 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01008073 another protocol or access method. Using multiple store-request rules with
8074 the same table is possible and may be used to find the best criterion to rely
8075 on, by arranging the rules by decreasing preference order. Only the first
8076 extracted criterion for a given table will be stored. All subsequent store-
8077 request rules referencing the same table will be skipped and their ACLs will
8078 not be evaluated.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008079
8080 The "store-request" rules are evaluated once the server connection has been
8081 established, so that the table will contain the real server that processed
8082 the request.
8083
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008084 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8085 unless you know what you do : memory is not shared between the
8086 processes, which can result in random behaviours.
8087
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008088 Example :
8089 # forward SMTP users to the same server they just used for POP in the
8090 # last 30 minutes
8091 backend pop
8092 mode tcp
8093 balance roundrobin
8094 stick store-request src
8095 stick-table type ip size 200k expire 30m
8096 server s1 192.168.1.1:110
8097 server s2 192.168.1.1:110
8098
8099 backend smtp
8100 mode tcp
8101 balance roundrobin
8102 stick match src table pop
8103 server s1 192.168.1.1:25
8104 server s2 192.168.1.1:25
8105
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008106 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02008107 about ACLs and sample fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008108
8109
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008110stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Emeric Brunf099e792010-09-27 12:05:28 +02008111 size <size> [expire <expire>] [nopurge] [peers <peersect>]
8112 [store <data_type>]*
Godbach64cef792013-12-04 16:08:22 +08008113 Configure the stickiness table for the current section
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008114 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +02008115 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008116
8117 Arguments :
8118 ip a table declared with "type ip" will only store IPv4 addresses.
8119 This form is very compact (about 50 bytes per entry) and allows
8120 very fast entry lookup and stores with almost no overhead. This
8121 is mainly used to store client source IP addresses.
8122
David du Colombier9a6d3c92011-03-17 10:40:24 +01008123 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
8124 This form is very compact (about 60 bytes per entry) and allows
8125 very fast entry lookup and stores with almost no overhead. This
8126 is mainly used to store client source IP addresses.
8127
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008128 integer a table declared with "type integer" will store 32bit integers
8129 which can represent a client identifier found in a request for
8130 instance.
8131
8132 string a table declared with "type string" will store substrings of up
8133 to <len> characters. If the string provided by the pattern
8134 extractor is larger than <len>, it will be truncated before
8135 being stored. During matching, at most <len> characters will be
8136 compared between the string in the table and the extracted
8137 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008138 to 32 characters.
8139
8140 binary a table declared with "type binary" will store binary blocks
8141 of <len> bytes. If the block provided by the pattern
8142 extractor is larger than <len>, it will be truncated before
Willy Tarreaube722a22014-06-13 16:31:59 +02008143 being stored. If the block provided by the sample expression
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008144 is shorter than <len>, it will be padded by 0. When not
8145 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008146
8147 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008148 "string" type table (See type "string" above). Or the number
8149 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008150 changing this parameter as memory usage will proportionally
8151 increase.
8152
8153 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +01008154 value directly impacts memory usage. Count approximately
8155 50 bytes per entry, plus the size of a string if any. The size
8156 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008157
8158 [nopurge] indicates that we refuse to purge older entries when the table
8159 is full. When not specified and the table is full when haproxy
8160 wants to store an entry in it, it will flush a few of the oldest
8161 entries in order to release some space for the new ones. This is
8162 most often the desired behaviour. In some specific cases, it
8163 be desirable to refuse new entries instead of purging the older
8164 ones. That may be the case when the amount of data to store is
8165 far above the hardware limits and we prefer not to offer access
8166 to new clients than to reject the ones already connected. When
8167 using this parameter, be sure to properly set the "expire"
8168 parameter (see below).
8169
Emeric Brunf099e792010-09-27 12:05:28 +02008170 <peersect> is the name of the peers section to use for replication. Entries
8171 which associate keys to server IDs are kept synchronized with
8172 the remote peers declared in this section. All entries are also
8173 automatically learned from the local peer (old process) during a
8174 soft restart.
8175
Willy Tarreau1abc6732015-05-01 19:21:02 +02008176 NOTE : each peers section may be referenced only by tables
8177 belonging to the same unique process.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008178
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008179 <expire> defines the maximum duration of an entry in the table since it
8180 was last created, refreshed or matched. The expiration delay is
8181 defined using the standard time format, similarly as the various
8182 timeouts. The maximum duration is slightly above 24 days. See
8183 section 2.2 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +02008184 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008185 be removed once full. Be sure not to use the "nopurge" parameter
8186 if not expiration delay is specified.
8187
Willy Tarreau08d5f982010-06-06 13:34:54 +02008188 <data_type> is used to store additional information in the stick-table. This
8189 may be used by ACLs in order to control various criteria related
8190 to the activity of the client matching the stick-table. For each
8191 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +02008192 that the additional data can fit. Several data types may be
8193 stored with an entry. Multiple data types may be specified after
8194 the "store" keyword, as a comma-separated list. Alternatively,
8195 it is possible to repeat the "store" keyword followed by one or
8196 several data types. Except for the "server_id" type which is
8197 automatically detected and enabled, all data types must be
8198 explicitly declared to be stored. If an ACL references a data
8199 type which is not stored, the ACL will simply not match. Some
8200 data types require an argument which must be passed just after
8201 the type between parenthesis. See below for the supported data
8202 types and their arguments.
8203
8204 The data types that can be stored with an entry are the following :
8205 - server_id : this is an integer which holds the numeric ID of the server a
8206 request was assigned to. It is used by the "stick match", "stick store",
8207 and "stick on" rules. It is automatically enabled when referenced.
8208
8209 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
8210 integer which may be used for anything. Most of the time it will be used
8211 to put a special tag on some entries, for instance to note that a
8212 specific behaviour was detected and must be known for future matches.
8213
Willy Tarreauba2ffd12013-05-29 15:54:14 +02008214 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
8215 over a period. It is a positive 32-bit integer integer which may be used
8216 for anything. Just like <gpc0>, it counts events, but instead of keeping
8217 a cumulative count, it maintains the rate at which the counter is
8218 incremented. Most of the time it will be used to measure the frequency of
8219 occurrence of certain events (eg: requests to a specific URL).
8220
Willy Tarreauc9705a12010-07-27 20:05:50 +02008221 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
8222 the absolute number of connections received from clients which matched
8223 this entry. It does not mean the connections were accepted, just that
8224 they were received.
8225
8226 - conn_cur : Current Connections. It is a positive 32-bit integer which
8227 stores the concurrent connection counts for the entry. It is incremented
8228 once an incoming connection matches the entry, and decremented once the
8229 connection leaves. That way it is possible to know at any time the exact
8230 number of concurrent connections for an entry.
8231
8232 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8233 integer parameter <period> which indicates in milliseconds the length
8234 of the period over which the average is measured. It reports the average
8235 incoming connection rate over that period, in connections per period. The
8236 result is an integer which can be matched using ACLs.
8237
8238 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
8239 the absolute number of sessions received from clients which matched this
8240 entry. A session is a connection that was accepted by the layer 4 rules.
8241
8242 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8243 integer parameter <period> which indicates in milliseconds the length
8244 of the period over which the average is measured. It reports the average
8245 incoming session rate over that period, in sessions per period. The
8246 result is an integer which can be matched using ACLs.
8247
8248 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
8249 counts the absolute number of HTTP requests received from clients which
8250 matched this entry. It does not matter whether they are valid requests or
8251 not. Note that this is different from sessions when keep-alive is used on
8252 the client side.
8253
8254 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8255 integer parameter <period> which indicates in milliseconds the length
8256 of the period over which the average is measured. It reports the average
8257 HTTP request rate over that period, in requests per period. The result is
8258 an integer which can be matched using ACLs. It does not matter whether
8259 they are valid requests or not. Note that this is different from sessions
8260 when keep-alive is used on the client side.
8261
8262 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
8263 counts the absolute number of HTTP requests errors induced by clients
8264 which matched this entry. Errors are counted on invalid and truncated
8265 requests, as well as on denied or tarpitted requests, and on failed
8266 authentications. If the server responds with 4xx, then the request is
8267 also counted as an error since it's an error triggered by the client
8268 (eg: vulnerability scan).
8269
8270 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8271 integer parameter <period> which indicates in milliseconds the length
8272 of the period over which the average is measured. It reports the average
8273 HTTP request error rate over that period, in requests per period (see
8274 http_err_cnt above for what is accounted as an error). The result is an
8275 integer which can be matched using ACLs.
8276
8277 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
8278 integer which counts the cumulated amount of bytes received from clients
8279 which matched this entry. Headers are included in the count. This may be
8280 used to limit abuse of upload features on photo or video servers.
8281
8282 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8283 integer parameter <period> which indicates in milliseconds the length
8284 of the period over which the average is measured. It reports the average
8285 incoming bytes rate over that period, in bytes per period. It may be used
8286 to detect users which upload too much and too fast. Warning: with large
8287 uploads, it is possible that the amount of uploaded data will be counted
8288 once upon termination, thus causing spikes in the average transfer speed
8289 instead of having a smooth one. This may partially be smoothed with
8290 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
8291 recommended for better fairness.
8292
8293 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
8294 integer which counts the cumulated amount of bytes sent to clients which
8295 matched this entry. Headers are included in the count. This may be used
8296 to limit abuse of bots sucking the whole site.
8297
8298 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
8299 an integer parameter <period> which indicates in milliseconds the length
8300 of the period over which the average is measured. It reports the average
8301 outgoing bytes rate over that period, in bytes per period. It may be used
8302 to detect users which download too much and too fast. Warning: with large
8303 transfers, it is possible that the amount of transferred data will be
8304 counted once upon termination, thus causing spikes in the average
8305 transfer speed instead of having a smooth one. This may partially be
8306 smoothed with "option contstats" though this is not perfect yet. Use of
8307 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +02008308
Willy Tarreauc00cdc22010-06-06 16:48:26 +02008309 There is only one stick-table per proxy. At the moment of writing this doc,
8310 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008311 to be required, simply create a dummy backend with a stick-table in it and
8312 reference it.
8313
8314 It is important to understand that stickiness based on learning information
8315 has some limitations, including the fact that all learned associations are
Baptiste Assmann123ff042016-03-06 23:29:28 +01008316 lost upon restart unless peers are properly configured to transfer such
8317 information upon restart (recommended). In general it can be good as a
8318 complement but not always as an exclusive stickiness.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008319
Willy Tarreauc9705a12010-07-27 20:05:50 +02008320 Last, memory requirements may be important when storing many data types.
8321 Indeed, storing all indicators above at once in each entry requires 116 bytes
8322 per entry, or 116 MB for a 1-million entries table. This is definitely not
8323 something that can be ignored.
8324
8325 Example:
8326 # Keep track of counters of up to 1 million IP addresses over 5 minutes
8327 # and store a general purpose counter and the average connection rate
8328 # computed over a sliding window of 30 seconds.
8329 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
8330
8331 See also : "stick match", "stick on", "stick store-request", section 2.2
David du Colombiera13d1b92011-03-17 10:40:22 +01008332 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008333
8334
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008335stick store-response <pattern> [table <table>] [{if | unless} <condition>]
Baptiste Assmann2f2d2ec2016-03-06 23:27:24 +01008336 Define a response pattern used to create an entry in a stickiness table
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008337 May be used in sections : defaults | frontend | listen | backend
8338 no | no | yes | yes
8339
8340 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008341 <pattern> is a sample expression rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008342 describes what elements of the response or connection will
8343 be analysed, extracted and stored in the table once a
8344 server is selected.
8345
8346 <table> is an optional stickiness table name. If unspecified, the same
8347 backend's table is used. A stickiness table is declared using
8348 the "stick-table" statement.
8349
8350 <cond> is an optional storage condition. It makes it possible to store
8351 certain criteria only when some conditions are met (or not met).
8352 For instance, it could be used to store the SSL session ID only
8353 when the response is a SSL server hello.
8354
8355 Some protocols or applications require complex stickiness rules and cannot
8356 always simply rely on cookies nor hashing. The "stick store-response"
8357 statement describes a rule to decide what to extract from the response and
8358 when to do it, in order to store it into a stickiness table for further
8359 requests to match it using the "stick match" statement. Obviously the
8360 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008361 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008362 See section 7 for a complete list of possible patterns and transformation
8363 rules.
8364
8365 The table has to be declared using the "stick-table" statement. It must be of
8366 a type compatible with the pattern. By default it is the one which is present
8367 in the same backend. It is possible to share a table with other backends by
8368 referencing it using the "table" keyword. If another table is referenced,
8369 the server's ID inside the backends are used. By default, all server IDs
8370 start at 1 in each backend, so the server ordering is enough. But in case of
8371 doubt, it is highly recommended to force server IDs using their "id" setting.
8372
8373 It is possible to restrict the conditions where a "stick store-response"
8374 statement will apply, using "if" or "unless" followed by a condition. This
8375 condition will be evaluated while parsing the response, so any criteria can
8376 be used. See section 7 for ACL based conditions.
8377
8378 There is no limit on the number of "stick store-response" statements, but
8379 there is a limit of 8 simultaneous stores per request or response. This
8380 makes it possible to store up to 8 criteria, all extracted from either the
8381 request or the response, regardless of the number of rules. Only the 8 first
8382 ones which match will be kept. Using this, it is possible to feed multiple
8383 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01008384 another protocol or access method. Using multiple store-response rules with
8385 the same table is possible and may be used to find the best criterion to rely
8386 on, by arranging the rules by decreasing preference order. Only the first
8387 extracted criterion for a given table will be stored. All subsequent store-
8388 response rules referencing the same table will be skipped and their ACLs will
8389 not be evaluated. However, even if a store-request rule references a table, a
8390 store-response rule may also use the same table. This means that each table
8391 may learn exactly one element from the request and one element from the
8392 response at once.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008393
8394 The table will contain the real server that processed the request.
8395
8396 Example :
8397 # Learn SSL session ID from both request and response and create affinity.
8398 backend https
8399 mode tcp
8400 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +02008401 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008402 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008403
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008404 acl clienthello req_ssl_hello_type 1
8405 acl serverhello rep_ssl_hello_type 2
8406
8407 # use tcp content accepts to detects ssl client and server hello.
8408 tcp-request inspect-delay 5s
8409 tcp-request content accept if clienthello
8410
8411 # no timeout on response inspect delay by default.
8412 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008413
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008414 # SSL session ID (SSLID) may be present on a client or server hello.
8415 # Its length is coded on 1 byte at offset 43 and its value starts
8416 # at offset 44.
8417
8418 # Match and learn on request if client hello.
8419 stick on payload_lv(43,1) if clienthello
8420
8421 # Learn on response if server hello.
8422 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +02008423
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008424 server s1 192.168.1.1:443
8425 server s2 192.168.1.1:443
8426
8427 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
8428 extraction.
8429
8430
Willy Tarreau938c7fe2014-04-25 14:21:39 +02008431tcp-check connect [params*]
8432 Opens a new connection
8433 May be used in sections: defaults | frontend | listen | backend
8434 no | no | yes | yes
8435
8436 When an application lies on more than a single TCP port or when HAProxy
8437 load-balance many services in a single backend, it makes sense to probe all
8438 the services individually before considering a server as operational.
8439
8440 When there are no TCP port configured on the server line neither server port
8441 directive, then the 'tcp-check connect port <port>' must be the first step
8442 of the sequence.
8443
8444 In a tcp-check ruleset a 'connect' is required, it is also mandatory to start
8445 the ruleset with a 'connect' rule. Purpose is to ensure admin know what they
8446 do.
8447
8448 Parameters :
8449 They are optional and can be used to describe how HAProxy should open and
8450 use the TCP connection.
8451
8452 port if not set, check port or server port is used.
8453 It tells HAProxy where to open the connection to.
8454 <port> must be a valid TCP port source integer, from 1 to 65535.
8455
8456 send-proxy send a PROXY protocol string
8457
8458 ssl opens a ciphered connection
8459
8460 Examples:
8461 # check HTTP and HTTPs services on a server.
8462 # first open port 80 thanks to server line port directive, then
8463 # tcp-check opens port 443, ciphered and run a request on it:
8464 option tcp-check
8465 tcp-check connect
8466 tcp-check send GET\ /\ HTTP/1.0\r\n
8467 tcp-check send Host:\ haproxy.1wt.eu\r\n
8468 tcp-check send \r\n
8469 tcp-check expect rstring (2..|3..)
8470 tcp-check connect port 443 ssl
8471 tcp-check send GET\ /\ HTTP/1.0\r\n
8472 tcp-check send Host:\ haproxy.1wt.eu\r\n
8473 tcp-check send \r\n
8474 tcp-check expect rstring (2..|3..)
8475 server www 10.0.0.1 check port 80
8476
8477 # check both POP and IMAP from a single server:
8478 option tcp-check
8479 tcp-check connect port 110
8480 tcp-check expect string +OK\ POP3\ ready
8481 tcp-check connect port 143
8482 tcp-check expect string *\ OK\ IMAP4\ ready
8483 server mail 10.0.0.1 check
8484
8485 See also : "option tcp-check", "tcp-check send", "tcp-check expect"
8486
8487
8488tcp-check expect [!] <match> <pattern>
8489 Specify data to be collected and analysed during a generic health check
8490 May be used in sections: defaults | frontend | listen | backend
8491 no | no | yes | yes
8492
8493 Arguments :
8494 <match> is a keyword indicating how to look for a specific pattern in the
8495 response. The keyword may be one of "string", "rstring" or
8496 binary.
8497 The keyword may be preceded by an exclamation mark ("!") to negate
8498 the match. Spaces are allowed between the exclamation mark and the
8499 keyword. See below for more details on the supported keywords.
8500
8501 <pattern> is the pattern to look for. It may be a string or a regular
8502 expression. If the pattern contains spaces, they must be escaped
8503 with the usual backslash ('\').
8504 If the match is set to binary, then the pattern must be passed as
8505 a serie of hexadecimal digits in an even number. Each sequence of
8506 two digits will represent a byte. The hexadecimal digits may be
8507 used upper or lower case.
8508
8509
8510 The available matches are intentionally similar to their http-check cousins :
8511
8512 string <string> : test the exact string matches in the response buffer.
8513 A health check response will be considered valid if the
8514 response's buffer contains this exact string. If the
8515 "string" keyword is prefixed with "!", then the response
8516 will be considered invalid if the body contains this
8517 string. This can be used to look for a mandatory pattern
8518 in a protocol response, or to detect a failure when a
8519 specific error appears in a protocol banner.
8520
8521 rstring <regex> : test a regular expression on the response buffer.
8522 A health check response will be considered valid if the
8523 response's buffer matches this expression. If the
8524 "rstring" keyword is prefixed with "!", then the response
8525 will be considered invalid if the body matches the
8526 expression.
8527
8528 binary <hexstring> : test the exact string in its hexadecimal form matches
8529 in the response buffer. A health check response will
8530 be considered valid if the response's buffer contains
8531 this exact hexadecimal string.
8532 Purpose is to match data on binary protocols.
8533
8534 It is important to note that the responses will be limited to a certain size
8535 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
8536 Thus, too large responses may not contain the mandatory pattern when using
8537 "string", "rstring" or binary. If a large response is absolutely required, it
8538 is possible to change the default max size by setting the global variable.
8539 However, it is worth keeping in mind that parsing very large responses can
8540 waste some CPU cycles, especially when regular expressions are used, and that
8541 it is always better to focus the checks on smaller resources. Also, in its
8542 current state, the check will not find any string nor regex past a null
8543 character in the response. Similarly it is not possible to request matching
8544 the null character.
8545
8546 Examples :
8547 # perform a POP check
8548 option tcp-check
8549 tcp-check expect string +OK\ POP3\ ready
8550
8551 # perform an IMAP check
8552 option tcp-check
8553 tcp-check expect string *\ OK\ IMAP4\ ready
8554
8555 # look for the redis master server
8556 option tcp-check
8557 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02008558 tcp-check expect string +PONG
Willy Tarreau938c7fe2014-04-25 14:21:39 +02008559 tcp-check send info\ replication\r\n
8560 tcp-check expect string role:master
8561 tcp-check send QUIT\r\n
8562 tcp-check expect string +OK
8563
8564
8565 See also : "option tcp-check", "tcp-check connect", "tcp-check send",
8566 "tcp-check send-binary", "http-check expect", tune.chksize
8567
8568
8569tcp-check send <data>
8570 Specify a string to be sent as a question during a generic health check
8571 May be used in sections: defaults | frontend | listen | backend
8572 no | no | yes | yes
8573
8574 <data> : the data to be sent as a question during a generic health check
8575 session. For now, <data> must be a string.
8576
8577 Examples :
8578 # look for the redis master server
8579 option tcp-check
8580 tcp-check send info\ replication\r\n
8581 tcp-check expect string role:master
8582
8583 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
8584 "tcp-check send-binary", tune.chksize
8585
8586
8587tcp-check send-binary <hexastring>
8588 Specify an hexa digits string to be sent as a binary question during a raw
8589 tcp health check
8590 May be used in sections: defaults | frontend | listen | backend
8591 no | no | yes | yes
8592
8593 <data> : the data to be sent as a question during a generic health check
8594 session. For now, <data> must be a string.
8595 <hexastring> : test the exact string in its hexadecimal form matches in the
8596 response buffer. A health check response will be considered
8597 valid if the response's buffer contains this exact
8598 hexadecimal string.
8599 Purpose is to send binary data to ask on binary protocols.
8600
8601 Examples :
8602 # redis check in binary
8603 option tcp-check
8604 tcp-check send-binary 50494e470d0a # PING\r\n
8605 tcp-check expect binary 2b504F4e47 # +PONG
8606
8607
8608 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
8609 "tcp-check send", tune.chksize
8610
8611
Willy Tarreaue9656522010-08-17 15:40:09 +02008612tcp-request connection <action> [{if | unless} <condition>]
8613 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +02008614 May be used in sections : defaults | frontend | listen | backend
8615 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +02008616 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02008617 <action> defines the action to perform if the condition applies. See
8618 below.
Willy Tarreau1a687942010-05-23 22:40:30 +02008619
Willy Tarreaue9656522010-08-17 15:40:09 +02008620 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008621
8622 Immediately after acceptance of a new incoming connection, it is possible to
8623 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +02008624 or dropped or have its counters tracked. Those conditions cannot make use of
8625 any data contents because the connection has not been read from yet, and the
8626 buffers are not yet allocated. This is used to selectively and very quickly
8627 accept or drop connections from various sources with a very low overhead. If
8628 some contents need to be inspected in order to take the decision, the
8629 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008630
Willy Tarreaue9656522010-08-17 15:40:09 +02008631 The "tcp-request connection" rules are evaluated in their exact declaration
8632 order. If no rule matches or if there is no rule, the default action is to
8633 accept the incoming connection. There is no specific limit to the number of
8634 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008635
Willy Tarreaua9083d02015-05-08 15:27:59 +02008636 Four types of actions are supported :
Willy Tarreaue9656522010-08-17 15:40:09 +02008637 - accept :
8638 accepts the connection if the condition is true (when used with "if")
8639 or false (when used with "unless"). The first such rule executed ends
8640 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008641
Willy Tarreaue9656522010-08-17 15:40:09 +02008642 - reject :
8643 rejects the connection if the condition is true (when used with "if")
8644 or false (when used with "unless"). The first such rule executed ends
8645 the rules evaluation. Rejected connections do not even become a
8646 session, which is why they are accounted separately for in the stats,
8647 as "denied connections". They are not considered for the session
8648 rate-limit and are not logged either. The reason is that these rules
8649 should only be used to filter extremely high connection rates such as
8650 the ones encountered during a massive DDoS attack. Under these extreme
8651 conditions, the simple action of logging each event would make the
8652 system collapse and would considerably lower the filtering capacity. If
8653 logging is absolutely desired, then "tcp-request content" rules should
8654 be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008655
Willy Tarreau4f0d9192013-06-11 20:40:55 +02008656 - expect-proxy layer4 :
8657 configures the client-facing connection to receive a PROXY protocol
8658 header before any byte is read from the socket. This is equivalent to
8659 having the "accept-proxy" keyword on the "bind" line, except that using
8660 the TCP rule allows the PROXY protocol to be accepted only for certain
8661 IP address ranges using an ACL. This is convenient when multiple layers
8662 of load balancers are passed through by traffic coming from public
8663 hosts.
8664
Bertrand Jacquin90759682016-06-06 15:35:39 +01008665 - expect-netscaler-cip layer4 :
8666 configures the client-facing connection to receive a NetScaler Client
8667 IP insertion protocol header before any byte is read from the socket.
8668 This is equivalent to having the "accept-netscaler-cip" keyword on the
8669 "bind" line, except that using the TCP rule allows the PROXY protocol
8670 to be accepted only for certain IP address ranges using an ACL. This
8671 is convenient when multiple layers of load balancers are passed
8672 through by traffic coming from public hosts.
8673
Willy Tarreau18bf01e2014-06-13 16:18:52 +02008674 - capture <sample> len <length> :
8675 This only applies to "tcp-request content" rules. It captures sample
8676 expression <sample> from the request buffer, and converts it to a
8677 string of at most <len> characters. The resulting string is stored into
8678 the next request "capture" slot, so it will possibly appear next to
8679 some captured HTTP headers. It will then automatically appear in the
8680 logs, and it will be possible to extract it using sample fetch rules to
8681 feed it into headers or anything. The length should be limited given
8682 that this size will be allocated for each capture during the whole
Willy Tarreaua9083d02015-05-08 15:27:59 +02008683 session life. Please check section 7.3 (Fetching samples) and "capture
8684 request header" for more information.
Willy Tarreau18bf01e2014-06-13 16:18:52 +02008685
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008686 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +02008687 enables tracking of sticky counters from current connection. These
Willy Tarreau09448f72014-06-25 18:12:15 +02008688 rules do not stop evaluation and do not change default action. 3 sets
Willy Tarreaue9656522010-08-17 15:40:09 +02008689 of counters may be simultaneously tracked by the same connection. The
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008690 first "track-sc0" rule executed enables tracking of the counters of the
8691 specified table as the first set. The first "track-sc1" rule executed
Willy Tarreaue9656522010-08-17 15:40:09 +02008692 enables tracking of the counters of the specified table as the second
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008693 set. The first "track-sc2" rule executed enables tracking of the
8694 counters of the specified table as the third set. It is a recommended
8695 practice to use the first set of counters for the per-frontend counters
8696 and the second set for the per-backend ones. But this is just a
8697 guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008698
Willy Tarreaue9656522010-08-17 15:40:09 +02008699 These actions take one or two arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008700 <key> is mandatory, and is a sample expression rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +02008701 in section 7.3. It describes what elements of the incoming
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008702 request or connection will be analysed, extracted, combined,
8703 and used to select which table entry to update the counters.
8704 Note that "tcp-request connection" cannot use content-based
8705 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008706
Willy Tarreaue9656522010-08-17 15:40:09 +02008707 <table> is an optional table to be used instead of the default one,
8708 which is the stick-table declared in the current proxy. All
8709 the counters for the matches and updates for the key will
8710 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008711
Willy Tarreaue9656522010-08-17 15:40:09 +02008712 Once a "track-sc*" rule is executed, the key is looked up in the table
8713 and if it is not found, an entry is allocated for it. Then a pointer to
8714 that entry is kept during all the session's life, and this entry's
8715 counters are updated as often as possible, every time the session's
8716 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008717 Counters are only updated for events that happen after the tracking has
8718 been started. For example, connection counters will not be updated when
8719 tracking layer 7 information, since the connection event happens before
8720 layer7 information is extracted.
8721
Willy Tarreaue9656522010-08-17 15:40:09 +02008722 If the entry tracks concurrent connection counters, one connection is
8723 counted for as long as the entry is tracked, and the entry will not
8724 expire during that time. Tracking counters also provides a performance
8725 advantage over just checking the keys, because only one table lookup is
8726 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008727
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02008728 - sc-inc-gpc0(<sc-id>):
8729 The "sc-inc-gpc0" increments the GPC0 counter according to the sticky
8730 counter designated by <sc-id>. If an error occurs, this action silently
8731 fails and the actions evaluation continues.
8732
Thierry FOURNIER236657b2015-08-19 08:25:14 +02008733 - sc-set-gpt0(<sc-id>) <int>:
8734 This action sets the GPT0 tag according to the sticky counter designated
8735 by <sc-id> and the value of <int>. The expected result is a boolean. If
8736 an error occurs, this action silently fails and the actions evaluation
8737 continues.
8738
William Lallemand2e785f22016-05-25 01:48:42 +02008739 - set-src <expr> :
8740 Is used to set the source IP address to the value of specified
8741 expression. Useful if you want to mask source IP for privacy.
8742 If you want to provide an IP from a HTTP header use "http-request
8743 set-src"
8744
8745 <expr> Is a standard HAProxy expression formed by a sample-fetch
8746 followed by some converters.
8747
8748 Example:
8749
8750 tcp-request connection set-src src,ipmask(24)
8751
8752 When set-src is successful, the source port is set to 0.
8753
William Lallemand44be6402016-05-25 01:51:35 +02008754 - set-src-port <expr> :
8755 Is used to set the source port address to the value of specified
8756 expression.
8757
8758 <expr> Is a standard HAProxy expression formed by a sample-fetch
8759 followed by some converters.
8760
8761 Example:
8762
8763 tcp-request connection set-src-port int(4000)
8764
8765 Be careful to use "set-src-port" after "set-src", because "set-src" sets
8766 the source port to 0.
8767
William Lallemand13e9b0c2016-05-25 02:34:07 +02008768 - set-dst <expr> :
8769 Is used to set the destination IP address to the value of specified
8770 expression. Useful if you want to mask IP for privacy in log.
8771 If you want to provide an IP from a HTTP header use "http-request
8772 set-dst". If you want to connect to the new address/port, use
8773 '0.0.0.0:0' as a server address in the backend.
8774
8775 <expr> Is a standard HAProxy expression formed by a sample-fetch
8776 followed by some converters.
8777
8778 Example:
8779
8780 tcp-request connection set-dst dst,ipmask(24)
8781 tcp-request connection set-dst ipv4(10.0.0.1)
8782
8783 - set-dst-port <expr> :
8784 Is used to set the destination port address to the value of specified
8785 expression. If you want to connect to the new address/port, use
8786 '0.0.0.0:0' as a server address in the backend.
8787
8788
8789 <expr> Is a standard HAProxy expression formed by a sample-fetch
8790 followed by some converters.
8791
8792 Example:
8793
8794 tcp-request connection set-dst-port int(4000)
8795
Willy Tarreau2d392c22015-08-24 01:43:45 +02008796 - "silent-drop" :
8797 This stops the evaluation of the rules and makes the client-facing
8798 connection suddenly disappear using a system-dependant way that tries
8799 to prevent the client from being notified. The effect it then that the
8800 client still sees an established connection while there's none on
8801 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
8802 except that it doesn't use any local resource at all on the machine
8803 running HAProxy. It can resist much higher loads than "tarpit", and
8804 slow down stronger attackers. It is important to undestand the impact
8805 of using this mechanism. All stateful equipments placed between the
8806 client and HAProxy (firewalls, proxies, load balancers) will also keep
8807 the established connection for a long time and may suffer from this
8808 action. On modern Linux systems running with enough privileges, the
8809 TCP_REPAIR socket option is used to block the emission of a TCP
8810 reset. On other systems, the socket's TTL is reduced to 1 so that the
8811 TCP reset doesn't pass the first router, though it's still delivered to
8812 local networks. Do not use it unless you fully understand how it works.
8813
Willy Tarreaue9656522010-08-17 15:40:09 +02008814 Note that the "if/unless" condition is optional. If no condition is set on
8815 the action, it is simply performed unconditionally. That can be useful for
8816 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008817
Willy Tarreaue9656522010-08-17 15:40:09 +02008818 Example: accept all connections from white-listed hosts, reject too fast
8819 connection without counting them, and track accepted connections.
8820 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008821
Willy Tarreaue9656522010-08-17 15:40:09 +02008822 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008823 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008824 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008825
Willy Tarreaue9656522010-08-17 15:40:09 +02008826 Example: accept all connections from white-listed hosts, count all other
8827 connections and reject too fast ones. This results in abusive ones
8828 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008829
Willy Tarreaue9656522010-08-17 15:40:09 +02008830 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008831 tcp-request connection track-sc0 src
8832 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008833
Willy Tarreau4f0d9192013-06-11 20:40:55 +02008834 Example: enable the PROXY protocol for traffic coming from all known proxies.
8835
8836 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
8837
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008838 See section 7 about ACL usage.
8839
Willy Tarreaue9656522010-08-17 15:40:09 +02008840 See also : "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008841
8842
Willy Tarreaue9656522010-08-17 15:40:09 +02008843tcp-request content <action> [{if | unless} <condition>]
8844 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +02008845 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02008846 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +02008847 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02008848 <action> defines the action to perform if the condition applies. See
8849 below.
Willy Tarreau62644772008-07-16 18:36:06 +02008850
Willy Tarreaue9656522010-08-17 15:40:09 +02008851 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +02008852
Willy Tarreaue9656522010-08-17 15:40:09 +02008853 A request's contents can be analysed at an early stage of request processing
8854 called "TCP content inspection". During this stage, ACL-based rules are
8855 evaluated every time the request contents are updated, until either an
8856 "accept" or a "reject" rule matches, or the TCP request inspection delay
8857 expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +02008858
Willy Tarreaue9656522010-08-17 15:40:09 +02008859 The first difference between these rules and "tcp-request connection" rules
8860 is that "tcp-request content" rules can make use of contents to take a
8861 decision. Most often, these decisions will consider a protocol recognition or
8862 validity. The second difference is that content-based rules can be used in
Willy Tarreauf3338342014-01-28 21:40:28 +01008863 both frontends and backends. In case of HTTP keep-alive with the client, all
8864 tcp-request content rules are evaluated again, so haproxy keeps a record of
8865 what sticky counters were assigned by a "tcp-request connection" versus a
8866 "tcp-request content" rule, and flushes all the content-related ones after
8867 processing an HTTP request, so that they may be evaluated again by the rules
8868 being evaluated again for the next request. This is of particular importance
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008869 when the rule tracks some L7 information or when it is conditioned by an
Willy Tarreauf3338342014-01-28 21:40:28 +01008870 L7-based ACL, since tracking may change between requests.
Willy Tarreau62644772008-07-16 18:36:06 +02008871
Willy Tarreaue9656522010-08-17 15:40:09 +02008872 Content-based rules are evaluated in their exact declaration order. If no
8873 rule matches or if there is no rule, the default action is to accept the
8874 contents. There is no specific limit to the number of rules which may be
8875 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +02008876
Thierry FOURNIER236657b2015-08-19 08:25:14 +02008877 Several types of actions are supported :
Willy Tarreau18bf01e2014-06-13 16:18:52 +02008878 - accept : the request is accepted
8879 - reject : the request is rejected and the connection is closed
8880 - capture : the specified sample expression is captured
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008881 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02008882 - sc-inc-gpc0(<sc-id>)
Thierry Fournierb9125672016-03-29 19:34:37 +02008883 - sc-set-gpt0(<sc-id>) <int>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02008884 - set-var(<var-name>) <expr>
Willy Tarreau2d392c22015-08-24 01:43:45 +02008885 - silent-drop
Willy Tarreau62644772008-07-16 18:36:06 +02008886
Willy Tarreaue9656522010-08-17 15:40:09 +02008887 They have the same meaning as their counter-parts in "tcp-request connection"
8888 so please refer to that section for a complete description.
Willy Tarreau62644772008-07-16 18:36:06 +02008889
Willy Tarreauf3338342014-01-28 21:40:28 +01008890 While there is nothing mandatory about it, it is recommended to use the
8891 track-sc0 in "tcp-request connection" rules, track-sc1 for "tcp-request
8892 content" rules in the frontend, and track-sc2 for "tcp-request content"
8893 rules in the backend, because that makes the configuration more readable
8894 and easier to troubleshoot, but this is just a guideline and all counters
8895 may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +02008896
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008897 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +02008898 the action, it is simply performed unconditionally. That can be useful for
8899 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +02008900
Willy Tarreaue9656522010-08-17 15:40:09 +02008901 It is perfectly possible to match layer 7 contents with "tcp-request content"
Willy Tarreauc0239e02012-04-16 14:42:55 +02008902 rules, since HTTP-specific ACL matches are able to preliminarily parse the
8903 contents of a buffer before extracting the required data. If the buffered
8904 contents do not parse as a valid HTTP message, then the ACL does not match.
8905 The parser which is involved there is exactly the same as for all other HTTP
Willy Tarreauf3338342014-01-28 21:40:28 +01008906 processing, so there is no risk of parsing something differently. In an HTTP
8907 backend connected to from an HTTP frontend, it is guaranteed that HTTP
8908 contents will always be immediately present when the rule is evaluated first.
Willy Tarreau62644772008-07-16 18:36:06 +02008909
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008910 Tracking layer7 information is also possible provided that the information
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02008911 are present when the rule is processed. The rule processing engine is able to
8912 wait until the inspect delay expires when the data to be tracked is not yet
8913 available.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008914
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02008915 The "set-var" is used to set the content of a variable. The variable is
8916 declared inline.
8917
Daniel Schneller0b547052016-03-21 20:46:57 +01008918 <var-name> The name of the variable starts with an indication about
8919 its scope. The scopes allowed are:
8920 "sess" : the variable is shared with the whole session
8921 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02008922 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +01008923 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02008924 processing
Daniel Schneller0b547052016-03-21 20:46:57 +01008925 "res" : the variable is shared only during response
8926 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02008927 This prefix is followed by a name. The separator is a '.'.
Daniel Schneller0b547052016-03-21 20:46:57 +01008928 The name may only contain characters 'a-z', 'A-Z', '0-9'
8929 and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02008930
8931 <expr> Is a standard HAProxy expression formed by a sample-fetch
8932 followed by some converters.
8933
8934 Example:
8935
8936 tcp-request content set-var(sess.my_var) src
8937
Willy Tarreau62644772008-07-16 18:36:06 +02008938 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +02008939 # Accept HTTP requests containing a Host header saying "example.com"
8940 # and reject everything else.
8941 acl is_host_com hdr(Host) -i example.com
8942 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +02008943 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +02008944 tcp-request content reject
8945
8946 Example:
Willy Tarreau62644772008-07-16 18:36:06 +02008947 # reject SMTP connection if client speaks first
8948 tcp-request inspect-delay 30s
8949 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008950 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +02008951
8952 # Forward HTTPS connection only if client speaks
8953 tcp-request inspect-delay 30s
8954 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008955 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +02008956 tcp-request content reject
8957
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008958 Example:
8959 # Track the last IP from X-Forwarded-For
8960 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02008961 tcp-request content track-sc0 hdr(x-forwarded-for,-1)
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008962
8963 Example:
8964 # track request counts per "base" (concatenation of Host+URL)
8965 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02008966 tcp-request content track-sc0 base table req-rate
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008967
Willy Tarreaue9656522010-08-17 15:40:09 +02008968 Example: track per-frontend and per-backend counters, block abusers at the
8969 frontend when the backend detects abuse.
8970
8971 frontend http
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008972 # Use General Purpose Couter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +02008973 # protecting all our sites
8974 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008975 tcp-request connection track-sc0 src
8976 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +02008977 ...
8978 use_backend http_dynamic if { path_end .php }
8979
8980 backend http_dynamic
8981 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008982 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +02008983 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008984 acl click_too_fast sc1_http_req_rate gt 10
8985 acl mark_as_abuser sc0_inc_gpc0 gt 0
8986 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +02008987 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +02008988
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008989 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +02008990
Willy Tarreaue9656522010-08-17 15:40:09 +02008991 See also : "tcp-request connection", "tcp-request inspect-delay"
Willy Tarreau62644772008-07-16 18:36:06 +02008992
8993
8994tcp-request inspect-delay <timeout>
8995 Set the maximum allowed time to wait for data during content inspection
8996 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02008997 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +02008998 Arguments :
8999 <timeout> is the timeout value specified in milliseconds by default, but
9000 can be in any other unit if the number is suffixed by the unit,
9001 as explained at the top of this document.
9002
9003 People using haproxy primarily as a TCP relay are often worried about the
9004 risk of passing any type of protocol to a server without any analysis. In
9005 order to be able to analyze the request contents, we must first withhold
9006 the data then analyze them. This statement simply enables withholding of
9007 data for at most the specified amount of time.
9008
Willy Tarreaufb356202010-08-03 14:02:05 +02009009 TCP content inspection applies very early when a connection reaches a
9010 frontend, then very early when the connection is forwarded to a backend. This
9011 means that a connection may experience a first delay in the frontend and a
9012 second delay in the backend if both have tcp-request rules.
9013
Willy Tarreau62644772008-07-16 18:36:06 +02009014 Note that when performing content inspection, haproxy will evaluate the whole
9015 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009016 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +02009017 a last check is performed upon expiration, this time considering that the
Willy Tarreaud869b242009-03-15 14:43:58 +01009018 contents are definitive. If no delay is set, haproxy will not wait at all
9019 and will immediately apply a verdict based on the available information.
9020 Obviously this is unlikely to be very useful and might even be racy, so such
9021 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +02009022
9023 As soon as a rule matches, the request is released and continues as usual. If
9024 the timeout is reached and no rule matches, the default policy will be to let
9025 it pass through unaffected.
9026
9027 For most protocols, it is enough to set it to a few seconds, as most clients
9028 send the full request immediately upon connection. Add 3 or more seconds to
9029 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +01009030 to use large values, for instance to ensure that the client never talks
Willy Tarreau62644772008-07-16 18:36:06 +02009031 before the server (eg: SMTP), or to wait for a client to talk before passing
9032 data to the server (eg: SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +02009033 least the inspection delay, otherwise it will expire first. If the client
9034 closes the connection or if the buffer is full, the delay immediately expires
9035 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +02009036
Willy Tarreau55165fe2009-05-10 12:02:55 +02009037 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +02009038 "timeout client".
9039
9040
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009041tcp-response content <action> [{if | unless} <condition>]
9042 Perform an action on a session response depending on a layer 4-7 condition
9043 May be used in sections : defaults | frontend | listen | backend
9044 no | no | yes | yes
9045 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02009046 <action> defines the action to perform if the condition applies. See
9047 below.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009048
9049 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
9050
9051 Response contents can be analysed at an early stage of response processing
9052 called "TCP content inspection". During this stage, ACL-based rules are
9053 evaluated every time the response contents are updated, until either an
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02009054 "accept", "close" or a "reject" rule matches, or a TCP response inspection
9055 delay is set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009056
9057 Most often, these decisions will consider a protocol recognition or validity.
9058
9059 Content-based rules are evaluated in their exact declaration order. If no
9060 rule matches or if there is no rule, the default action is to accept the
9061 contents. There is no specific limit to the number of rules which may be
9062 inserted.
9063
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009064 Several types of actions are supported :
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009065 - accept :
9066 accepts the response if the condition is true (when used with "if")
9067 or false (when used with "unless"). The first such rule executed ends
9068 the rules evaluation.
9069
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02009070 - close :
9071 immediately closes the connection with the server if the condition is
9072 true (when used with "if"), or false (when used with "unless"). The
9073 first such rule executed ends the rules evaluation. The main purpose of
9074 this action is to force a connection to be finished between a client
9075 and a server after an exchange when the application protocol expects
9076 some long time outs to elapse first. The goal is to eliminate idle
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009077 connections which take significant resources on servers with certain
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02009078 protocols.
9079
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009080 - reject :
9081 rejects the response if the condition is true (when used with "if")
9082 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009083 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009084
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009085 - set-var(<var-name>) <expr>
9086 Sets a variable.
9087
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02009088 - sc-inc-gpc0(<sc-id>):
9089 This action increments the GPC0 counter according to the sticky
9090 counter designated by <sc-id>. If an error occurs, this action fails
9091 silently and the actions evaluation continues.
9092
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009093 - sc-set-gpt0(<sc-id>) <int> :
9094 This action sets the GPT0 tag according to the sticky counter designated
9095 by <sc-id> and the value of <int>. The expected result is a boolean. If
9096 an error occurs, this action silently fails and the actions evaluation
9097 continues.
9098
Willy Tarreau2d392c22015-08-24 01:43:45 +02009099 - "silent-drop" :
9100 This stops the evaluation of the rules and makes the client-facing
9101 connection suddenly disappear using a system-dependant way that tries
9102 to prevent the client from being notified. The effect it then that the
9103 client still sees an established connection while there's none on
9104 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
9105 except that it doesn't use any local resource at all on the machine
9106 running HAProxy. It can resist much higher loads than "tarpit", and
9107 slow down stronger attackers. It is important to undestand the impact
9108 of using this mechanism. All stateful equipments placed between the
9109 client and HAProxy (firewalls, proxies, load balancers) will also keep
9110 the established connection for a long time and may suffer from this
9111 action. On modern Linux systems running with enough privileges, the
9112 TCP_REPAIR socket option is used to block the emission of a TCP
9113 reset. On other systems, the socket's TTL is reduced to 1 so that the
9114 TCP reset doesn't pass the first router, though it's still delivered to
9115 local networks. Do not use it unless you fully understand how it works.
9116
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009117 Note that the "if/unless" condition is optional. If no condition is set on
9118 the action, it is simply performed unconditionally. That can be useful for
9119 for changing the default action to a reject.
9120
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009121 It is perfectly possible to match layer 7 contents with "tcp-response
9122 content" rules, but then it is important to ensure that a full response has
9123 been buffered, otherwise no contents will match. In order to achieve this,
9124 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009125 period.
9126
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009127 The "set-var" is used to set the content of a variable. The variable is
9128 declared inline.
9129
Daniel Schneller0b547052016-03-21 20:46:57 +01009130 <var-name> The name of the variable starts with an indication about
9131 its scope. The scopes allowed are:
9132 "sess" : the variable is shared with the whole session
9133 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009134 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +01009135 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009136 processing
Daniel Schneller0b547052016-03-21 20:46:57 +01009137 "res" : the variable is shared only during response
9138 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009139 This prefix is followed by a name. The separator is a '.'.
Daniel Schneller0b547052016-03-21 20:46:57 +01009140 The name may only contain characters 'a-z', 'A-Z', '0-9'
9141 and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009142
9143 <expr> Is a standard HAProxy expression formed by a sample-fetch
9144 followed by some converters.
9145
9146 Example:
9147
9148 tcp-request content set-var(sess.my_var) src
9149
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009150 See section 7 about ACL usage.
9151
9152 See also : "tcp-request content", "tcp-response inspect-delay"
9153
9154
9155tcp-response inspect-delay <timeout>
9156 Set the maximum allowed time to wait for a response during content inspection
9157 May be used in sections : defaults | frontend | listen | backend
9158 no | no | yes | yes
9159 Arguments :
9160 <timeout> is the timeout value specified in milliseconds by default, but
9161 can be in any other unit if the number is suffixed by the unit,
9162 as explained at the top of this document.
9163
9164 See also : "tcp-response content", "tcp-request inspect-delay".
9165
9166
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009167timeout check <timeout>
9168 Set additional check timeout, but only after a connection has been already
9169 established.
9170
9171 May be used in sections: defaults | frontend | listen | backend
9172 yes | no | yes | yes
9173 Arguments:
9174 <timeout> is the timeout value specified in milliseconds by default, but
9175 can be in any other unit if the number is suffixed by the unit,
9176 as explained at the top of this document.
9177
9178 If set, haproxy uses min("timeout connect", "inter") as a connect timeout
9179 for check and "timeout check" as an additional read timeout. The "min" is
9180 used so that people running with *very* long "timeout connect" (eg. those
9181 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +01009182 (Please also note that there is no valid reason to have such long connect
9183 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
9184 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009185
9186 If "timeout check" is not set haproxy uses "inter" for complete check
9187 timeout (connect + read) exactly like all <1.3.15 version.
9188
9189 In most cases check request is much simpler and faster to handle than normal
9190 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +01009191 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009192
9193 This parameter is specific to backends, but can be specified once for all in
9194 "defaults" sections. This is in fact one of the easiest solutions not to
9195 forget about it.
9196
Willy Tarreau41a340d2008-01-22 12:25:31 +01009197 See also: "timeout connect", "timeout queue", "timeout server",
9198 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009199
9200
Willy Tarreau0ba27502007-12-24 16:55:16 +01009201timeout client <timeout>
9202timeout clitimeout <timeout> (deprecated)
9203 Set the maximum inactivity time on the client side.
9204 May be used in sections : defaults | frontend | listen | backend
9205 yes | yes | yes | no
9206 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01009207 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01009208 can be in any other unit if the number is suffixed by the unit,
9209 as explained at the top of this document.
9210
9211 The inactivity timeout applies when the client is expected to acknowledge or
9212 send data. In HTTP mode, this timeout is particularly important to consider
9213 during the first phase, when the client sends the request, and during the
Baptiste Assmann2e1941e2016-03-06 23:24:12 +01009214 response while it is reading data sent by the server. That said, for the
9215 first phase, it is preferable to set the "timeout http-request" to better
9216 protect HAProxy from Slowloris like attacks. The value is specified in
9217 milliseconds by default, but can be in any other unit if the number is
Willy Tarreau0ba27502007-12-24 16:55:16 +01009218 suffixed by the unit, as specified at the top of this document. In TCP mode
9219 (and to a lesser extent, in HTTP mode), it is highly recommended that the
9220 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01009221 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01009222 losses by specifying timeouts that are slightly above multiples of 3 seconds
Willy Tarreauce887fd2012-05-12 12:50:00 +02009223 (eg: 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
9224 sessions (eg: WebSocket and HTTP), it's worth considering "timeout tunnel",
Willy Tarreau05cdd962014-05-10 14:30:07 +02009225 which overrides "timeout client" and "timeout server" for tunnels, as well as
9226 "timeout client-fin" for half-closed connections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01009227
9228 This parameter is specific to frontends, but can be specified once for all in
9229 "defaults" sections. This is in fact one of the easiest solutions not to
9230 forget about it. An unspecified timeout results in an infinite timeout, which
9231 is not recommended. Such a usage is accepted and works but reports a warning
9232 during startup because it may results in accumulation of expired sessions in
9233 the system if the system's timeouts are not configured either.
9234
9235 This parameter replaces the old, deprecated "clitimeout". It is recommended
9236 to use it to write new configurations. The form "timeout clitimeout" is
9237 provided only by backwards compatibility but its use is strongly discouraged.
9238
Baptiste Assmann2e1941e2016-03-06 23:24:12 +01009239 See also : "clitimeout", "timeout server", "timeout tunnel",
9240 "timeout http-request".
Willy Tarreau0ba27502007-12-24 16:55:16 +01009241
9242
Willy Tarreau05cdd962014-05-10 14:30:07 +02009243timeout client-fin <timeout>
9244 Set the inactivity timeout on the client side for half-closed connections.
9245 May be used in sections : defaults | frontend | listen | backend
9246 yes | yes | yes | no
9247 Arguments :
9248 <timeout> is the timeout value specified in milliseconds by default, but
9249 can be in any other unit if the number is suffixed by the unit,
9250 as explained at the top of this document.
9251
9252 The inactivity timeout applies when the client is expected to acknowledge or
9253 send data while one direction is already shut down. This timeout is different
9254 from "timeout client" in that it only applies to connections which are closed
9255 in one direction. This is particularly useful to avoid keeping connections in
9256 FIN_WAIT state for too long when clients do not disconnect cleanly. This
9257 problem is particularly common long connections such as RDP or WebSocket.
9258 Note that this timeout can override "timeout tunnel" when a connection shuts
9259 down in one direction.
9260
9261 This parameter is specific to frontends, but can be specified once for all in
9262 "defaults" sections. By default it is not set, so half-closed connections
9263 will use the other timeouts (timeout.client or timeout.tunnel).
9264
9265 See also : "timeout client", "timeout server-fin", and "timeout tunnel".
9266
9267
Willy Tarreau0ba27502007-12-24 16:55:16 +01009268timeout connect <timeout>
9269timeout contimeout <timeout> (deprecated)
9270 Set the maximum time to wait for a connection attempt to a server to succeed.
9271 May be used in sections : defaults | frontend | listen | backend
9272 yes | no | yes | yes
9273 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01009274 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01009275 can be in any other unit if the number is suffixed by the unit,
9276 as explained at the top of this document.
9277
9278 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01009279 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01009280 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01009281 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009282 connect timeout also presets both queue and tarpit timeouts to the same value
9283 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +01009284
9285 This parameter is specific to backends, but can be specified once for all in
9286 "defaults" sections. This is in fact one of the easiest solutions not to
9287 forget about it. An unspecified timeout results in an infinite timeout, which
9288 is not recommended. Such a usage is accepted and works but reports a warning
9289 during startup because it may results in accumulation of failed sessions in
9290 the system if the system's timeouts are not configured either.
9291
9292 This parameter replaces the old, deprecated "contimeout". It is recommended
9293 to use it to write new configurations. The form "timeout contimeout" is
9294 provided only by backwards compatibility but its use is strongly discouraged.
9295
Willy Tarreau41a340d2008-01-22 12:25:31 +01009296 See also: "timeout check", "timeout queue", "timeout server", "contimeout",
9297 "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +01009298
9299
Willy Tarreaub16a5742010-01-10 14:46:16 +01009300timeout http-keep-alive <timeout>
9301 Set the maximum allowed time to wait for a new HTTP request to appear
9302 May be used in sections : defaults | frontend | listen | backend
9303 yes | yes | yes | yes
9304 Arguments :
9305 <timeout> is the timeout value specified in milliseconds by default, but
9306 can be in any other unit if the number is suffixed by the unit,
9307 as explained at the top of this document.
9308
9309 By default, the time to wait for a new request in case of keep-alive is set
9310 by "timeout http-request". However this is not always convenient because some
9311 people want very short keep-alive timeouts in order to release connections
9312 faster, and others prefer to have larger ones but still have short timeouts
9313 once the request has started to present itself.
9314
9315 The "http-keep-alive" timeout covers these needs. It will define how long to
9316 wait for a new HTTP request to start coming after a response was sent. Once
9317 the first byte of request has been seen, the "http-request" timeout is used
9318 to wait for the complete request to come. Note that empty lines prior to a
9319 new request do not refresh the timeout and are not counted as a new request.
9320
9321 There is also another difference between the two timeouts : when a connection
9322 expires during timeout http-keep-alive, no error is returned, the connection
9323 just closes. If the connection expires in "http-request" while waiting for a
9324 connection to complete, a HTTP 408 error is returned.
9325
9326 In general it is optimal to set this value to a few tens to hundreds of
9327 milliseconds, to allow users to fetch all objects of a page at once but
9328 without waiting for further clicks. Also, if set to a very small value (eg:
9329 1 millisecond) it will probably only accept pipelined requests but not the
9330 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +02009331 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +01009332
9333 If this parameter is not set, the "http-request" timeout applies, and if both
9334 are not set, "timeout client" still applies at the lower level. It should be
9335 set in the frontend to take effect, unless the frontend is in TCP mode, in
9336 which case the HTTP backend's timeout will be used.
9337
9338 See also : "timeout http-request", "timeout client".
9339
9340
Willy Tarreau036fae02008-01-06 13:24:40 +01009341timeout http-request <timeout>
9342 Set the maximum allowed time to wait for a complete HTTP request
9343 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +02009344 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +01009345 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01009346 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +01009347 can be in any other unit if the number is suffixed by the unit,
9348 as explained at the top of this document.
9349
9350 In order to offer DoS protection, it may be required to lower the maximum
9351 accepted time to receive a complete HTTP request without affecting the client
9352 timeout. This helps protecting against established connections on which
9353 nothing is sent. The client timeout cannot offer a good protection against
9354 this abuse because it is an inactivity timeout, which means that if the
9355 attacker sends one character every now and then, the timeout will not
9356 trigger. With the HTTP request timeout, no matter what speed the client
Willy Tarreau2705a612014-05-23 17:38:34 +02009357 types, the request will be aborted if it does not complete in time. When the
9358 timeout expires, an HTTP 408 response is sent to the client to inform it
9359 about the problem, and the connection is closed. The logs will report
9360 termination codes "cR". Some recent browsers are having problems with this
9361 standard, well-documented behaviour, so it might be needed to hide the 408
Willy Tarreau0f228a02015-05-01 15:37:53 +02009362 code using "option http-ignore-probes" or "errorfile 408 /dev/null". See
9363 more details in the explanations of the "cR" termination code in section 8.5.
Willy Tarreau036fae02008-01-06 13:24:40 +01009364
Baptiste Assmanneccdf432015-10-28 13:49:01 +01009365 By default, this timeout only applies to the header part of the request,
9366 and not to any data. As soon as the empty line is received, this timeout is
9367 not used anymore. When combined with "option http-buffer-request", this
9368 timeout also applies to the body of the request..
9369 It is used again on keep-alive connections to wait for a second
Willy Tarreaub16a5742010-01-10 14:46:16 +01009370 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +01009371
9372 Generally it is enough to set it to a few seconds, as most clients send the
9373 full request immediately upon connection. Add 3 or more seconds to cover TCP
9374 retransmits but that's all. Setting it to very low values (eg: 50 ms) will
9375 generally work on local networks as long as there are no packet losses. This
9376 will prevent people from sending bare HTTP requests using telnet.
9377
9378 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +02009379 chunk of the incoming request. It should be set in the frontend to take
9380 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
9381 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +01009382
Willy Tarreau0f228a02015-05-01 15:37:53 +02009383 See also : "errorfile", "http-ignore-probes", "timeout http-keep-alive", and
Baptiste Assmanneccdf432015-10-28 13:49:01 +01009384 "timeout client", "option http-buffer-request".
Willy Tarreau036fae02008-01-06 13:24:40 +01009385
Willy Tarreau844e3c52008-01-11 16:28:18 +01009386
9387timeout queue <timeout>
9388 Set the maximum time to wait in the queue for a connection slot to be free
9389 May be used in sections : defaults | frontend | listen | backend
9390 yes | no | yes | yes
9391 Arguments :
9392 <timeout> is the timeout value specified in milliseconds by default, but
9393 can be in any other unit if the number is suffixed by the unit,
9394 as explained at the top of this document.
9395
9396 When a server's maxconn is reached, connections are left pending in a queue
9397 which may be server-specific or global to the backend. In order not to wait
9398 indefinitely, a timeout is applied to requests pending in the queue. If the
9399 timeout is reached, it is considered that the request will almost never be
9400 served, so it is dropped and a 503 error is returned to the client.
9401
9402 The "timeout queue" statement allows to fix the maximum time for a request to
9403 be left pending in a queue. If unspecified, the same value as the backend's
9404 connection timeout ("timeout connect") is used, for backwards compatibility
9405 with older versions with no "timeout queue" parameter.
9406
9407 See also : "timeout connect", "contimeout".
9408
9409
9410timeout server <timeout>
9411timeout srvtimeout <timeout> (deprecated)
9412 Set the maximum inactivity time on the server side.
9413 May be used in sections : defaults | frontend | listen | backend
9414 yes | no | yes | yes
9415 Arguments :
9416 <timeout> is the timeout value specified in milliseconds by default, but
9417 can be in any other unit if the number is suffixed by the unit,
9418 as explained at the top of this document.
9419
9420 The inactivity timeout applies when the server is expected to acknowledge or
9421 send data. In HTTP mode, this timeout is particularly important to consider
9422 during the first phase of the server's response, when it has to send the
9423 headers, as it directly represents the server's processing time for the
9424 request. To find out what value to put there, it's often good to start with
9425 what would be considered as unacceptable response times, then check the logs
9426 to observe the response time distribution, and adjust the value accordingly.
9427
9428 The value is specified in milliseconds by default, but can be in any other
9429 unit if the number is suffixed by the unit, as specified at the top of this
9430 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
9431 recommended that the client timeout remains equal to the server timeout in
9432 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01009433 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01009434 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreauce887fd2012-05-12 12:50:00 +02009435 seconds (eg: 4 or 5 seconds minimum). If some long-lived sessions are mixed
9436 with short-lived sessions (eg: WebSocket and HTTP), it's worth considering
9437 "timeout tunnel", which overrides "timeout client" and "timeout server" for
9438 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009439
9440 This parameter is specific to backends, but can be specified once for all in
9441 "defaults" sections. This is in fact one of the easiest solutions not to
9442 forget about it. An unspecified timeout results in an infinite timeout, which
9443 is not recommended. Such a usage is accepted and works but reports a warning
9444 during startup because it may results in accumulation of expired sessions in
9445 the system if the system's timeouts are not configured either.
9446
9447 This parameter replaces the old, deprecated "srvtimeout". It is recommended
9448 to use it to write new configurations. The form "timeout srvtimeout" is
9449 provided only by backwards compatibility but its use is strongly discouraged.
9450
Willy Tarreauce887fd2012-05-12 12:50:00 +02009451 See also : "srvtimeout", "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +01009452
Willy Tarreau05cdd962014-05-10 14:30:07 +02009453
9454timeout server-fin <timeout>
9455 Set the inactivity timeout on the server side for half-closed connections.
9456 May be used in sections : defaults | frontend | listen | backend
9457 yes | no | yes | yes
9458 Arguments :
9459 <timeout> is the timeout value specified in milliseconds by default, but
9460 can be in any other unit if the number is suffixed by the unit,
9461 as explained at the top of this document.
9462
9463 The inactivity timeout applies when the server is expected to acknowledge or
9464 send data while one direction is already shut down. This timeout is different
9465 from "timeout server" in that it only applies to connections which are closed
9466 in one direction. This is particularly useful to avoid keeping connections in
9467 FIN_WAIT state for too long when a remote server does not disconnect cleanly.
9468 This problem is particularly common long connections such as RDP or WebSocket.
9469 Note that this timeout can override "timeout tunnel" when a connection shuts
9470 down in one direction. This setting was provided for completeness, but in most
9471 situations, it should not be needed.
9472
9473 This parameter is specific to backends, but can be specified once for all in
9474 "defaults" sections. By default it is not set, so half-closed connections
9475 will use the other timeouts (timeout.server or timeout.tunnel).
9476
9477 See also : "timeout client-fin", "timeout server", and "timeout tunnel".
9478
Willy Tarreau844e3c52008-01-11 16:28:18 +01009479
9480timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +01009481 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +01009482 May be used in sections : defaults | frontend | listen | backend
9483 yes | yes | yes | yes
9484 Arguments :
9485 <timeout> is the tarpit duration specified in milliseconds by default, but
9486 can be in any other unit if the number is suffixed by the unit,
9487 as explained at the top of this document.
9488
9489 When a connection is tarpitted using "reqtarpit", it is maintained open with
9490 no activity for a certain amount of time, then closed. "timeout tarpit"
9491 defines how long it will be maintained open.
9492
9493 The value is specified in milliseconds by default, but can be in any other
9494 unit if the number is suffixed by the unit, as specified at the top of this
9495 document. If unspecified, the same value as the backend's connection timeout
9496 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +01009497 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009498
9499 See also : "timeout connect", "contimeout".
9500
9501
Willy Tarreauce887fd2012-05-12 12:50:00 +02009502timeout tunnel <timeout>
9503 Set the maximum inactivity time on the client and server side for tunnels.
9504 May be used in sections : defaults | frontend | listen | backend
9505 yes | no | yes | yes
9506 Arguments :
9507 <timeout> is the timeout value specified in milliseconds by default, but
9508 can be in any other unit if the number is suffixed by the unit,
9509 as explained at the top of this document.
9510
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009511 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +02009512 between a client and a server, and the connection remains inactive in both
9513 directions. This timeout supersedes both the client and server timeouts once
9514 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
9515 analyser remains attached to either connection (eg: tcp content rules are
9516 accepted). In HTTP, this timeout is used when a connection is upgraded (eg:
9517 when switching to the WebSocket protocol, or forwarding a CONNECT request
9518 to a proxy), or after the first response when no keepalive/close option is
9519 specified.
9520
Willy Tarreau05cdd962014-05-10 14:30:07 +02009521 Since this timeout is usually used in conjunction with long-lived connections,
9522 it usually is a good idea to also set "timeout client-fin" to handle the
9523 situation where a client suddenly disappears from the net and does not
9524 acknowledge a close, or sends a shutdown and does not acknowledge pending
9525 data anymore. This can happen in lossy networks where firewalls are present,
9526 and is detected by the presence of large amounts of sessions in a FIN_WAIT
9527 state.
9528
Willy Tarreauce887fd2012-05-12 12:50:00 +02009529 The value is specified in milliseconds by default, but can be in any other
9530 unit if the number is suffixed by the unit, as specified at the top of this
9531 document. Whatever the expected normal idle time, it is a good practice to
9532 cover at least one or several TCP packet losses by specifying timeouts that
9533 are slightly above multiples of 3 seconds (eg: 4 or 5 seconds minimum).
9534
9535 This parameter is specific to backends, but can be specified once for all in
9536 "defaults" sections. This is in fact one of the easiest solutions not to
9537 forget about it.
9538
9539 Example :
9540 defaults http
9541 option http-server-close
9542 timeout connect 5s
9543 timeout client 30s
Willy Tarreau05cdd962014-05-10 14:30:07 +02009544 timeout client-fin 30s
Willy Tarreauce887fd2012-05-12 12:50:00 +02009545 timeout server 30s
9546 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
9547
Willy Tarreau05cdd962014-05-10 14:30:07 +02009548 See also : "timeout client", "timeout client-fin", "timeout server".
Willy Tarreauce887fd2012-05-12 12:50:00 +02009549
9550
Willy Tarreau844e3c52008-01-11 16:28:18 +01009551transparent (deprecated)
9552 Enable client-side transparent proxying
9553 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01009554 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01009555 Arguments : none
9556
9557 This keyword was introduced in order to provide layer 7 persistence to layer
9558 3 load balancers. The idea is to use the OS's ability to redirect an incoming
9559 connection for a remote address to a local process (here HAProxy), and let
9560 this process know what address was initially requested. When this option is
9561 used, sessions without cookies will be forwarded to the original destination
9562 IP address of the incoming request (which should match that of another
9563 equipment), while requests with cookies will still be forwarded to the
9564 appropriate server.
9565
9566 The "transparent" keyword is deprecated, use "option transparent" instead.
9567
9568 Note that contrary to a common belief, this option does NOT make HAProxy
9569 present the client's IP to the server when establishing the connection.
9570
Willy Tarreau844e3c52008-01-11 16:28:18 +01009571 See also: "option transparent"
9572
William Lallemanda73203e2012-03-12 12:48:57 +01009573unique-id-format <string>
9574 Generate a unique ID for each request.
9575 May be used in sections : defaults | frontend | listen | backend
9576 yes | yes | yes | no
9577 Arguments :
9578 <string> is a log-format string.
9579
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009580 This keyword creates a ID for each request using the custom log format. A
9581 unique ID is useful to trace a request passing through many components of
9582 a complex infrastructure. The newly created ID may also be logged using the
9583 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +01009584
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009585 The format should be composed from elements that are guaranteed to be
9586 unique when combined together. For instance, if multiple haproxy instances
9587 are involved, it might be important to include the node name. It is often
9588 needed to log the incoming connection's source and destination addresses
9589 and ports. Note that since multiple requests may be performed over the same
9590 connection, including a request counter may help differentiate them.
9591 Similarly, a timestamp may protect against a rollover of the counter.
9592 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +01009593
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009594 It is recommended to use hexadecimal notation for many fields since it
9595 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +01009596
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009597 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01009598
Julien Vehentf21be322014-03-07 08:27:34 -05009599 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +01009600
9601 will generate:
9602
9603 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
9604
9605 See also: "unique-id-header"
9606
9607unique-id-header <name>
9608 Add a unique ID header in the HTTP request.
9609 May be used in sections : defaults | frontend | listen | backend
9610 yes | yes | yes | no
9611 Arguments :
9612 <name> is the name of the header.
9613
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009614 Add a unique-id header in the HTTP request sent to the server, using the
9615 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +01009616
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009617 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01009618
Julien Vehentf21be322014-03-07 08:27:34 -05009619 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +01009620 unique-id-header X-Unique-ID
9621
9622 will generate:
9623
9624 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
9625
9626 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +01009627
Willy Tarreauf51658d2014-04-23 01:21:56 +02009628use_backend <backend> [{if | unless} <condition>]
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02009629 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009630 May be used in sections : defaults | frontend | listen | backend
9631 no | yes | yes | no
9632 Arguments :
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01009633 <backend> is the name of a valid backend or "listen" section, or a
9634 "log-format" string resolving to a backend name.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009635
Willy Tarreauf51658d2014-04-23 01:21:56 +02009636 <condition> is a condition composed of ACLs, as described in section 7. If
9637 it is omitted, the rule is unconditionally applied.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009638
9639 When doing content-switching, connections arrive on a frontend and are then
9640 dispatched to various backends depending on a number of conditions. The
9641 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02009642 "use_backend" keyword. While it is normally used with HTTP processing, it can
9643 also be used in pure TCP, either without content using stateless ACLs (eg:
9644 source address validation) or combined with a "tcp-request" rule to wait for
9645 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009646
9647 There may be as many "use_backend" rules as desired. All of these rules are
9648 evaluated in their declaration order, and the first one which matches will
9649 assign the backend.
9650
9651 In the first form, the backend will be used if the condition is met. In the
9652 second form, the backend will be used if the condition is not met. If no
9653 condition is valid, the backend defined with "default_backend" will be used.
9654 If no default backend is defined, either the servers in the same section are
9655 used (in case of a "listen" section) or, in case of a frontend, no server is
9656 used and a 503 service unavailable response is returned.
9657
Willy Tarreau51aecc72009-07-12 09:47:04 +02009658 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009659 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +02009660 and backend processing will immediately follow, or the backend will wait for
9661 a complete HTTP request to get in. This feature is useful when a frontend
9662 must decode several protocols on a unique port, one of them being HTTP.
9663
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01009664 When <backend> is a simple name, it is resolved at configuration time, and an
9665 error is reported if the specified backend does not exist. If <backend> is
9666 a log-format string instead, no check may be done at configuration time, so
9667 the backend name is resolved dynamically at run time. If the resulting
9668 backend name does not correspond to any valid backend, no other rule is
9669 evaluated, and the default_backend directive is applied instead. Note that
9670 when using dynamic backend names, it is highly recommended to use a prefix
9671 that no other backend uses in order to ensure that an unauthorized backend
9672 cannot be forced from the request.
9673
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009674 It is worth mentioning that "use_backend" rules with an explicit name are
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01009675 used to detect the association between frontends and backends to compute the
9676 backend's "fullconn" setting. This cannot be done for dynamic names.
9677
9678 See also: "default_backend", "tcp-request", "fullconn", "log-format", and
9679 section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +01009680
Willy Tarreau036fae02008-01-06 13:24:40 +01009681
Willy Tarreau4a5cade2012-04-05 21:09:48 +02009682use-server <server> if <condition>
9683use-server <server> unless <condition>
9684 Only use a specific server if/unless an ACL-based condition is matched.
9685 May be used in sections : defaults | frontend | listen | backend
9686 no | no | yes | yes
9687 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009688 <server> is the name of a valid server in the same backend section.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02009689
9690 <condition> is a condition composed of ACLs, as described in section 7.
9691
9692 By default, connections which arrive to a backend are load-balanced across
9693 the available servers according to the configured algorithm, unless a
9694 persistence mechanism such as a cookie is used and found in the request.
9695
9696 Sometimes it is desirable to forward a particular request to a specific
9697 server without having to declare a dedicated backend for this server. This
9698 can be achieved using the "use-server" rules. These rules are evaluated after
9699 the "redirect" rules and before evaluating cookies, and they have precedence
9700 on them. There may be as many "use-server" rules as desired. All of these
9701 rules are evaluated in their declaration order, and the first one which
9702 matches will assign the server.
9703
9704 If a rule designates a server which is down, and "option persist" is not used
9705 and no force-persist rule was validated, it is ignored and evaluation goes on
9706 with the next rules until one matches.
9707
9708 In the first form, the server will be used if the condition is met. In the
9709 second form, the server will be used if the condition is not met. If no
9710 condition is valid, the processing continues and the server will be assigned
9711 according to other persistence mechanisms.
9712
9713 Note that even if a rule is matched, cookie processing is still performed but
9714 does not assign the server. This allows prefixed cookies to have their prefix
9715 stripped.
9716
9717 The "use-server" statement works both in HTTP and TCP mode. This makes it
9718 suitable for use with content-based inspection. For instance, a server could
9719 be selected in a farm according to the TLS SNI field. And if these servers
9720 have their weight set to zero, they will not be used for other traffic.
9721
9722 Example :
9723 # intercept incoming TLS requests based on the SNI field
9724 use-server www if { req_ssl_sni -i www.example.com }
9725 server www 192.168.0.1:443 weight 0
9726 use-server mail if { req_ssl_sni -i mail.example.com }
9727 server mail 192.168.0.1:587 weight 0
9728 use-server imap if { req_ssl_sni -i imap.example.com }
9729 server mail 192.168.0.1:993 weight 0
9730 # all the rest is forwarded to this server
9731 server default 192.168.0.2:443 check
9732
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009733 See also: "use_backend", section 5 about server and section 7 about ACLs.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02009734
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009735
97365. Bind and Server options
9737--------------------------
9738
9739The "bind", "server" and "default-server" keywords support a number of settings
9740depending on some build options and on the system HAProxy was built on. These
9741settings generally each consist in one word sometimes followed by a value,
9742written on the same line as the "bind" or "server" line. All these options are
9743described in this section.
9744
9745
97465.1. Bind options
9747-----------------
9748
9749The "bind" keyword supports a certain number of settings which are all passed
9750as arguments on the same line. The order in which those arguments appear makes
9751no importance, provided that they appear after the bind address. All of these
9752parameters are optional. Some of them consist in a single words (booleans),
9753while other ones expect a value after them. In this case, the value must be
9754provided immediately after the setting name.
9755
9756The currently supported settings are the following ones.
9757
Bertrand Jacquin93b227d2016-06-04 15:11:10 +01009758accept-netscaler-cip <magic number>
9759 Enforces the use of the NetScaler Client IP insertion protocol over any
9760 connection accepted by any of the TCP sockets declared on the same line. The
9761 NetScaler Client IP insertion protocol dictates the layer 3/4 addresses of
9762 the incoming connection to be used everywhere an address is used, with the
9763 only exception of "tcp-request connection" rules which will only see the
9764 real connection address. Logs will reflect the addresses indicated in the
9765 protocol, unless it is violated, in which case the real address will still
9766 be used. This keyword combined with support from external components can be
9767 used as an efficient and reliable alternative to the X-Forwarded-For
Bertrand Jacquin90759682016-06-06 15:35:39 +01009768 mechanism which is not always reliable and not even always usable. See also
9769 "tcp-request connection expect-netscaler-cip" for a finer-grained setting of
9770 which client is allowed to use the protocol.
Bertrand Jacquin93b227d2016-06-04 15:11:10 +01009771
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009772accept-proxy
9773 Enforces the use of the PROXY protocol over any connection accepted by any of
Willy Tarreau77992672014-06-14 11:06:17 +02009774 the sockets declared on the same line. Versions 1 and 2 of the PROXY protocol
9775 are supported and correctly detected. The PROXY protocol dictates the layer
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009776 3/4 addresses of the incoming connection to be used everywhere an address is
9777 used, with the only exception of "tcp-request connection" rules which will
9778 only see the real connection address. Logs will reflect the addresses
9779 indicated in the protocol, unless it is violated, in which case the real
9780 address will still be used. This keyword combined with support from external
9781 components can be used as an efficient and reliable alternative to the
9782 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +02009783 usable. See also "tcp-request connection expect-proxy" for a finer-grained
9784 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009785
Willy Tarreauab861d32013-04-02 02:30:41 +02009786alpn <protocols>
9787 This enables the TLS ALPN extension and advertises the specified protocol
9788 list as supported on top of ALPN. The protocol list consists in a comma-
9789 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
9790 quotes). This requires that the SSL library is build with support for TLS
9791 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
9792 initial NPN extension.
9793
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009794backlog <backlog>
9795 Sets the socket's backlog to this value. If unspecified, the frontend's
9796 backlog is used instead, which generally defaults to the maxconn value.
9797
Emeric Brun7fb34422012-09-28 15:26:15 +02009798ecdhe <named curve>
9799 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +01009800 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
9801 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +02009802
Emeric Brunfd33a262012-10-11 16:28:27 +02009803ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +02009804 This setting is only available when support for OpenSSL was built in. It
9805 designates a PEM file from which to load CA certificates used to verify
9806 client's certificate.
9807
Emeric Brunb6dc9342012-09-28 17:55:37 +02009808ca-ignore-err [all|<errorID>,...]
9809 This setting is only available when support for OpenSSL was built in.
9810 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
9811 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
9812 error is ignored.
9813
Christopher Faulet31af49d2015-06-09 17:29:50 +02009814ca-sign-file <cafile>
9815 This setting is only available when support for OpenSSL was built in. It
9816 designates a PEM file containing both the CA certificate and the CA private
9817 key used to create and sign server's certificates. This is a mandatory
9818 setting when the dynamic generation of certificates is enabled. See
9819 'generate-certificates' for details.
9820
9821ca-sign-passphrase <passphrase>
9822 This setting is only available when support for OpenSSL was built in. It is
9823 the CA private key passphrase. This setting is optional and used only when
9824 the dynamic generation of certificates is enabled. See
9825 'generate-certificates' for details.
9826
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009827ciphers <ciphers>
9828 This setting is only available when support for OpenSSL was built in. It sets
9829 the string describing the list of cipher algorithms ("cipher suite") that are
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009830 negotiated during the SSL/TLS handshake. The format of the string is defined
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009831 in "man 1 ciphers" from OpenSSL man pages, and can be for instance a string
9832 such as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes).
9833
Emeric Brunfd33a262012-10-11 16:28:27 +02009834crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +02009835 This setting is only available when support for OpenSSL was built in. It
9836 designates a PEM file from which to load certificate revocation list used
9837 to verify client's certificate.
9838
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009839crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +00009840 This setting is only available when support for OpenSSL was built in. It
9841 designates a PEM file containing both the required certificates and any
9842 associated private keys. This file can be built by concatenating multiple
9843 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
9844 requires an intermediate certificate, this can also be concatenated into this
9845 file.
9846
9847 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
9848 are loaded.
9849
9850 If a directory name is used instead of a PEM file, then all files found in
Cyril Bonté3180f7b2015-01-25 00:16:08 +01009851 that directory will be loaded in alphabetic order unless their name ends with
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +01009852 '.issuer', '.ocsp' or '.sctl' (reserved extensions). This directive may be
9853 specified multiple times in order to load certificates from multiple files or
9854 directories. The certificates will be presented to clients who provide a
9855 valid TLS Server Name Indication field matching one of their CN or alt
9856 subjects. Wildcards are supported, where a wildcard character '*' is used
9857 instead of the first hostname component (eg: *.example.org matches
9858 www.example.org but not www.sub.example.org).
Alex Davies0fbf0162013-03-02 16:04:50 +00009859
9860 If no SNI is provided by the client or if the SSL library does not support
9861 TLS extensions, or if the client provides an SNI hostname which does not
9862 match any certificate, then the first loaded certificate will be presented.
9863 This means that when loading certificates from a directory, it is highly
Cyril Bonté3180f7b2015-01-25 00:16:08 +01009864 recommended to load the default one first as a file or to ensure that it will
9865 always be the first one in the directory.
Alex Davies0fbf0162013-03-02 16:04:50 +00009866
Emeric Brune032bfa2012-09-28 13:01:45 +02009867 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009868
Alex Davies0fbf0162013-03-02 16:04:50 +00009869 Some CAs (such as Godaddy) offer a drop down list of server types that do not
9870 include HAProxy when obtaining a certificate. If this happens be sure to
Godbach8bf60a12014-04-21 21:42:41 +08009871 choose a webserver that the CA believes requires an intermediate CA (for
Alex Davies0fbf0162013-03-02 16:04:50 +00009872 Godaddy, selection Apache Tomcat will get the correct bundle, but many
9873 others, e.g. nginx, result in a wrong bundle that will not work for some
9874 clients).
9875
Emeric Brun4147b2e2014-06-16 18:36:30 +02009876 For each PEM file, haproxy checks for the presence of file at the same path
9877 suffixed by ".ocsp". If such file is found, support for the TLS Certificate
9878 Status Request extension (also known as "OCSP stapling") is automatically
9879 enabled. The content of this file is optional. If not empty, it must contain
9880 a valid OCSP Response in DER format. In order to be valid an OCSP Response
9881 must comply with the following rules: it has to indicate a good status,
9882 it has to be a single response for the certificate of the PEM file, and it
9883 has to be valid at the moment of addition. If these rules are not respected
9884 the OCSP Response is ignored and a warning is emitted. In order to identify
9885 which certificate an OCSP Response applies to, the issuer's certificate is
9886 necessary. If the issuer's certificate is not found in the PEM file, it will
9887 be loaded from a file at the same path as the PEM file suffixed by ".issuer"
9888 if it exists otherwise it will fail with an error.
9889
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +01009890 For each PEM file, haproxy also checks for the presence of file at the same
9891 path suffixed by ".sctl". If such file is found, support for Certificate
9892 Transparency (RFC6962) TLS extension is enabled. The file must contain a
9893 valid Signed Certificate Timestamp List, as described in RFC. File is parsed
9894 to check basic syntax, but no signatures are verified.
9895
yanbzhu6c25e9e2016-01-05 12:52:02 -05009896 There are cases where it is desirable to support multiple key types, e.g. RSA
9897 and ECDSA in the cipher suites offered to the clients. This allows clients
9898 that support EC certificates to be able to use EC ciphers, while
9899 simultaneously supporting older, RSA only clients.
yanbzhud19630c2015-12-14 15:10:25 -05009900
9901 In order to provide this functionality, multiple PEM files, each with a
9902 different key type, are required. To associate these PEM files into a
9903 "cert bundle" that is recognized by haproxy, they must be named in the
9904 following way: All PEM files that are to be bundled must have the same base
9905 name, with a suffix indicating the key type. Currently, three suffixes are
9906 supported: rsa, dsa and ecdsa. For example, if www.example.com has two PEM
9907 files, an RSA file and an ECDSA file, they must be named: "example.pem.rsa"
9908 and "example.pem.ecdsa". The first part of the filename is arbitrary; only the
9909 suffix matters. To load this bundle into haproxy, specify the base name only:
9910
9911 Example : bind :8443 ssl crt example.pem
9912
yanbzhu6c25e9e2016-01-05 12:52:02 -05009913 Note that the suffix is not given to haproxy; this tells haproxy to look for
yanbzhud19630c2015-12-14 15:10:25 -05009914 a cert bundle.
9915
9916 Haproxy will load all PEM files in the bundle at the same time to try to
9917 support multiple key types. PEM files are combined based on Common Name
9918 (CN) and Subject Alternative Name (SAN) to support SNI lookups. This means
9919 that even if you give haproxy a cert bundle, if there are no shared CN/SAN
9920 entries in the certificates in that bundle, haproxy will not be able to
9921 provide multi-cert support.
9922
9923 Assuming bundle in the example above contained the following:
9924
9925 Filename | CN | SAN
9926 -------------------+-----------------+-------------------
9927 example.pem.rsa | www.example.com | rsa.example.com
yanbzhu6c25e9e2016-01-05 12:52:02 -05009928 -------------------+-----------------+-------------------
yanbzhud19630c2015-12-14 15:10:25 -05009929 example.pem.ecdsa | www.example.com | ecdsa.example.com
9930 -------------------+-----------------+-------------------
9931
9932 Users connecting with an SNI of "www.example.com" will be able
9933 to use both RSA and ECDSA cipher suites. Users connecting with an SNI of
9934 "rsa.example.com" will only be able to use RSA cipher suites, and users
9935 connecting with "ecdsa.example.com" will only be able to use ECDSA cipher
9936 suites.
9937
9938 If a directory name is given as the <cert> argument, haproxy will
9939 automatically search and load bundled files in that directory.
9940
9941 OSCP files (.ocsp) and issuer files (.issuer) are supported with multi-cert
9942 bundling. Each certificate can have its own .ocsp and .issuer file. At this
9943 time, sctl is not supported in multi-certificate bundling.
9944
Emeric Brunb6dc9342012-09-28 17:55:37 +02009945crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +00009946 This setting is only available when support for OpenSSL was built in. Sets a
9947 comma separated list of errorIDs to ignore during verify at depth == 0. If
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009948 set to 'all', all errors are ignored. SSL handshake is not aborted if an error
Alex Davies0fbf0162013-03-02 16:04:50 +00009949 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +02009950
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009951crt-list <file>
9952 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02009953 designates a list of PEM file with an optional list of SNI filter per
9954 certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009955
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02009956 <crtfile> [[!]<snifilter> ...]
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009957
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02009958 Wildcards are supported in the SNI filter. Negative filter are also supported,
9959 only useful in combination with a wildcard filter to exclude a particular SNI.
9960 The certificates will be presented to clients who provide a valid TLS Server
9961 Name Indication field matching one of the SNI filters. If no SNI filter is
9962 specified, the CN and alt subjects are used. This directive may be specified
9963 multiple times. See the "crt" option for more information. The default
9964 certificate is still needed to meet OpenSSL expectations. If it is not used,
9965 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009966
yanbzhu6c25e9e2016-01-05 12:52:02 -05009967 Multi-cert bundling (see "crt") is supported with crt-list, as long as only
Emmanuel Hocdetd294aea2016-05-13 11:14:06 +02009968 the base name is given in the crt-list. SNI filter will do the same work on
9969 all bundled certificates.
yanbzhud19630c2015-12-14 15:10:25 -05009970
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009971defer-accept
9972 Is an optional keyword which is supported only on certain Linux kernels. It
9973 states that a connection will only be accepted once some data arrive on it,
9974 or at worst after the first retransmit. This should be used only on protocols
9975 for which the client talks first (eg: HTTP). It can slightly improve
9976 performance by ensuring that most of the request is already available when
9977 the connection is accepted. On the other hand, it will not be able to detect
9978 connections which don't talk. It is important to note that this option is
9979 broken in all kernels up to 2.6.31, as the connection is never accepted until
9980 the client talks. This can cause issues with front firewalls which would see
9981 an established connection while the proxy will only see it in SYN_RECV. This
9982 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
9983
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009984force-sslv3
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009985 This option enforces use of SSLv3 only on SSL connections instantiated from
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009986 this listener. SSLv3 is generally less expensive than the TLS counterparts
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009987 for high connection rates. This option is also available on global statement
9988 "ssl-default-bind-options". See also "no-tlsv*" and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009989
9990force-tlsv10
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009991 This option enforces use of TLSv1.0 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009992 this listener. This option is also available on global statement
9993 "ssl-default-bind-options". See also "no-tlsv*" and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009994
9995force-tlsv11
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009996 This option enforces use of TLSv1.1 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009997 this listener. This option is also available on global statement
9998 "ssl-default-bind-options". See also "no-tlsv*", and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009999
10000force-tlsv12
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010001 This option enforces use of TLSv1.2 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010002 this listener. This option is also available on global statement
10003 "ssl-default-bind-options". See also "no-tlsv*", and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010004
Christopher Faulet31af49d2015-06-09 17:29:50 +020010005generate-certificates
10006 This setting is only available when support for OpenSSL was built in. It
10007 enables the dynamic SSL certificates generation. A CA certificate and its
10008 private key are necessary (see 'ca-sign-file'). When HAProxy is configured as
10009 a transparent forward proxy, SSL requests generate errors because of a common
10010 name mismatch on the certificate presented to the client. With this option
10011 enabled, HAProxy will try to forge a certificate using the SNI hostname
10012 indicated by the client. This is done only if no certificate matches the SNI
10013 hostname (see 'crt-list'). If an error occurs, the default certificate is
10014 used, else the 'strict-sni' option is set.
10015 It can also be used when HAProxy is configured as a reverse proxy to ease the
10016 deployment of an architecture with many backends.
10017
10018 Creating a SSL certificate is an expensive operation, so a LRU cache is used
10019 to store forged certificates (see 'tune.ssl.ssl-ctx-cache-size'). It
10020 increases the HAProxy's memroy footprint to reduce latency when the same
10021 certificate is used many times.
10022
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010023gid <gid>
10024 Sets the group of the UNIX sockets to the designated system gid. It can also
10025 be set by default in the global section's "unix-bind" statement. Note that
10026 some platforms simply ignore this. This setting is equivalent to the "group"
10027 setting except that the group ID is used instead of its name. This setting is
10028 ignored by non UNIX sockets.
10029
10030group <group>
10031 Sets the group of the UNIX sockets to the designated system group. It can
10032 also be set by default in the global section's "unix-bind" statement. Note
10033 that some platforms simply ignore this. This setting is equivalent to the
10034 "gid" setting except that the group name is used instead of its gid. This
10035 setting is ignored by non UNIX sockets.
10036
10037id <id>
10038 Fixes the socket ID. By default, socket IDs are automatically assigned, but
10039 sometimes it is more convenient to fix them to ease monitoring. This value
10040 must be strictly positive and unique within the listener/frontend. This
10041 option can only be used when defining only a single socket.
10042
10043interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +010010044 Restricts the socket to a specific interface. When specified, only packets
10045 received from that particular interface are processed by the socket. This is
10046 currently only supported on Linux. The interface must be a primary system
10047 interface, not an aliased interface. It is also possible to bind multiple
10048 frontends to the same address if they are bound to different interfaces. Note
10049 that binding to a network interface requires root privileges. This parameter
10050 is only compatible with TCPv4/TCPv6 sockets.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010051
Willy Tarreauabb175f2012-09-24 12:43:26 +020010052level <level>
10053 This setting is used with the stats sockets only to restrict the nature of
10054 the commands that can be issued on the socket. It is ignored by other
10055 sockets. <level> can be one of :
10056 - "user" is the least privileged level ; only non-sensitive stats can be
10057 read, and no change is allowed. It would make sense on systems where it
10058 is not easy to restrict access to the socket.
10059 - "operator" is the default level and fits most common uses. All data can
10060 be read, and only non-sensitive changes are permitted (eg: clear max
10061 counters).
10062 - "admin" should be used with care, as everything is permitted (eg: clear
10063 all counters).
10064
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010065maxconn <maxconn>
10066 Limits the sockets to this number of concurrent connections. Extraneous
10067 connections will remain in the system's backlog until a connection is
10068 released. If unspecified, the limit will be the same as the frontend's
10069 maxconn. Note that in case of port ranges or multiple addresses, the same
10070 value will be applied to each socket. This setting enables different
10071 limitations on expensive sockets, for instance SSL entries which may easily
10072 eat all memory.
10073
10074mode <mode>
10075 Sets the octal mode used to define access permissions on the UNIX socket. It
10076 can also be set by default in the global section's "unix-bind" statement.
10077 Note that some platforms simply ignore this. This setting is ignored by non
10078 UNIX sockets.
10079
10080mss <maxseg>
10081 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
10082 connections. This can be used to force a lower MSS for certain specific
10083 ports, for instance for connections passing through a VPN. Note that this
10084 relies on a kernel feature which is theoretically supported under Linux but
10085 was buggy in all versions prior to 2.6.28. It may or may not work on other
10086 operating systems. It may also not change the advertised value but change the
10087 effective size of outgoing segments. The commonly advertised value for TCPv4
10088 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
10089 positive, it will be used as the advertised MSS. If it is negative, it will
10090 indicate by how much to reduce the incoming connection's advertised MSS for
10091 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
10092
10093name <name>
10094 Sets an optional name for these sockets, which will be reported on the stats
10095 page.
10096
Willy Tarreaud72f0f32015-10-13 14:50:22 +020010097namespace <name>
10098 On Linux, it is possible to specify which network namespace a socket will
10099 belong to. This directive makes it possible to explicitly bind a listener to
10100 a namespace different from the default one. Please refer to your operating
10101 system's documentation to find more details about network namespaces.
10102
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010103nice <nice>
10104 Sets the 'niceness' of connections initiated from the socket. Value must be
10105 in the range -1024..1024 inclusive, and defaults to zero. Positive values
10106 means that such connections are more friendly to others and easily offer
10107 their place in the scheduler. On the opposite, negative values mean that
10108 connections want to run with a higher priority than others. The difference
10109 only happens under high loads when the system is close to saturation.
10110 Negative values are appropriate for low-latency or administration services,
10111 and high values are generally recommended for CPU intensive tasks such as SSL
10112 processing or bulk transfers which are less sensible to latency. For example,
10113 it may make sense to use a positive value for an SMTP socket and a negative
10114 one for an RDP socket.
10115
Emeric Brun9b3009b2012-10-05 11:55:06 +020010116no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010117 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010118 disables support for SSLv3 on any sockets instantiated from the listener when
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010119 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010120 be enabled using any configuration option. This option is also available on
10121 global statement "ssl-default-bind-options". See also "force-tls*",
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010122 and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010123
Emeric Brun90ad8722012-10-02 14:00:59 +020010124no-tls-tickets
10125 This setting is only available when support for OpenSSL was built in. It
10126 disables the stateless session resumption (RFC 5077 TLS Ticket
10127 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010128 session resumption is more expensive in CPU usage. This option is also
10129 available on global statement "ssl-default-bind-options".
Emeric Brun90ad8722012-10-02 14:00:59 +020010130
Emeric Brun9b3009b2012-10-05 11:55:06 +020010131no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010132 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010133 disables support for TLSv1.0 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010134 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010135 cannot be enabled using any configuration option. This option is also
10136 available on global statement "ssl-default-bind-options". See also
10137 "force-tlsv*", and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010138
Emeric Brun9b3009b2012-10-05 11:55:06 +020010139no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +020010140 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010141 disables support for TLSv1.1 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010142 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010143 cannot be enabled using any configuration option. This option is also
10144 available on global statement "ssl-default-bind-options". See also
10145 "force-tlsv*", and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +020010146
Emeric Brun9b3009b2012-10-05 11:55:06 +020010147no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +020010148 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010149 disables support for TLSv1.2 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010150 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010151 cannot be enabled using any configuration option. This option is also
10152 available on global statement "ssl-default-bind-options". See also
10153 "force-tlsv*", and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +020010154
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020010155npn <protocols>
10156 This enables the NPN TLS extension and advertises the specified protocol list
10157 as supported on top of NPN. The protocol list consists in a comma-delimited
10158 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
10159 This requires that the SSL library is build with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +020010160 enabled (check with haproxy -vv). Note that the NPN extension has been
10161 replaced with the ALPN extension (see the "alpn" keyword).
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020010162
Willy Tarreau6ae1ba62014-05-07 19:01:58 +020010163process [ all | odd | even | <number 1-64>[-<number 1-64>] ]
10164 This restricts the list of processes on which this listener is allowed to
10165 run. It does not enforce any process but eliminates those which do not match.
10166 If the frontend uses a "bind-process" setting, the intersection between the
10167 two is applied. If in the end the listener is not allowed to run on any
10168 remaining process, a warning is emitted, and the listener will either run on
10169 the first process of the listener if a single process was specified, or on
10170 all of its processes if multiple processes were specified. For the unlikely
Willy Tarreauae302532014-05-07 19:22:24 +020010171 case where several ranges are needed, this directive may be repeated. The
10172 main purpose of this directive is to be used with the stats sockets and have
10173 one different socket per process. The second purpose is to have multiple bind
10174 lines sharing the same IP:port but not the same process in a listener, so
10175 that the system can distribute the incoming connections into multiple queues
10176 and allow a smoother inter-process load balancing. Currently Linux 3.9 and
10177 above is known for supporting this. See also "bind-process" and "nbproc".
Willy Tarreau6ae1ba62014-05-07 19:01:58 +020010178
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010179ssl
10180 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010181 enables SSL deciphering on connections instantiated from this listener. A
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010182 certificate is necessary (see "crt" above). All contents in the buffers will
10183 appear in clear text, so that ACLs and HTTP processing will only have access
10184 to deciphered contents.
10185
Emmanuel Hocdet65623372013-01-24 17:17:15 +010010186strict-sni
10187 This setting is only available when support for OpenSSL was built in. The
10188 SSL/TLS negotiation is allow only if the client provided an SNI which match
10189 a certificate. The default certificate is not used.
10190 See the "crt" option for more information.
10191
Willy Tarreau2af207a2015-02-04 00:45:58 +010010192tcp-ut <delay>
10193 Sets the TCP User Timeout for all incoming connections instanciated from this
10194 listening socket. This option is available on Linux since version 2.6.37. It
10195 allows haproxy to configure a timeout for sockets which contain data not
10196 receiving an acknoledgement for the configured delay. This is especially
10197 useful on long-lived connections experiencing long idle periods such as
10198 remote terminals or database connection pools, where the client and server
10199 timeouts must remain high to allow a long period of idle, but where it is
10200 important to detect that the client has disappeared in order to release all
10201 resources associated with its connection (and the server's session). The
10202 argument is a delay expressed in milliseconds by default. This only works
10203 for regular TCP connections, and is ignored for other protocols.
10204
Willy Tarreau1c862c52012-10-05 16:21:00 +020010205tfo
Lukas Tribus0defb902013-02-13 23:35:39 +010010206 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +020010207 enables TCP Fast Open on the listening socket, which means that clients which
10208 support this feature will be able to send a request and receive a response
10209 during the 3-way handshake starting from second connection, thus saving one
10210 round-trip after the first connection. This only makes sense with protocols
10211 that use high connection rates and where each round trip matters. This can
10212 possibly cause issues with many firewalls which do not accept data on SYN
10213 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +020010214 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
10215 need to build HAProxy with USE_TFO=1 if your libc doesn't define
10216 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +020010217
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010010218tls-ticket-keys <keyfile>
10219 Sets the TLS ticket keys file to load the keys from. The keys need to be 48
10220 bytes long, encoded with base64 (ex. openssl rand -base64 48). Number of keys
10221 is specified by the TLS_TICKETS_NO build option (default 3) and at least as
10222 many keys need to be present in the file. Last TLS_TICKETS_NO keys will be
10223 used for decryption and the penultimate one for encryption. This enables easy
10224 key rotation by just appending new key to the file and reloading the process.
10225 Keys must be periodically rotated (ex. every 12h) or Perfect Forward Secrecy
10226 is compromised. It is also a good idea to keep the keys off any permanent
10227 storage such as hard drives (hint: use tmpfs and don't swap those files).
10228 Lifetime hint can be changed using tune.ssl.timeout.
10229
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010230transparent
10231 Is an optional keyword which is supported only on certain Linux kernels. It
10232 indicates that the addresses will be bound even if they do not belong to the
10233 local machine, and that packets targeting any of these addresses will be
10234 intercepted just as if the addresses were locally configured. This normally
10235 requires that IP forwarding is enabled. Caution! do not use this with the
10236 default address '*', as it would redirect any traffic for the specified port.
10237 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
10238 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
10239 kernel version. Some distribution kernels include backports of the feature,
10240 so check for support with your vendor.
10241
Willy Tarreau77e3af92012-11-24 15:07:23 +010010242v4v6
10243 Is an optional keyword which is supported only on most recent systems
10244 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
10245 and IPv6 when it uses the default address. Doing so is sometimes necessary
10246 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010247 sockets, and is overridden by the "v6only" option.
Willy Tarreau77e3af92012-11-24 15:07:23 +010010248
Willy Tarreau9b6700f2012-11-24 11:55:28 +010010249v6only
10250 Is an optional keyword which is supported only on most recent systems
10251 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
10252 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +010010253 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
10254 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +010010255
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010256uid <uid>
10257 Sets the owner of the UNIX sockets to the designated system uid. It can also
10258 be set by default in the global section's "unix-bind" statement. Note that
10259 some platforms simply ignore this. This setting is equivalent to the "user"
10260 setting except that the user numeric ID is used instead of its name. This
10261 setting is ignored by non UNIX sockets.
10262
10263user <user>
10264 Sets the owner of the UNIX sockets to the designated system user. It can also
10265 be set by default in the global section's "unix-bind" statement. Note that
10266 some platforms simply ignore this. This setting is equivalent to the "uid"
10267 setting except that the user name is used instead of its uid. This setting is
10268 ignored by non UNIX sockets.
10269
Emeric Brun1a073b42012-09-28 17:07:34 +020010270verify [none|optional|required]
10271 This setting is only available when support for OpenSSL was built in. If set
10272 to 'none', client certificate is not requested. This is the default. In other
10273 cases, a client certificate is requested. If the client does not provide a
10274 certificate after the request and if 'verify' is set to 'required', then the
10275 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +020010276 certificate provided by the client is always verified using CAs from
10277 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
10278 is aborted, regardless of the 'verify' option, unless the error code exactly
10279 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020010280
Willy Tarreaub6205fd2012-09-24 12:27:33 +0200102815.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +010010282------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020010283
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010010284The "server" and "default-server" keywords support a certain number of settings
10285which are all passed as arguments on the server line. The order in which those
10286arguments appear does not count, and they are all optional. Some of those
10287settings are single words (booleans) while others expect one or several values
10288after them. In this case, the values must immediately follow the setting name.
10289Except default-server, all those settings must be specified after the server's
10290address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +020010291
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010292 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010010293 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +020010294
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010295The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010296
Willy Tarreauceb4ac92012-04-28 00:41:46 +020010297addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010298 Using the "addr" parameter, it becomes possible to use a different IP address
Baptiste Assmann13f83532016-03-06 23:14:36 +010010299 to send health-checks or to probe the agent-check. On some servers, it may be
10300 desirable to dedicate an IP address to specific component able to perform
10301 complex tests which are more suitable to health-checks than the application.
10302 This parameter is ignored if the "check" parameter is not set. See also the
10303 "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010304
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010305 Supported in default-server: No
10306
Simon Hormand60d6912013-11-25 10:46:36 +090010307agent-check
10308 Enable an auxiliary agent check which is run independently of a regular
Willy Tarreau81f5d942013-12-09 20:51:51 +010010309 health check. An agent health check is performed by making a TCP connection
10310 to the port set by the "agent-port" parameter and reading an ASCII string.
10311 The string is made of a series of words delimited by spaces, tabs or commas
10312 in any order, optionally terminated by '\r' and/or '\n', each consisting of :
Simon Hormand60d6912013-11-25 10:46:36 +090010313
Willy Tarreau81f5d942013-12-09 20:51:51 +010010314 - An ASCII representation of a positive integer percentage, e.g. "75%".
Simon Hormand60d6912013-11-25 10:46:36 +090010315 Values in this format will set the weight proportional to the initial
Willy Tarreauc5af3a62014-10-07 15:27:33 +020010316 weight of a server as configured when haproxy starts. Note that a zero
10317 weight is reported on the stats page as "DRAIN" since it has the same
10318 effect on the server (it's removed from the LB farm).
Simon Hormand60d6912013-11-25 10:46:36 +090010319
Nenad Merdanovic174dd372016-04-24 23:10:06 +020010320 - The string "maxconn:" followed by an integer (no space between). Values in
10321 this format will set the maxconn of a server. The maximum number of
10322 connections advertised needs to be multipled by the number of load balancers
10323 and different backends that use this health check to get the total number
10324 of connections the server might receive. Example: maxconn:30
10325
Willy Tarreau81f5d942013-12-09 20:51:51 +010010326 - The word "ready". This will turn the server's administrative state to the
10327 READY mode, thus cancelling any DRAIN or MAINT state
Simon Hormand60d6912013-11-25 10:46:36 +090010328
Willy Tarreau81f5d942013-12-09 20:51:51 +010010329 - The word "drain". This will turn the server's administrative state to the
10330 DRAIN mode, thus it will not accept any new connections other than those
10331 that are accepted via persistence.
Simon Hormand60d6912013-11-25 10:46:36 +090010332
Willy Tarreau81f5d942013-12-09 20:51:51 +010010333 - The word "maint". This will turn the server's administrative state to the
10334 MAINT mode, thus it will not accept any new connections at all, and health
10335 checks will be stopped.
Simon Hormand60d6912013-11-25 10:46:36 +090010336
Willy Tarreau81f5d942013-12-09 20:51:51 +010010337 - The words "down", "failed", or "stopped", optionally followed by a
10338 description string after a sharp ('#'). All of these mark the server's
10339 operating state as DOWN, but since the word itself is reported on the stats
10340 page, the difference allows an administrator to know if the situation was
10341 expected or not : the service may intentionally be stopped, may appear up
10342 but fail some validity tests, or may be seen as down (eg: missing process,
10343 or port not responding).
Simon Hormand60d6912013-11-25 10:46:36 +090010344
Willy Tarreau81f5d942013-12-09 20:51:51 +010010345 - The word "up" sets back the server's operating state as UP if health checks
10346 also report that the service is accessible.
Simon Hormand60d6912013-11-25 10:46:36 +090010347
Willy Tarreau81f5d942013-12-09 20:51:51 +010010348 Parameters which are not advertised by the agent are not changed. For
10349 example, an agent might be designed to monitor CPU usage and only report a
10350 relative weight and never interact with the operating status. Similarly, an
10351 agent could be designed as an end-user interface with 3 radio buttons
10352 allowing an administrator to change only the administrative state. However,
10353 it is important to consider that only the agent may revert its own actions,
10354 so if a server is set to DRAIN mode or to DOWN state using the agent, the
10355 agent must implement the other equivalent actions to bring the service into
10356 operations again.
Simon Hormand60d6912013-11-25 10:46:36 +090010357
Simon Horman2f1f9552013-11-25 10:46:37 +090010358 Failure to connect to the agent is not considered an error as connectivity
10359 is tested by the regular health check which is enabled by the "check"
Willy Tarreau81f5d942013-12-09 20:51:51 +010010360 parameter. Warning though, it is not a good idea to stop an agent after it
10361 reports "down", since only an agent reporting "up" will be able to turn the
10362 server up again. Note that the CLI on the Unix stats socket is also able to
Willy Tarreau989222a2016-01-15 10:26:26 +010010363 force an agent's result in order to work around a bogus agent if needed.
Simon Horman2f1f9552013-11-25 10:46:37 +090010364
Willy Tarreau81f5d942013-12-09 20:51:51 +010010365 Requires the "agent-port" parameter to be set. See also the "agent-inter"
10366 parameter.
Simon Hormand60d6912013-11-25 10:46:36 +090010367
10368 Supported in default-server: No
10369
James Brown55f9ff12015-10-21 18:19:05 -070010370agent-send <string>
10371 If this option is specified, haproxy will send the given string (verbatim)
10372 to the agent server upon connection. You could, for example, encode
10373 the backend name into this string, which would enable your agent to send
10374 different responses based on the backend. Make sure to include a '\n' if
10375 you want to terminate your request with a newline.
10376
Simon Hormand60d6912013-11-25 10:46:36 +090010377agent-inter <delay>
10378 The "agent-inter" parameter sets the interval between two agent checks
10379 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
10380
10381 Just as with every other time-based parameter, it may be entered in any
10382 other explicit unit among { us, ms, s, m, h, d }. The "agent-inter"
10383 parameter also serves as a timeout for agent checks "timeout check" is
10384 not set. In order to reduce "resonance" effects when multiple servers are
10385 hosted on the same hardware, the agent and health checks of all servers
10386 are started with a small time offset between them. It is also possible to
10387 add some random noise in the agent and health checks interval using the
10388 global "spread-checks" keyword. This makes sense for instance when a lot
10389 of backends use the same servers.
10390
10391 See also the "agent-check" and "agent-port" parameters.
10392
10393 Supported in default-server: Yes
10394
10395agent-port <port>
10396 The "agent-port" parameter sets the TCP port used for agent checks.
10397
10398 See also the "agent-check" and "agent-inter" parameters.
10399
10400 Supported in default-server: Yes
10401
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010402backup
10403 When "backup" is present on a server line, the server is only used in load
10404 balancing when all other non-backup servers are unavailable. Requests coming
10405 with a persistence cookie referencing the server will always be served
10406 though. By default, only the first operational backup server is used, unless
10407 the "allbackups" option is set in the backend. See also the "allbackups"
10408 option.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010409
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010410 Supported in default-server: No
10411
Emeric Brunef42d922012-10-11 16:11:36 +020010412ca-file <cafile>
10413 This setting is only available when support for OpenSSL was built in. It
10414 designates a PEM file from which to load CA certificates used to verify
10415 server's certificate.
10416
10417 Supported in default-server: No
10418
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010419check
10420 This option enables health checks on the server. By default, a server is
Patrick Mézardb7aeec62012-01-22 16:01:22 +010010421 always considered available. If "check" is set, the server is available when
10422 accepting periodic TCP connections, to ensure that it is really able to serve
10423 requests. The default address and port to send the tests to are those of the
10424 server, and the default source is the same as the one defined in the
10425 backend. It is possible to change the address using the "addr" parameter, the
10426 port using the "port" parameter, the source address using the "source"
10427 address, and the interval and timers using the "inter", "rise" and "fall"
Simon Hormanafc47ee2013-11-25 10:46:35 +090010428 parameters. The request method is define in the backend using the "httpchk",
10429 "smtpchk", "mysql-check", "pgsql-check" and "ssl-hello-chk" options. Please
10430 refer to those options and parameters for more information.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010431
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010432 Supported in default-server: No
10433
Willy Tarreau6c16adc2012-10-05 00:04:16 +020010434check-send-proxy
10435 This option forces emission of a PROXY protocol line with outgoing health
10436 checks, regardless of whether the server uses send-proxy or not for the
10437 normal traffic. By default, the PROXY protocol is enabled for health checks
10438 if it is already enabled for normal traffic and if no "port" nor "addr"
10439 directive is present. However, if such a directive is present, the
10440 "check-send-proxy" option needs to be used to force the use of the
10441 protocol. See also the "send-proxy" option for more information.
10442
10443 Supported in default-server: No
10444
Willy Tarreau763a95b2012-10-04 23:15:39 +020010445check-ssl
10446 This option forces encryption of all health checks over SSL, regardless of
10447 whether the server uses SSL or not for the normal traffic. This is generally
10448 used when an explicit "port" or "addr" directive is specified and SSL health
10449 checks are not inherited. It is important to understand that this option
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010450 inserts an SSL transport layer below the checks, so that a simple TCP connect
Willy Tarreau763a95b2012-10-04 23:15:39 +020010451 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
10452 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
10453 All SSL settings are common to health checks and traffic (eg: ciphers).
10454 See the "ssl" option for more information.
10455
10456 Supported in default-server: No
10457
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010458ciphers <ciphers>
10459 This option sets the string describing the list of cipher algorithms that is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010460 is negotiated during the SSL/TLS handshake with the server. The format of the
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010461 string is defined in "man 1 ciphers". When SSL is used to communicate with
10462 servers on the local network, it is common to see a weaker set of algorithms
10463 than what is used over the internet. Doing so reduces CPU usage on both the
10464 server and haproxy while still keeping it compatible with deployed software.
10465 Some algorithms such as RC4-SHA1 are reasonably cheap. If no security at all
10466 is needed and just connectivity, using DES can be appropriate.
10467
Willy Tarreau763a95b2012-10-04 23:15:39 +020010468 Supported in default-server: No
10469
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010470cookie <value>
10471 The "cookie" parameter sets the cookie value assigned to the server to
10472 <value>. This value will be checked in incoming requests, and the first
10473 operational server possessing the same value will be selected. In return, in
10474 cookie insertion or rewrite modes, this value will be assigned to the cookie
10475 sent to the client. There is nothing wrong in having several servers sharing
10476 the same cookie value, and it is in fact somewhat common between normal and
10477 backup servers. See also the "cookie" keyword in backend section.
10478
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010479 Supported in default-server: No
10480
Emeric Brunef42d922012-10-11 16:11:36 +020010481crl-file <crlfile>
10482 This setting is only available when support for OpenSSL was built in. It
10483 designates a PEM file from which to load certificate revocation list used
10484 to verify server's certificate.
10485
10486 Supported in default-server: No
10487
Emeric Bruna7aa3092012-10-26 12:58:00 +020010488crt <cert>
10489 This setting is only available when support for OpenSSL was built in.
10490 It designates a PEM file from which to load both a certificate and the
10491 associated private key. This file can be built by concatenating both PEM
10492 files into one. This certificate will be sent if the server send a client
10493 certificate request.
10494
10495 Supported in default-server: No
10496
Willy Tarreau96839092010-03-29 10:02:24 +020010497disabled
10498 The "disabled" keyword starts the server in the "disabled" state. That means
10499 that it is marked down in maintenance mode, and no connection other than the
10500 ones allowed by persist mode will reach it. It is very well suited to setup
10501 new servers, because normal traffic will never reach them, while it is still
10502 possible to test the service by making use of the force-persist mechanism.
10503
10504 Supported in default-server: No
10505
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010506error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +010010507 If health observing is enabled, the "error-limit" parameter specifies the
10508 number of consecutive errors that triggers event selected by the "on-error"
10509 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010510
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010511 Supported in default-server: Yes
10512
10513 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010514
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010515fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010516 The "fall" parameter states that a server will be considered as dead after
10517 <count> consecutive unsuccessful health checks. This value defaults to 3 if
10518 unspecified. See also the "check", "inter" and "rise" parameters.
10519
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010520 Supported in default-server: Yes
10521
Emeric Brun8694b9a2012-10-05 14:39:07 +020010522force-sslv3
10523 This option enforces use of SSLv3 only when SSL is used to communicate with
10524 the server. SSLv3 is generally less expensive than the TLS counterparts for
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010525 high connection rates. This option is also available on global statement
10526 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010527
10528 Supported in default-server: No
10529
10530force-tlsv10
10531 This option enforces use of TLSv1.0 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010532 the server. This option is also available on global statement
10533 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010534
10535 Supported in default-server: No
10536
10537force-tlsv11
10538 This option enforces use of TLSv1.1 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010539 the server. This option is also available on global statement
10540 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010541
10542 Supported in default-server: No
10543
10544force-tlsv12
10545 This option enforces use of TLSv1.2 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010546 the server. This option is also available on global statement
10547 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010548
10549 Supported in default-server: No
10550
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010551id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +020010552 Set a persistent ID for the server. This ID must be positive and unique for
10553 the proxy. An unused ID will automatically be assigned if unset. The first
10554 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010555
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010556 Supported in default-server: No
10557
10558inter <delay>
10559fastinter <delay>
10560downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010561 The "inter" parameter sets the interval between two consecutive health checks
10562 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
10563 It is also possible to use "fastinter" and "downinter" to optimize delays
10564 between checks depending on the server state :
10565
Pieter Baauw44fc9df2015-09-17 21:30:46 +020010566 Server state | Interval used
10567 ----------------------------------------+----------------------------------
10568 UP 100% (non-transitional) | "inter"
10569 ----------------------------------------+----------------------------------
10570 Transitionally UP (going down "fall"), | "fastinter" if set,
10571 Transitionally DOWN (going up "rise"), | "inter" otherwise.
10572 or yet unchecked. |
10573 ----------------------------------------+----------------------------------
10574 DOWN 100% (non-transitional) | "downinter" if set,
10575 | "inter" otherwise.
10576 ----------------------------------------+----------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +010010577
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010578 Just as with every other time-based parameter, they can be entered in any
10579 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
10580 serves as a timeout for health checks sent to servers if "timeout check" is
10581 not set. In order to reduce "resonance" effects when multiple servers are
Simon Hormand60d6912013-11-25 10:46:36 +090010582 hosted on the same hardware, the agent and health checks of all servers
10583 are started with a small time offset between them. It is also possible to
10584 add some random noise in the agent and health checks interval using the
10585 global "spread-checks" keyword. This makes sense for instance when a lot
10586 of backends use the same servers.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010587
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010588 Supported in default-server: Yes
10589
10590maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010591 The "maxconn" parameter specifies the maximal number of concurrent
10592 connections that will be sent to this server. If the number of incoming
10593 concurrent requests goes higher than this value, they will be queued, waiting
10594 for a connection to be released. This parameter is very important as it can
10595 save fragile servers from going down under extreme loads. If a "minconn"
10596 parameter is specified, the limit becomes dynamic. The default value is "0"
10597 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
10598 the backend's "fullconn" keyword.
10599
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010600 Supported in default-server: Yes
10601
10602maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010603 The "maxqueue" parameter specifies the maximal number of connections which
10604 will wait in the queue for this server. If this limit is reached, next
10605 requests will be redispatched to other servers instead of indefinitely
10606 waiting to be served. This will break persistence but may allow people to
10607 quickly re-log in when the server they try to connect to is dying. The
10608 default value is "0" which means the queue is unlimited. See also the
10609 "maxconn" and "minconn" parameters.
10610
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010611 Supported in default-server: Yes
10612
10613minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010614 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
10615 limit following the backend's load. The server will always accept at least
10616 <minconn> connections, never more than <maxconn>, and the limit will be on
10617 the ramp between both values when the backend has less than <fullconn>
10618 concurrent connections. This makes it possible to limit the load on the
10619 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010620 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010621 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010622
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010623 Supported in default-server: Yes
10624
Willy Tarreaud72f0f32015-10-13 14:50:22 +020010625namespace <name>
10626 On Linux, it is possible to specify which network namespace a socket will
10627 belong to. This directive makes it possible to explicitly bind a server to
10628 a namespace different from the default one. Please refer to your operating
10629 system's documentation to find more details about network namespaces.
10630
Willy Tarreau2a3fb1c2015-02-05 16:47:07 +010010631no-ssl-reuse
10632 This option disables SSL session reuse when SSL is used to communicate with
10633 the server. It will force the server to perform a full handshake for every
10634 new connection. It's probably only useful for benchmarking, troubleshooting,
10635 and for paranoid users.
10636
10637 Supported in default-server: No
10638
Emeric Brun9b3009b2012-10-05 11:55:06 +020010639no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010640 This option disables support for SSLv3 when SSL is used to communicate with
10641 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emeric Brun8694b9a2012-10-05 14:39:07 +020010642 using any configuration option. See also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010643
Willy Tarreau763a95b2012-10-04 23:15:39 +020010644 Supported in default-server: No
10645
Emeric Brunf9c5c472012-10-11 15:28:34 +020010646no-tls-tickets
10647 This setting is only available when support for OpenSSL was built in. It
10648 disables the stateless session resumption (RFC 5077 TLS Ticket
10649 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010650 session resumption is more expensive in CPU usage for servers. This option
10651 is also available on global statement "ssl-default-server-options".
Emeric Brunf9c5c472012-10-11 15:28:34 +020010652
10653 Supported in default-server: No
10654
Emeric Brun9b3009b2012-10-05 11:55:06 +020010655no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +020010656 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020010657 the server. Note that SSLv2 is disabled in the code and cannot be enabled
10658 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010659 often makes sense to disable it when communicating with local servers. This
10660 option is also available on global statement "ssl-default-server-options".
10661 See also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +020010662
Willy Tarreau763a95b2012-10-04 23:15:39 +020010663 Supported in default-server: No
10664
Emeric Brun9b3009b2012-10-05 11:55:06 +020010665no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +020010666 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020010667 the server. Note that SSLv2 is disabled in the code and cannot be enabled
10668 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010669 often makes sense to disable it when communicating with local servers. This
10670 option is also available on global statement "ssl-default-server-options".
10671 See also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +020010672
Willy Tarreau763a95b2012-10-04 23:15:39 +020010673 Supported in default-server: No
10674
Emeric Brun9b3009b2012-10-05 11:55:06 +020010675no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +020010676 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010677 the server. Note that SSLv2 is disabled in the code and cannot be enabled
10678 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010679 often makes sense to disable it when communicating with local servers. This
10680 option is also available on global statement "ssl-default-server-options".
10681 See also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010682
Willy Tarreau763a95b2012-10-04 23:15:39 +020010683 Supported in default-server: No
10684
Simon Hormanfa461682011-06-25 09:39:49 +090010685non-stick
10686 Never add connections allocated to this sever to a stick-table.
10687 This may be used in conjunction with backup to ensure that
10688 stick-table persistence is disabled for backup servers.
10689
Willy Tarreau763a95b2012-10-04 23:15:39 +020010690 Supported in default-server: No
10691
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010692observe <mode>
10693 This option enables health adjusting based on observing communication with
10694 the server. By default this functionality is disabled and enabling it also
10695 requires to enable health checks. There are two supported modes: "layer4" and
10696 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
10697 significant. In layer7, which is only allowed for http proxies, responses
10698 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +010010699 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010700
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010701 Supported in default-server: No
10702
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010703 See also the "check", "on-error" and "error-limit".
10704
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010705on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010706 Select what should happen when enough consecutive errors are detected.
10707 Currently, four modes are available:
10708 - fastinter: force fastinter
10709 - fail-check: simulate a failed check, also forces fastinter (default)
10710 - sudden-death: simulate a pre-fatal failed health check, one more failed
10711 check will mark a server down, forces fastinter
10712 - mark-down: mark the server immediately down and force fastinter
10713
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010714 Supported in default-server: Yes
10715
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010716 See also the "check", "observe" and "error-limit".
10717
Simon Hormane0d1bfb2011-06-21 14:34:58 +090010718on-marked-down <action>
10719 Modify what occurs when a server is marked down.
10720 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070010721 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
10722 all connections to the server are immediately terminated when the server
10723 goes down. It might be used if the health check detects more complex cases
10724 than a simple connection status, and long timeouts would cause the service
10725 to remain unresponsive for too long a time. For instance, a health check
10726 might detect that a database is stuck and that there's no chance to reuse
10727 existing connections anymore. Connections killed this way are logged with
10728 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +090010729
10730 Actions are disabled by default
10731
10732 Supported in default-server: Yes
10733
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070010734on-marked-up <action>
10735 Modify what occurs when a server is marked up.
10736 Currently one action is available:
10737 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
10738 done only if the server is not in backup state and if it is not disabled
10739 (it must have an effective weight > 0). This can be used sometimes to force
10740 an active server to take all the traffic back after recovery when dealing
10741 with long sessions (eg: LDAP, SQL, ...). Doing this can cause more trouble
10742 than it tries to solve (eg: incomplete transactions), so use this feature
10743 with extreme care. Sessions killed because a server comes up are logged
10744 with an 'U' termination code (for "Up").
10745
10746 Actions are disabled by default
10747
10748 Supported in default-server: Yes
10749
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010750port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010751 Using the "port" parameter, it becomes possible to use a different port to
10752 send health-checks. On some servers, it may be desirable to dedicate a port
10753 to a specific component able to perform complex tests which are more suitable
10754 to health-checks than the application. It is common to run a simple script in
10755 inetd for instance. This parameter is ignored if the "check" parameter is not
10756 set. See also the "addr" parameter.
10757
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010758 Supported in default-server: Yes
10759
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010760redir <prefix>
10761 The "redir" parameter enables the redirection mode for all GET and HEAD
10762 requests addressing this server. This means that instead of having HAProxy
10763 forward the request to the server, it will send an "HTTP 302" response with
10764 the "Location" header composed of this prefix immediately followed by the
10765 requested URI beginning at the leading '/' of the path component. That means
10766 that no trailing slash should be used after <prefix>. All invalid requests
10767 will be rejected, and all non-GET or HEAD requests will be normally served by
10768 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010769 mangling nor cookie insertion is possible in the response. However, cookies in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010770 requests are still analysed, making this solution completely usable to direct
10771 users to a remote location in case of local disaster. Main use consists in
10772 increasing bandwidth for static servers by having the clients directly
10773 connect to them. Note: never use a relative location here, it would cause a
10774 loop between the client and HAProxy!
10775
10776 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
10777
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010778 Supported in default-server: No
10779
10780rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010781 The "rise" parameter states that a server will be considered as operational
10782 after <count> consecutive successful health checks. This value defaults to 2
10783 if unspecified. See also the "check", "inter" and "fall" parameters.
10784
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010785 Supported in default-server: Yes
10786
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010787resolve-prefer <family>
10788 When DNS resolution is enabled for a server and multiple IP addresses from
10789 different families are returned, HAProxy will prefer using an IP address
10790 from the family mentioned in the "resolve-prefer" parameter.
10791 Available families: "ipv4" and "ipv6"
10792
Baptiste Assmannc4aabae2015-08-04 22:43:06 +020010793 Default value: ipv6
10794
10795 Supported in default-server: Yes
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010796
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020010797 Example:
10798
10799 server s1 app1.domain.com:80 resolvers mydns resolve-prefer ipv6
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010800
Thierry Fournierac88cfe2016-02-17 22:05:30 +010010801resolve-net <network>[,<network[,...]]
10802 This options prioritize th choice of an ip address matching a network. This is
10803 useful with clouds to prefer a local ip. In some cases, a cloud high
10804 avalailibility service can be announced with many ip addresses on many
10805 differents datacenters. The latency between datacenter is not negligible, so
10806 this patch permitsto prefers a local datacenter. If none address matchs the
10807 configured network, another address is selected.
10808
10809 Supported in default-server: Yes
10810
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020010811 Example:
10812
10813 server s1 app1.domain.com:80 resolvers mydns resolve-net 10.0.0.0/8
Thierry Fournierac88cfe2016-02-17 22:05:30 +010010814
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010815resolvers <id>
10816 Points to an existing "resolvers" section to resolve current server's
10817 hostname.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020010818 In order to be operational, DNS resolution requires that health check is
10819 enabled on the server. Actually, health checks triggers the DNS resolution.
10820 You must precise one 'resolvers' parameter on each server line where DNS
10821 resolution is required.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010822
Baptiste Assmann62b75b42015-09-09 01:11:36 +020010823 Supported in default-server: No
10824
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020010825 Example:
10826
10827 server s1 app1.domain.com:80 check resolvers mydns
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010828
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020010829 See also section 5.3
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010830
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010010831send-proxy
10832 The "send-proxy" parameter enforces use of the PROXY protocol over any
10833 connection established to this server. The PROXY protocol informs the other
10834 end about the layer 3/4 addresses of the incoming connection, so that it can
10835 know the client's address or the public address it accessed to, whatever the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010010836 upper layer protocol. For connections accepted by an "accept-proxy" or
10837 "accept-netscaler-cip" listener, the advertised address will be used. Only
10838 TCPv4 and TCPv6 address families are supported. Other families such as
10839 Unix sockets, will report an UNKNOWN family. Servers using this option can
10840 fully be chained to another instance of haproxy listening with an
10841 "accept-proxy" setting. This setting must not be used if the server isn't
10842 aware of the protocol. When health checks are sent to the server, the PROXY
10843 protocol is automatically used when this option is set, unless there is an
10844 explicit "port" or "addr" directive, in which case an explicit
10845 "check-send-proxy" directive would also be needed to use the PROXY protocol.
10846 See also the "accept-proxy" and "accept-netscaler-cip" option of the "bind"
10847 keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010010848
10849 Supported in default-server: No
10850
David Safb76832014-05-08 23:42:08 -040010851send-proxy-v2
10852 The "send-proxy-v2" parameter enforces use of the PROXY protocol version 2
10853 over any connection established to this server. The PROXY protocol informs
10854 the other end about the layer 3/4 addresses of the incoming connection, so
10855 that it can know the client's address or the public address it accessed to,
10856 whatever the upper layer protocol. This setting must not be used if the
10857 server isn't aware of this version of the protocol. See also the "send-proxy"
10858 option of the "bind" keyword.
10859
10860 Supported in default-server: No
10861
10862send-proxy-v2-ssl
10863 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
10864 2 over any connection established to this server. The PROXY protocol informs
10865 the other end about the layer 3/4 addresses of the incoming connection, so
10866 that it can know the client's address or the public address it accessed to,
10867 whatever the upper layer protocol. In addition, the SSL information extension
10868 of the PROXY protocol is added to the PROXY protocol header. This setting
10869 must not be used if the server isn't aware of this version of the protocol.
10870 See also the "send-proxy-v2" option of the "bind" keyword.
10871
10872 Supported in default-server: No
10873
10874send-proxy-v2-ssl-cn
10875 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
10876 2 over any connection established to this server. The PROXY protocol informs
10877 the other end about the layer 3/4 addresses of the incoming connection, so
10878 that it can know the client's address or the public address it accessed to,
10879 whatever the upper layer protocol. In addition, the SSL information extension
10880 of the PROXY protocol, along along with the Common Name from the subject of
10881 the client certificate (if any), is added to the PROXY protocol header. This
10882 setting must not be used if the server isn't aware of this version of the
10883 protocol. See also the "send-proxy-v2" option of the "bind" keyword.
10884
10885 Supported in default-server: No
10886
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010887slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010888 The "slowstart" parameter for a server accepts a value in milliseconds which
10889 indicates after how long a server which has just come back up will run at
10890 full speed. Just as with every other time-based parameter, it can be entered
10891 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
10892 linearly from 0 to 100% during this time. The limitation applies to two
10893 parameters :
10894
10895 - maxconn: the number of connections accepted by the server will grow from 1
10896 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
10897
10898 - weight: when the backend uses a dynamic weighted algorithm, the weight
10899 grows linearly from 1 to 100%. In this case, the weight is updated at every
10900 health-check. For this reason, it is important that the "inter" parameter
10901 is smaller than the "slowstart", in order to maximize the number of steps.
10902
10903 The slowstart never applies when haproxy starts, otherwise it would cause
10904 trouble to running servers. It only applies when a server has been previously
10905 seen as failed.
10906
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010907 Supported in default-server: Yes
10908
Willy Tarreau732eac42015-07-09 11:40:25 +020010909sni <expression>
10910 The "sni" parameter evaluates the sample fetch expression, converts it to a
10911 string and uses the result as the host name sent in the SNI TLS extension to
10912 the server. A typical use case is to send the SNI received from the client in
10913 a bridged HTTPS scenario, using the "ssl_fc_sni" sample fetch for the
10914 expression, though alternatives such as req.hdr(host) can also make sense.
10915
10916 Supported in default-server: no
10917
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020010918source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +020010919source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020010920source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010921 The "source" parameter sets the source address which will be used when
10922 connecting to the server. It follows the exact same parameters and principle
10923 as the backend "source" keyword, except that it only applies to the server
10924 referencing it. Please consult the "source" keyword for details.
10925
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020010926 Additionally, the "source" statement on a server line allows one to specify a
10927 source port range by indicating the lower and higher bounds delimited by a
10928 dash ('-'). Some operating systems might require a valid IP address when a
10929 source port range is specified. It is permitted to have the same IP/range for
10930 several servers. Doing so makes it possible to bypass the maximum of 64k
10931 total concurrent connections. The limit will then reach 64k connections per
10932 server.
10933
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010934 Supported in default-server: No
10935
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010936ssl
Willy Tarreau44f65392013-06-25 07:56:20 +020010937 This option enables SSL ciphering on outgoing connections to the server. It
10938 is critical to verify server certificates using "verify" when using SSL to
10939 connect to servers, otherwise the communication is prone to trivial man in
10940 the-middle attacks rendering SSL useless. When this option is used, health
10941 checks are automatically sent in SSL too unless there is a "port" or an
10942 "addr" directive indicating the check should be sent to a different location.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010943 See the "check-ssl" option to force SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +020010944
10945 Supported in default-server: No
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010946
Willy Tarreau163d4622015-10-13 16:16:41 +020010947tcp-ut <delay>
10948 Sets the TCP User Timeout for all outgoing connections to this server. This
10949 option is available on Linux since version 2.6.37. It allows haproxy to
10950 configure a timeout for sockets which contain data not receiving an
10951 acknoledgement for the configured delay. This is especially useful on
10952 long-lived connections experiencing long idle periods such as remote
10953 terminals or database connection pools, where the client and server timeouts
10954 must remain high to allow a long period of idle, but where it is important to
10955 detect that the server has disappeared in order to release all resources
10956 associated with its connection (and the client's session). One typical use
10957 case is also to force dead server connections to die when health checks are
10958 too slow or during a soft reload since health checks are then disabled. The
10959 argument is a delay expressed in milliseconds by default. This only works for
10960 regular TCP connections, and is ignored for other protocols.
10961
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010962track [<proxy>/]<server>
Willy Tarreau32091232014-05-16 13:52:00 +020010963 This option enables ability to set the current state of the server by tracking
10964 another one. It is possible to track a server which itself tracks another
10965 server, provided that at the end of the chain, a server has health checks
10966 enabled. If <proxy> is omitted the current one is used. If disable-on-404 is
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010967 used, it has to be enabled on both proxies.
10968
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010969 Supported in default-server: No
10970
Emeric Brunef42d922012-10-11 16:11:36 +020010971verify [none|required]
10972 This setting is only available when support for OpenSSL was built in. If set
Emeric Brun850efd52014-01-29 12:24:34 +010010973 to 'none', server certificate is not verified. In the other case, The
10974 certificate provided by the server is verified using CAs from 'ca-file'
10975 and optional CRLs from 'crl-file'. If 'ssl_server_verify' is not specified
10976 in global section, this is the default. On verify failure the handshake
Willy Tarreau44f65392013-06-25 07:56:20 +020010977 is aborted. It is critically important to verify server certificates when
10978 using SSL to connect to servers, otherwise the communication is prone to
10979 trivial man-in-the-middle attacks rendering SSL totally useless.
Emeric Brunef42d922012-10-11 16:11:36 +020010980
10981 Supported in default-server: No
10982
Evan Broderbe554312013-06-27 00:05:25 -070010983verifyhost <hostname>
10984 This setting is only available when support for OpenSSL was built in, and
10985 only takes effect if 'verify required' is also specified. When set, the
10986 hostnames in the subject and subjectAlternateNames of the certificate
10987 provided by the server are checked. If none of the hostnames in the
10988 certificate match the specified hostname, the handshake is aborted. The
10989 hostnames in the server-provided certificate may include wildcards.
10990
10991 Supported in default-server: No
10992
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010993weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010994 The "weight" parameter is used to adjust the server's weight relative to
10995 other servers. All servers will receive a load proportional to their weight
10996 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +020010997 load. The default weight is 1, and the maximal value is 256. A value of 0
10998 means the server will not participate in load-balancing but will still accept
10999 persistent connections. If this parameter is used to distribute the load
11000 according to server's capacity, it is recommended to start with values which
11001 can both grow and shrink, for instance between 10 and 100 to leave enough
11002 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011003
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011004 Supported in default-server: Yes
11005
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011006
Cyril Bonté46175dd2015-07-02 22:45:32 +0200110075.3. Server IP address resolution using DNS
11008-------------------------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011009
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011010HAProxy allows using a host name on the server line to retrieve its IP address
11011using name servers. By default, HAProxy resolves the name when parsing the
11012configuration file, at startup and cache the result for the process' life.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011013This is not sufficient in some cases, such as in Amazon where a server's IP
11014can change after a reboot or an ELB Virtual IP can change based on current
11015workload.
11016This chapter describes how HAProxy can be configured to process server's name
11017resolution at run time.
11018Whether run time server name resolution has been enable or not, HAProxy will
11019carry on doing the first resolution when parsing the configuration.
11020
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011021Bear in mind that DNS resolution is triggered by health checks. This makes
11022health checks mandatory to allow DNS resolution.
11023
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011024
Cyril Bonté46175dd2015-07-02 22:45:32 +0200110255.3.1. Global overview
11026----------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011027
11028As we've seen in introduction, name resolution in HAProxy occurs at two
11029different steps of the process life:
11030
11031 1. when starting up, HAProxy parses the server line definition and matches a
11032 host name. It uses libc functions to get the host name resolved. This
11033 resolution relies on /etc/resolv.conf file.
11034
11035 2. at run time, when HAProxy gets prepared to run a health check on a server,
11036 it verifies if the current name resolution is still considered as valid.
11037 If not, it processes a new resolution, in parallel of the health check.
11038
11039A few other events can trigger a name resolution at run time:
11040 - when a server's health check ends up in a connection timeout: this may be
11041 because the server has a new IP address. So we need to trigger a name
11042 resolution to know this new IP.
11043
11044A few things important to notice:
11045 - all the name servers are queried in the mean time. HAProxy will process the
11046 first valid response.
11047
11048 - a resolution is considered as invalid (NX, timeout, refused), when all the
11049 servers return an error.
11050
11051
Cyril Bonté46175dd2015-07-02 22:45:32 +0200110525.3.2. The resolvers section
11053----------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011054
11055This section is dedicated to host information related to name resolution in
11056HAProxy.
11057There can be as many as resolvers section as needed. Each section can contain
11058many name servers.
11059
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011060When multiple name servers are configured in a resolvers section, then HAProxy
11061uses the first valid response. In case of invalid responses, only the last one
11062is treated. Purpose is to give the chance to a slow server to deliver a valid
11063answer after a fast faulty or outdated server.
11064
11065When each server returns a different error type, then only the last error is
11066used by HAProxy to decide what type of behavior to apply.
11067
11068Two types of behavior can be applied:
11069 1. stop DNS resolution
11070 2. replay the DNS query with a new query type
11071 In such case, the following types are applied in this exact order:
11072 1. ANY query type
11073 2. query type corresponding to family pointed by resolve-prefer
11074 server's parameter
11075 3. remaining family type
11076
11077HAProxy stops DNS resolution when the following errors occur:
11078 - invalid DNS response packet
11079 - wrong name in the query section of the response
11080 - NX domain
11081 - Query refused by server
11082 - CNAME not pointing to an IP address
11083
11084HAProxy tries a new query type when the following errors occur:
11085 - no Answer records in the response
11086 - DNS response truncated
11087 - Error in DNS response
11088 - No expected DNS records found in the response
11089 - name server timeout
11090
11091For example, with 2 name servers configured in a resolvers section:
11092 - first response is valid and is applied directly, second response is ignored
11093 - first response is invalid and second one is valid, then second response is
11094 applied;
11095 - first response is a NX domain and second one a truncated response, then
11096 HAProxy replays the query with a new type;
11097 - first response is truncated and second one is a NX Domain, then HAProxy
11098 stops resolution.
11099
11100
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011101resolvers <resolvers id>
11102 Creates a new name server list labelled <resolvers id>
11103
11104A resolvers section accept the following parameters:
11105
11106nameserver <id> <ip>:<port>
11107 DNS server description:
11108 <id> : label of the server, should be unique
11109 <ip> : IP address of the server
11110 <port> : port where the DNS service actually runs
11111
11112hold <status> <period>
11113 Defines <period> during which the last name resolution should be kept based
11114 on last resolution <status>
11115 <status> : last name resolution status. Only "valid" is accepted for now.
11116 <period> : interval between two successive name resolution when the last
11117 answer was in <status>. It follows the HAProxy time format.
11118 <period> is in milliseconds by default.
11119
11120 Default value is 10s for "valid".
11121
11122 Note: since the name resolution is triggered by the health checks, a new
11123 resolution is triggered after <period> modulo the <inter> parameter of
11124 the healch check.
11125
11126resolve_retries <nb>
11127 Defines the number <nb> of queries to send to resolve a server name before
11128 giving up.
11129 Default value: 3
11130
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011131 A retry occurs on name server timeout or when the full sequence of DNS query
11132 type failover is over and we need to start up from the default ANY query
11133 type.
11134
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011135timeout <event> <time>
11136 Defines timeouts related to name resolution
11137 <event> : the event on which the <time> timeout period applies to.
11138 events available are:
11139 - retry: time between two DNS queries, when no response have
11140 been received.
11141 Default value: 1s
11142 <time> : time related to the event. It follows the HAProxy time format.
11143 <time> is expressed in milliseconds.
11144
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020011145 Example:
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011146
11147 resolvers mydns
11148 nameserver dns1 10.0.0.1:53
11149 nameserver dns2 10.0.0.2:53
11150 resolve_retries 3
11151 timeout retry 1s
11152 hold valid 10s
11153
11154
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200111556. HTTP header manipulation
11156---------------------------
11157
11158In HTTP mode, it is possible to rewrite, add or delete some of the request and
11159response headers based on regular expressions. It is also possible to block a
11160request or a response if a particular header matches a regular expression,
11161which is enough to stop most elementary protocol attacks, and to protect
Willy Tarreau70dffda2014-01-30 03:07:23 +010011162against information leak from the internal network.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011163
Willy Tarreau70dffda2014-01-30 03:07:23 +010011164If HAProxy encounters an "Informational Response" (status code 1xx), it is able
11165to process all rsp* rules which can allow, deny, rewrite or delete a header,
11166but it will refuse to add a header to any such messages as this is not
11167HTTP-compliant. The reason for still processing headers in such responses is to
11168stop and/or fix any possible information leak which may happen, for instance
11169because another downstream equipment would unconditionally add a header, or if
11170a server name appears there. When such messages are seen, normal processing
11171still occurs on the next non-informational messages.
Willy Tarreau816b9792009-09-15 21:25:21 +020011172
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011173This section covers common usage of the following keywords, described in detail
11174in section 4.2 :
11175
11176 - reqadd <string>
11177 - reqallow <search>
11178 - reqiallow <search>
11179 - reqdel <search>
11180 - reqidel <search>
11181 - reqdeny <search>
11182 - reqideny <search>
11183 - reqpass <search>
11184 - reqipass <search>
11185 - reqrep <search> <replace>
11186 - reqirep <search> <replace>
11187 - reqtarpit <search>
11188 - reqitarpit <search>
11189 - rspadd <string>
11190 - rspdel <search>
11191 - rspidel <search>
11192 - rspdeny <search>
11193 - rspideny <search>
11194 - rsprep <search> <replace>
11195 - rspirep <search> <replace>
11196
11197With all these keywords, the same conventions are used. The <search> parameter
11198is a POSIX extended regular expression (regex) which supports grouping through
11199parenthesis (without the backslash). Spaces and other delimiters must be
11200prefixed with a backslash ('\') to avoid confusion with a field delimiter.
11201Other characters may be prefixed with a backslash to change their meaning :
11202
11203 \t for a tab
11204 \r for a carriage return (CR)
11205 \n for a new line (LF)
11206 \ to mark a space and differentiate it from a delimiter
11207 \# to mark a sharp and differentiate it from a comment
11208 \\ to use a backslash in a regex
11209 \\\\ to use a backslash in the text (*2 for regex, *2 for haproxy)
11210 \xXX to write the ASCII hex code XX as in the C language
11211
11212The <replace> parameter contains the string to be used to replace the largest
11213portion of text matching the regex. It can make use of the special characters
11214above, and can reference a substring which is delimited by parenthesis in the
11215regex, by writing a backslash ('\') immediately followed by one digit from 0 to
112169 indicating the group position (0 designating the entire line). This practice
11217is very common to users of the "sed" program.
11218
11219The <string> parameter represents the string which will systematically be added
11220after the last header line. It can also use special character sequences above.
11221
11222Notes related to these keywords :
11223---------------------------------
11224 - these keywords are not always convenient to allow/deny based on header
11225 contents. It is strongly recommended to use ACLs with the "block" keyword
11226 instead, resulting in far more flexible and manageable rules.
11227
11228 - lines are always considered as a whole. It is not possible to reference
11229 a header name only or a value only. This is important because of the way
11230 headers are written (notably the number of spaces after the colon).
11231
11232 - the first line is always considered as a header, which makes it possible to
11233 rewrite or filter HTTP requests URIs or response codes, but in turn makes
11234 it harder to distinguish between headers and request line. The regex prefix
11235 ^[^\ \t]*[\ \t] matches any HTTP method followed by a space, and the prefix
11236 ^[^ \t:]*: matches any header name followed by a colon.
11237
11238 - for performances reasons, the number of characters added to a request or to
11239 a response is limited at build time to values between 1 and 4 kB. This
11240 should normally be far more than enough for most usages. If it is too short
11241 on occasional usages, it is possible to gain some space by removing some
11242 useless headers before adding new ones.
11243
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011244 - keywords beginning with "reqi" and "rspi" are the same as their counterpart
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011245 without the 'i' letter except that they ignore case when matching patterns.
11246
11247 - when a request passes through a frontend then a backend, all req* rules
11248 from the frontend will be evaluated, then all req* rules from the backend
11249 will be evaluated. The reverse path is applied to responses.
11250
11251 - req* statements are applied after "block" statements, so that "block" is
11252 always the first one, but before "use_backend" in order to permit rewriting
Willy Tarreaud72758d2010-01-12 10:42:19 +010011253 before switching.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011254
11255
Willy Tarreau74ca5042013-06-11 23:12:07 +0200112567. Using ACLs and fetching samples
11257----------------------------------
11258
11259Haproxy is capable of extracting data from request or response streams, from
11260client or server information, from tables, environmental information etc...
11261The action of extracting such data is called fetching a sample. Once retrieved,
11262these samples may be used for various purposes such as a key to a stick-table,
11263but most common usages consist in matching them against predefined constant
11264data called patterns.
11265
11266
112677.1. ACL basics
11268---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011269
11270The use of Access Control Lists (ACL) provides a flexible solution to perform
11271content switching and generally to take decisions based on content extracted
11272from the request, the response or any environmental status. The principle is
11273simple :
11274
Willy Tarreau74ca5042013-06-11 23:12:07 +020011275 - extract a data sample from a stream, table or the environment
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011276 - optionally apply some format conversion to the extracted sample
Willy Tarreau74ca5042013-06-11 23:12:07 +020011277 - apply one or multiple pattern matching methods on this sample
11278 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011279
Willy Tarreau74ca5042013-06-11 23:12:07 +020011280The actions generally consist in blocking a request, selecting a backend, or
11281adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011282
11283In order to define a test, the "acl" keyword is used. The syntax is :
11284
Willy Tarreau74ca5042013-06-11 23:12:07 +020011285 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011286
11287This creates a new ACL <aclname> or completes an existing one with new tests.
11288Those tests apply to the portion of request/response specified in <criterion>
11289and may be adjusted with optional flags [flags]. Some criteria also support
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011290an operator which may be specified before the set of values. Optionally some
11291conversion operators may be applied to the sample, and they will be specified
11292as a comma-delimited list of keywords just after the first keyword. The values
11293are of the type supported by the criterion, and are separated by spaces.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011294
11295ACL names must be formed from upper and lower case letters, digits, '-' (dash),
11296'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
11297which means that "my_acl" and "My_Acl" are two different ACLs.
11298
11299There is no enforced limit to the number of ACLs. The unused ones do not affect
11300performance, they just consume a small amount of memory.
11301
Willy Tarreau74ca5042013-06-11 23:12:07 +020011302The criterion generally is the name of a sample fetch method, or one of its ACL
11303specific declinations. The default test method is implied by the output type of
11304this sample fetch method. The ACL declinations can describe alternate matching
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011305methods of a same sample fetch method. The sample fetch methods are the only
11306ones supporting a conversion.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011307
11308Sample fetch methods return data which can be of the following types :
11309 - boolean
11310 - integer (signed or unsigned)
11311 - IPv4 or IPv6 address
11312 - string
11313 - data block
11314
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011315Converters transform any of these data into any of these. For example, some
11316converters might convert a string to a lower-case string while other ones
11317would turn a string to an IPv4 address, or apply a netmask to an IP address.
11318The resulting sample is of the type of the last converter applied to the list,
11319which defaults to the type of the sample fetch method.
11320
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011321Each sample or converter returns data of a specific type, specified with its
11322keyword in this documentation. When an ACL is declared using a standard sample
11323fetch method, certain types automatically involved a default matching method
11324which are summarized in the table below :
11325
11326 +---------------------+-----------------+
11327 | Sample or converter | Default |
11328 | output type | matching method |
11329 +---------------------+-----------------+
11330 | boolean | bool |
11331 +---------------------+-----------------+
11332 | integer | int |
11333 +---------------------+-----------------+
11334 | ip | ip |
11335 +---------------------+-----------------+
11336 | string | str |
11337 +---------------------+-----------------+
11338 | binary | none, use "-m" |
11339 +---------------------+-----------------+
11340
11341Note that in order to match a binary samples, it is mandatory to specify a
11342matching method, see below.
11343
Willy Tarreau74ca5042013-06-11 23:12:07 +020011344The ACL engine can match these types against patterns of the following types :
11345 - boolean
11346 - integer or integer range
11347 - IP address / network
11348 - string (exact, substring, suffix, prefix, subdir, domain)
11349 - regular expression
11350 - hex block
11351
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011352The following ACL flags are currently supported :
11353
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011354 -i : ignore case during matching of all subsequent patterns.
11355 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011356 -m : use a specific pattern matching method
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010011357 -n : forbid the DNS resolutions
Thierry FOURNIER9860c412014-01-29 14:23:29 +010011358 -M : load the file pointed by -f like a map file.
Thierry FOURNIER3534d882014-01-20 17:01:44 +010011359 -u : force the unique id of the ACL
Willy Tarreau6a06a402007-07-15 20:15:28 +020011360 -- : force end of flags. Useful when a string looks like one of the flags.
11361
Willy Tarreau74ca5042013-06-11 23:12:07 +020011362The "-f" flag is followed by the name of a file from which all lines will be
11363read as individual values. It is even possible to pass multiple "-f" arguments
11364if the patterns are to be loaded from multiple files. Empty lines as well as
11365lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
11366will be stripped. If it is absolutely necessary to insert a valid pattern
11367beginning with a sharp, just prefix it with a space so that it is not taken for
11368a comment. Depending on the data type and match method, haproxy may load the
11369lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
11370exact string matching. In this case, duplicates will automatically be removed.
11371
Thierry FOURNIER9860c412014-01-29 14:23:29 +010011372The "-M" flag allows an ACL to use a map file. If this flag is set, the file is
11373parsed as two column file. The first column contains the patterns used by the
11374ACL, and the second column contain the samples. The sample can be used later by
11375a map. This can be useful in some rare cases where an ACL would just be used to
11376check for the existence of a pattern in a map before a mapping is applied.
11377
Thierry FOURNIER3534d882014-01-20 17:01:44 +010011378The "-u" flag forces the unique id of the ACL. This unique id is used with the
11379socket interface to identify ACL and dynamically change its values. Note that a
11380file is always identified by its name even if an id is set.
11381
Willy Tarreau74ca5042013-06-11 23:12:07 +020011382Also, note that the "-i" flag applies to subsequent entries and not to entries
11383loaded from files preceding it. For instance :
11384
11385 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
11386
11387In this example, each line of "exact-ua.lst" will be exactly matched against
11388the "user-agent" header of the request. Then each line of "generic-ua" will be
11389case-insensitively matched. Then the word "test" will be insensitively matched
11390as well.
11391
11392The "-m" flag is used to select a specific pattern matching method on the input
11393sample. All ACL-specific criteria imply a pattern matching method and generally
11394do not need this flag. However, this flag is useful with generic sample fetch
11395methods to describe how they're going to be matched against the patterns. This
11396is required for sample fetches which return data type for which there is no
11397obvious matching method (eg: string or binary). When "-m" is specified and
11398followed by a pattern matching method name, this method is used instead of the
11399default one for the criterion. This makes it possible to match contents in ways
11400that were not initially planned, or with sample fetch methods which return a
11401string. The matching method also affects the way the patterns are parsed.
11402
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010011403The "-n" flag forbids the dns resolutions. It is used with the load of ip files.
11404By default, if the parser cannot parse ip address it considers that the parsed
11405string is maybe a domain name and try dns resolution. The flag "-n" disable this
11406resolution. It is useful for detecting malformed ip lists. Note that if the DNS
11407server is not reachable, the haproxy configuration parsing may last many minutes
11408waiting fir the timeout. During this time no error messages are displayed. The
11409flag "-n" disable this behavior. Note also that during the runtime, this
11410function is disabled for the dynamic acl modifications.
11411
Willy Tarreau74ca5042013-06-11 23:12:07 +020011412There are some restrictions however. Not all methods can be used with all
11413sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
11414be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +020011415
11416 - "found" : only check if the requested sample could be found in the stream,
11417 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +020011418 to pass any pattern to avoid confusion. This matching method is
11419 particularly useful to detect presence of certain contents such
11420 as headers, cookies, etc... even if they are empty and without
11421 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011422
11423 - "bool" : check the value as a boolean. It can only be applied to fetches
11424 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011425 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011426
11427 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +020011428 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011429
11430 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +020011431 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011432
11433 - "bin" : match the contents against an hexadecimal string representing a
11434 binary sequence. This may be used with binary or string samples.
11435
11436 - "len" : match the sample's length as an integer. This may be used with
11437 binary or string samples.
11438
Willy Tarreau74ca5042013-06-11 23:12:07 +020011439 - "str" : exact match : match the contents against a string. This may be
11440 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011441
Willy Tarreau74ca5042013-06-11 23:12:07 +020011442 - "sub" : substring match : check that the contents contain at least one of
11443 the provided string patterns. This may be used with binary or
11444 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011445
Willy Tarreau74ca5042013-06-11 23:12:07 +020011446 - "reg" : regex match : match the contents against a list of regular
11447 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011448
Willy Tarreau74ca5042013-06-11 23:12:07 +020011449 - "beg" : prefix match : check that the contents begin like the provided
11450 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011451
Willy Tarreau74ca5042013-06-11 23:12:07 +020011452 - "end" : suffix match : check that the contents end like the provided
11453 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011454
Willy Tarreau74ca5042013-06-11 23:12:07 +020011455 - "dir" : subdir match : check that a slash-delimited portion of the
11456 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011457 This may be used with binary or string samples.
11458
Willy Tarreau74ca5042013-06-11 23:12:07 +020011459 - "dom" : domain match : check that a dot-delimited portion of the contents
11460 exactly match one of the provided string patterns. This may be
11461 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011462
11463For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
11464request, it is possible to do :
11465
11466 acl jsess_present cook(JSESSIONID) -m found
11467
11468In order to apply a regular expression on the 500 first bytes of data in the
11469buffer, one would use the following acl :
11470
11471 acl script_tag payload(0,500) -m reg -i <script>
11472
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011473On systems where the regex library is much slower when using "-i", it is
11474possible to convert the sample to lowercase before matching, like this :
11475
11476 acl script_tag payload(0,500),lower -m reg <script>
11477
Willy Tarreau74ca5042013-06-11 23:12:07 +020011478All ACL-specific criteria imply a default matching method. Most often, these
11479criteria are composed by concatenating the name of the original sample fetch
11480method and the matching method. For example, "hdr_beg" applies the "beg" match
11481to samples retrieved using the "hdr" fetch method. Since all ACL-specific
11482criteria rely on a sample fetch method, it is always possible instead to use
11483the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011484
Willy Tarreau74ca5042013-06-11 23:12:07 +020011485If an alternate match is specified using "-m" on an ACL-specific criterion,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011486the matching method is simply applied to the underlying sample fetch method.
11487For example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011488
Willy Tarreau74ca5042013-06-11 23:12:07 +020011489 acl short_form hdr_beg(host) www.
11490 acl alternate1 hdr_beg(host) -m beg www.
11491 acl alternate2 hdr_dom(host) -m beg www.
11492 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011493
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011494
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011495The table below summarizes the compatibility matrix between sample or converter
11496types and the pattern types to fetch against. It indicates for each compatible
11497combination the name of the matching method to be used, surrounded with angle
11498brackets ">" and "<" when the method is the default one and will work by
11499default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +010011500
Willy Tarreau74ca5042013-06-11 23:12:07 +020011501 +-------------------------------------------------+
11502 | Input sample type |
11503 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011504 | pattern type | boolean | integer | ip | string | binary |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011505 +----------------------+---------+---------+---------+---------+---------+
11506 | none (presence only) | found | found | found | found | found |
11507 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011508 | none (boolean value) |> bool <| bool | | bool | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011509 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011510 | integer (value) | int |> int <| int | int | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011511 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011512 | integer (length) | len | len | len | len | len |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011513 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011514 | IP address | | |> ip <| ip | ip |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011515 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011516 | exact string | str | str | str |> str <| str |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011517 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011518 | prefix | beg | beg | beg | beg | beg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011519 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011520 | suffix | end | end | end | end | end |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011521 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011522 | substring | sub | sub | sub | sub | sub |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011523 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011524 | subdir | dir | dir | dir | dir | dir |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011525 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011526 | domain | dom | dom | dom | dom | dom |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011527 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011528 | regex | reg | reg | reg | reg | reg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011529 +----------------------+---------+---------+---------+---------+---------+
11530 | hex block | | | | bin | bin |
11531 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +020011532
11533
Willy Tarreau74ca5042013-06-11 23:12:07 +0200115347.1.1. Matching booleans
11535------------------------
11536
11537In order to match a boolean, no value is needed and all values are ignored.
11538Boolean matching is used by default for all fetch methods of type "boolean".
11539When boolean matching is used, the fetched value is returned as-is, which means
11540that a boolean "true" will always match and a boolean "false" will never match.
11541
11542Boolean matching may also be enforced using "-m bool" on fetch methods which
11543return an integer value. Then, integer value 0 is converted to the boolean
11544"false" and all other values are converted to "true".
11545
Willy Tarreau6a06a402007-07-15 20:15:28 +020011546
Willy Tarreau74ca5042013-06-11 23:12:07 +0200115477.1.2. Matching integers
11548------------------------
11549
11550Integer matching applies by default to integer fetch methods. It can also be
11551enforced on boolean fetches using "-m int". In this case, "false" is converted
11552to the integer 0, and "true" is converted to the integer 1.
11553
11554Integer matching also supports integer ranges and operators. Note that integer
11555matching only applies to positive values. A range is a value expressed with a
11556lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011557
11558For instance, "1024:65535" is a valid range to represent a range of
11559unprivileged ports, and "1024:" would also work. "0:1023" is a valid
11560representation of privileged ports, and ":1023" would also work.
11561
Willy Tarreau62644772008-07-16 18:36:06 +020011562As a special case, some ACL functions support decimal numbers which are in fact
11563two integers separated by a dot. This is used with some version checks for
11564instance. All integer properties apply to those decimal numbers, including
11565ranges and operators.
11566
Willy Tarreau6a06a402007-07-15 20:15:28 +020011567For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +010011568operators with ranges does not make much sense and is strongly discouraged.
11569Similarly, it does not make much sense to perform order comparisons with a set
11570of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011571
Willy Tarreau0ba27502007-12-24 16:55:16 +010011572Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +020011573
11574 eq : true if the tested value equals at least one value
11575 ge : true if the tested value is greater than or equal to at least one value
11576 gt : true if the tested value is greater than at least one value
11577 le : true if the tested value is less than or equal to at least one value
11578 lt : true if the tested value is less than at least one value
11579
Willy Tarreau0ba27502007-12-24 16:55:16 +010011580For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +020011581
11582 acl negative-length hdr_val(content-length) lt 0
11583
Willy Tarreau62644772008-07-16 18:36:06 +020011584This one matches SSL versions between 3.0 and 3.1 (inclusive) :
11585
11586 acl sslv3 req_ssl_ver 3:3.1
11587
Willy Tarreau6a06a402007-07-15 20:15:28 +020011588
Willy Tarreau74ca5042013-06-11 23:12:07 +0200115897.1.3. Matching strings
11590-----------------------
11591
11592String matching applies to string or binary fetch methods, and exists in 6
11593different forms :
11594
11595 - exact match (-m str) : the extracted string must exactly match the
11596 patterns ;
11597
11598 - substring match (-m sub) : the patterns are looked up inside the
11599 extracted string, and the ACL matches if any of them is found inside ;
11600
11601 - prefix match (-m beg) : the patterns are compared with the beginning of
11602 the extracted string, and the ACL matches if any of them matches.
11603
11604 - suffix match (-m end) : the patterns are compared with the end of the
11605 extracted string, and the ACL matches if any of them matches.
11606
Baptiste Assmann33db6002016-03-06 23:32:10 +010011607 - subdir match (-m dir) : the patterns are looked up inside the extracted
Willy Tarreau74ca5042013-06-11 23:12:07 +020011608 string, delimited with slashes ("/"), and the ACL matches if any of them
11609 matches.
11610
11611 - domain match (-m dom) : the patterns are looked up inside the extracted
11612 string, delimited with dots ("."), and the ACL matches if any of them
11613 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011614
11615String matching applies to verbatim strings as they are passed, with the
11616exception of the backslash ("\") which makes it possible to escape some
11617characters such as the space. If the "-i" flag is passed before the first
11618string, then the matching will be performed ignoring the case. In order
11619to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +010011620before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020011621
11622
Willy Tarreau74ca5042013-06-11 23:12:07 +0200116237.1.4. Matching regular expressions (regexes)
11624---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020011625
11626Just like with string matching, regex matching applies to verbatim strings as
11627they are passed, with the exception of the backslash ("\") which makes it
11628possible to escape some characters such as the space. If the "-i" flag is
11629passed before the first regex, then the matching will be performed ignoring
11630the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +010011631the "--" flag before the first string. Same principle applies of course to
11632match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020011633
11634
Willy Tarreau74ca5042013-06-11 23:12:07 +0200116357.1.5. Matching arbitrary data blocks
11636-------------------------------------
11637
11638It is possible to match some extracted samples against a binary block which may
11639not safely be represented as a string. For this, the patterns must be passed as
11640a series of hexadecimal digits in an even number, when the match method is set
11641to binary. Each sequence of two digits will represent a byte. The hexadecimal
11642digits may be used upper or lower case.
11643
11644Example :
11645 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
11646 acl hello payload(0,6) -m bin 48656c6c6f0a
11647
11648
116497.1.6. Matching IPv4 and IPv6 addresses
11650---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020011651
11652IPv4 addresses values can be specified either as plain addresses or with a
11653netmask appended, in which case the IPv4 address matches whenever it is
11654within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010011655host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +010011656difficult to read and debug configurations. If hostnames are used, you should
11657at least ensure that they are present in /etc/hosts so that the configuration
11658does not depend on any random DNS match at the moment the configuration is
11659parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011660
Daniel Schnellereba56342016-04-13 00:26:52 +020011661The dotted IPv4 address notation is supported in both regular as well as the
11662abbreviated form with all-0-octets omitted:
11663
11664 +------------------+------------------+------------------+
11665 | Example 1 | Example 2 | Example 3 |
11666 +------------------+------------------+------------------+
11667 | 192.168.0.1 | 10.0.0.12 | 127.0.0.1 |
11668 | 192.168.1 | 10.12 | 127.1 |
11669 | 192.168.0.1/22 | 10.0.0.12/8 | 127.0.0.1/8 |
11670 | 192.168.1/22 | 10.12/8 | 127.1/8 |
11671 +------------------+------------------+------------------+
11672
11673Notice that this is different from RFC 4632 CIDR address notation in which
11674192.168.42/24 would be equivalent to 192.168.42.0/24.
11675
Willy Tarreauceb4ac92012-04-28 00:41:46 +020011676IPv6 may be entered in their usual form, with or without a netmask appended.
11677Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
11678trouble with randomly resolved IP addresses, host names are never allowed in
11679IPv6 patterns.
11680
11681HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
11682following situations :
11683 - tested address is IPv4, pattern address is IPv4, the match applies
11684 in IPv4 using the supplied mask if any.
11685 - tested address is IPv6, pattern address is IPv6, the match applies
11686 in IPv6 using the supplied mask if any.
11687 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
11688 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
11689 ::IPV4 or ::ffff:IPV4, otherwise it fails.
11690 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
11691 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
11692 applied in IPv6 using the supplied IPv6 mask.
11693
Willy Tarreau74ca5042013-06-11 23:12:07 +020011694
116957.2. Using ACLs to form conditions
11696----------------------------------
11697
11698Some actions are only performed upon a valid condition. A condition is a
11699combination of ACLs with operators. 3 operators are supported :
11700
11701 - AND (implicit)
11702 - OR (explicit with the "or" keyword or the "||" operator)
11703 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +020011704
Willy Tarreau74ca5042013-06-11 23:12:07 +020011705A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +020011706
Willy Tarreau74ca5042013-06-11 23:12:07 +020011707 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +020011708
Willy Tarreau74ca5042013-06-11 23:12:07 +020011709Such conditions are generally used after an "if" or "unless" statement,
11710indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +020011711
Willy Tarreau74ca5042013-06-11 23:12:07 +020011712For instance, to block HTTP requests to the "*" URL with methods other than
11713"OPTIONS", as well as POST requests without content-length, and GET or HEAD
11714requests with a content-length greater than 0, and finally every request which
11715is not either GET/HEAD/POST/OPTIONS !
11716
11717 acl missing_cl hdr_cnt(Content-length) eq 0
11718 block if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
11719 block if METH_GET HTTP_CONTENT
11720 block unless METH_GET or METH_POST or METH_OPTIONS
11721
11722To select a different backend for requests to static contents on the "www" site
11723and to every request on the "img", "video", "download" and "ftp" hosts :
11724
11725 acl url_static path_beg /static /images /img /css
11726 acl url_static path_end .gif .png .jpg .css .js
11727 acl host_www hdr_beg(host) -i www
11728 acl host_static hdr_beg(host) -i img. video. download. ftp.
11729
11730 # now use backend "static" for all static-only hosts, and for static urls
11731 # of host "www". Use backend "www" for the rest.
11732 use_backend static if host_static or host_www url_static
11733 use_backend www if host_www
11734
11735It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
11736expressions that are built on the fly without needing to be declared. They must
11737be enclosed between braces, with a space before and after each brace (because
11738the braces must be seen as independent words). Example :
11739
11740 The following rule :
11741
11742 acl missing_cl hdr_cnt(Content-length) eq 0
11743 block if METH_POST missing_cl
11744
11745 Can also be written that way :
11746
11747 block if METH_POST { hdr_cnt(Content-length) eq 0 }
11748
11749It is generally not recommended to use this construct because it's a lot easier
11750to leave errors in the configuration when written that way. However, for very
11751simple rules matching only one source IP address for instance, it can make more
11752sense to use them than to declare ACLs with random names. Another example of
11753good use is the following :
11754
11755 With named ACLs :
11756
11757 acl site_dead nbsrv(dynamic) lt 2
11758 acl site_dead nbsrv(static) lt 2
11759 monitor fail if site_dead
11760
11761 With anonymous ACLs :
11762
11763 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
11764
11765See section 4.2 for detailed help on the "block" and "use_backend" keywords.
11766
11767
117687.3. Fetching samples
11769---------------------
11770
11771Historically, sample fetch methods were only used to retrieve data to match
11772against patterns using ACLs. With the arrival of stick-tables, a new class of
11773sample fetch methods was created, most often sharing the same syntax as their
11774ACL counterpart. These sample fetch methods are also known as "fetches". As
11775of now, ACLs and fetches have converged. All ACL fetch methods have been made
11776available as fetch methods, and ACLs may use any sample fetch method as well.
11777
11778This section details all available sample fetch methods and their output type.
11779Some sample fetch methods have deprecated aliases that are used to maintain
11780compatibility with existing configurations. They are then explicitly marked as
11781deprecated and should not be used in new setups.
11782
11783The ACL derivatives are also indicated when available, with their respective
11784matching methods. These ones all have a well defined default pattern matching
11785method, so it is never necessary (though allowed) to pass the "-m" option to
11786indicate how the sample will be matched using ACLs.
11787
11788As indicated in the sample type versus matching compatibility matrix above,
11789when using a generic sample fetch method in an ACL, the "-m" option is
11790mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
11791the same keyword exists as an ACL keyword and as a standard fetch method, the
11792ACL engine will automatically pick the ACL-only one by default.
11793
11794Some of these keywords support one or multiple mandatory arguments, and one or
11795multiple optional arguments. These arguments are strongly typed and are checked
11796when the configuration is parsed so that there is no risk of running with an
11797incorrect argument (eg: an unresolved backend name). Fetch function arguments
11798are passed between parenthesis and are delimited by commas. When an argument
11799is optional, it will be indicated below between square brackets ('[ ]'). When
11800all arguments are optional, the parenthesis may be omitted.
11801
11802Thus, the syntax of a standard sample fetch method is one of the following :
11803 - name
11804 - name(arg1)
11805 - name(arg1,arg2)
11806
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011807
118087.3.1. Converters
11809-----------------
11810
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011811Sample fetch methods may be combined with transformations to be applied on top
11812of the fetched sample (also called "converters"). These combinations form what
11813is called "sample expressions" and the result is a "sample". Initially this
11814was only supported by "stick on" and "stick store-request" directives but this
11815has now be extended to all places where samples may be used (acls, log-format,
11816unique-id-format, add-header, ...).
11817
11818These transformations are enumerated as a series of specific keywords after the
11819sample fetch method. These keywords may equally be appended immediately after
11820the fetch keyword's argument, delimited by a comma. These keywords can also
11821support some arguments (eg: a netmask) which must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +010011822
Willy Tarreau97707872015-01-27 15:12:13 +010011823A certain category of converters are bitwise and arithmetic operators which
11824support performing basic operations on integers. Some bitwise operations are
11825supported (and, or, xor, cpl) and some arithmetic operations are supported
11826(add, sub, mul, div, mod, neg). Some comparators are provided (odd, even, not,
11827bool) which make it possible to report a match without having to write an ACL.
11828
Willy Tarreau74ca5042013-06-11 23:12:07 +020011829The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +010011830
Willy Tarreau97707872015-01-27 15:12:13 +010011831add(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011832 Adds <value> to the input value of type signed integer, and returns the
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011833 result as a signed integer. <value> can be a numeric value or a variable
Daniel Schneller0b547052016-03-21 20:46:57 +010011834 name. The name of the variable starts with an indication about its scope. The
11835 scopes allowed are:
11836 "sess" : the variable is shared with the whole session
11837 "txn" : the variable is shared with the transaction (request and response)
11838 "req" : the variable is shared only during request processing
11839 "res" : the variable is shared only during response processing
11840 This prefix is followed by a name. The separator is a '.'. The name may only
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011841 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011842
11843and(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011844 Performs a bitwise "AND" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011845 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010011846 numeric value or a variable name. The name of the variable starts with an
11847 indication about its scope. The scopes allowed are:
11848 "sess" : the variable is shared with the whole session
11849 "txn" : the variable is shared with the transaction (request and response)
11850 "req" : the variable is shared only during request processing
11851 "res" : the variable is shared only during response processing
11852 This prefix is followed by a name. The separator is a '.'. The name may only
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011853 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011854
Emeric Brun53d1a982014-04-30 18:21:37 +020011855base64
11856 Converts a binary input sample to a base64 string. It is used to log or
11857 transfer binary content in a way that can be reliably transferred (eg:
11858 an SSL ID can be copied in a header).
11859
Willy Tarreau97707872015-01-27 15:12:13 +010011860bool
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011861 Returns a boolean TRUE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010011862 non-null, otherwise returns FALSE. Used in conjunction with and(), it can be
11863 used to report true/false for bit testing on input values (eg: verify the
11864 presence of a flag).
11865
Emeric Brun54c4ac82014-11-03 15:32:43 +010011866bytes(<offset>[,<length>])
11867 Extracts some bytes from an input binary sample. The result is a binary
11868 sample starting at an offset (in bytes) of the original sample and
11869 optionnaly truncated at the given length.
11870
Willy Tarreau97707872015-01-27 15:12:13 +010011871cpl
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011872 Takes the input value of type signed integer, applies a ones-complement
11873 (flips all bits) and returns the result as an signed integer.
Willy Tarreau97707872015-01-27 15:12:13 +010011874
Willy Tarreau80599772015-01-20 19:35:24 +010011875crc32([<avalanche>])
11876 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32
11877 hash function. Optionally, it is possible to apply a full avalanche hash
11878 function to the output if the optional <avalanche> argument equals 1. This
11879 converter uses the same functions as used by the various hash-based load
11880 balancing algorithms, so it will provide exactly the same results. It is
11881 provided for compatibility with other software which want a CRC32 to be
11882 computed on some input keys, so it follows the most common implementation as
11883 found in Ethernet, Gzip, PNG, etc... It is slower than the other algorithms
11884 but may provide a better or at least less predictable distribution. It must
11885 not be used for security purposes as a 32-bit hash is trivial to break. See
11886 also "djb2", "sdbm", "wt6" and the "hash-type" directive.
11887
David Carlier29b3ca32015-09-25 14:09:21 +010011888da-csv-conv(<prop>[,<prop>*])
David Carlier4542b102015-06-01 13:54:29 +020011889 Asks the DeviceAtlas converter to identify the User Agent string passed on
11890 input, and to emit a string made of the concatenation of the properties
11891 enumerated in argument, delimited by the separator defined by the global
11892 keyword "deviceatlas-property-separator", or by default the pipe character
David Carlier840b0242016-03-16 10:09:55 +000011893 ('|'). There's a limit of 12 different properties imposed by the haproxy
David Carlier4542b102015-06-01 13:54:29 +020011894 configuration language.
11895
11896 Example:
11897 frontend www
Cyril Bonté307ee1e2015-09-28 23:16:06 +020011898 bind *:8881
11899 default_backend servers
David Carlier840b0242016-03-16 10:09:55 +000011900 http-request set-header X-DeviceAtlas-Data %[req.fhdr(User-Agent),da-csv(primaryHardwareType,osName,osVersion,browserName,browserVersion,browserRenderingEngine)]
David Carlier4542b102015-06-01 13:54:29 +020011901
Thierry FOURNIER9687c772015-05-07 15:46:29 +020011902debug
11903 This converter is used as debug tool. It dumps on screen the content and the
11904 type of the input sample. The sample is returned as is on its output. This
11905 converter only exists when haproxy was built with debugging enabled.
11906
Willy Tarreau97707872015-01-27 15:12:13 +010011907div(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011908 Divides the input value of type signed integer by <value>, and returns the
11909 result as an signed integer. If <value> is null, the largest unsigned
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011910 integer is returned (typically 2^63-1). <value> can be a numeric value or a
Daniel Schneller0b547052016-03-21 20:46:57 +010011911 variable name. The name of the variable starts with an indication about its
11912 scope. The scopes allowed are:
11913 "sess" : the variable is shared with the whole session
11914 "txn" : the variable is shared with the transaction (request and response)
11915 "req" : the variable is shared only during request processing
11916 "res" : the variable is shared only during response processing
11917 This prefix is followed by a name. The separator is a '.'. The name may only
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011918 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011919
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020011920djb2([<avalanche>])
11921 Hashes a binary input sample into an unsigned 32-bit quantity using the DJB2
11922 hash function. Optionally, it is possible to apply a full avalanche hash
11923 function to the output if the optional <avalanche> argument equals 1. This
11924 converter uses the same functions as used by the various hash-based load
11925 balancing algorithms, so it will provide exactly the same results. It is
11926 mostly intended for debugging, but can be used as a stick-table entry to
11927 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010011928 32-bit hash is trivial to break. See also "crc32", "sdbm", "wt6" and the
11929 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020011930
Willy Tarreau97707872015-01-27 15:12:13 +010011931even
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011932 Returns a boolean TRUE if the input value of type signed integer is even
Willy Tarreau97707872015-01-27 15:12:13 +010011933 otherwise returns FALSE. It is functionally equivalent to "not,and(1),bool".
11934
Emeric Brunf399b0d2014-11-03 17:07:03 +010011935field(<index>,<delimiters>)
11936 Extracts the substring at the given index considering given delimiters from
11937 an input string. Indexes start at 1 and delimiters are a string formatted
11938 list of chars.
11939
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011940hex
11941 Converts a binary input sample to an hex string containing two hex digits per
11942 input byte. It is used to log or transfer hex dumps of some binary input data
11943 in a way that can be reliably transferred (eg: an SSL ID can be copied in a
11944 header).
Thierry FOURNIER2f49d6d2014-03-12 15:01:52 +010011945
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011946http_date([<offset>])
11947 Converts an integer supposed to contain a date since epoch to a string
11948 representing this date in a format suitable for use in HTTP header fields. If
11949 an offset value is specified, then it is a number of seconds that is added to
11950 the date before the conversion is operated. This is particularly useful to
11951 emit Date header fields, Expires values in responses when combined with a
11952 positive offset, or Last-Modified values when the offset is negative.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011953
Willy Tarreaud9f316a2014-07-10 14:03:38 +020011954in_table(<table>)
11955 Uses the string representation of the input sample to perform a look up in
11956 the specified table. If the key is not found in the table, a boolean false
11957 is returned. Otherwise a boolean true is returned. This can be used to verify
11958 the presence of a certain key in a table tracking some elements (eg: whether
11959 or not a source IP address or an Authorization header was already seen).
11960
Willy Tarreauffcb2e42014-07-10 16:29:08 +020011961ipmask(<mask>)
11962 Apply a mask to an IPv4 address, and use the result for lookups and storage.
11963 This can be used to make all hosts within a certain mask to share the same
11964 table entries and as such use the same server. The mask can be passed in
11965 dotted form (eg: 255.255.255.0) or in CIDR form (eg: 24).
11966
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020011967json([<input-code>])
11968 Escapes the input string and produces an ASCII ouput string ready to use as a
11969 JSON string. The converter tries to decode the input string according to the
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020011970 <input-code> parameter. It can be "ascii", "utf8", "utf8s", "utf8p" or
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020011971 "utf8ps". The "ascii" decoder never fails. The "utf8" decoder detects 3 types
11972 of errors:
11973 - bad UTF-8 sequence (lone continuation byte, bad number of continuation
11974 bytes, ...)
11975 - invalid range (the decoded value is within a UTF-8 prohibited range),
11976 - code overlong (the value is encoded with more bytes than necessary).
11977
11978 The UTF-8 JSON encoding can produce a "too long value" error when the UTF-8
11979 character is greater than 0xffff because the JSON string escape specification
11980 only authorizes 4 hex digits for the value encoding. The UTF-8 decoder exists
11981 in 4 variants designated by a combination of two suffix letters : "p" for
11982 "permissive" and "s" for "silently ignore". The behaviors of the decoders
11983 are :
11984 - "ascii" : never fails ;
11985 - "utf8" : fails on any detected errors ;
11986 - "utf8s" : never fails, but removes characters corresponding to errors ;
11987 - "utf8p" : accepts and fixes the overlong errors, but fails on any other
11988 error ;
11989 - "utf8ps" : never fails, accepts and fixes the overlong errors, but removes
11990 characters corresponding to the other errors.
11991
11992 This converter is particularly useful for building properly escaped JSON for
11993 logging to servers which consume JSON-formated traffic logs.
11994
11995 Example:
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020011996 capture request header Host len 15
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020011997 capture request header user-agent len 150
11998 log-format '{"ip":"%[src]","user-agent":"%[capture.req.hdr(1),json(utf8s)]"}'
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020011999
12000 Input request from client 127.0.0.1:
12001 GET / HTTP/1.0
12002 User-Agent: Very "Ugly" UA 1/2
12003
12004 Output log:
12005 {"ip":"127.0.0.1","user-agent":"Very \"Ugly\" UA 1\/2"}
12006
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012007language(<value>[,<default>])
12008 Returns the value with the highest q-factor from a list as extracted from the
12009 "accept-language" header using "req.fhdr". Values with no q-factor have a
12010 q-factor of 1. Values with a q-factor of 0 are dropped. Only values which
12011 belong to the list of semi-colon delimited <values> will be considered. The
12012 argument <value> syntax is "lang[;lang[;lang[;...]]]". If no value matches the
12013 given list and a default value is provided, it is returned. Note that language
12014 names may have a variant after a dash ('-'). If this variant is present in the
12015 list, it will be matched, but if it is not, only the base language is checked.
12016 The match is case-sensitive, and the output string is always one of those
12017 provided in arguments. The ordering of arguments is meaningless, only the
12018 ordering of the values in the request counts, as the first value among
12019 multiple sharing the same q-factor is used.
Thierry FOURNIERad903512014-04-11 17:51:01 +020012020
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012021 Example :
Thierry FOURNIERad903512014-04-11 17:51:01 +020012022
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012023 # this configuration switches to the backend matching a
12024 # given language based on the request :
Thierry FOURNIERad903512014-04-11 17:51:01 +020012025
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012026 acl es req.fhdr(accept-language),language(es;fr;en) -m str es
12027 acl fr req.fhdr(accept-language),language(es;fr;en) -m str fr
12028 acl en req.fhdr(accept-language),language(es;fr;en) -m str en
12029 use_backend spanish if es
12030 use_backend french if fr
12031 use_backend english if en
12032 default_backend choose_your_language
Thierry FOURNIERad903512014-04-11 17:51:01 +020012033
Willy Tarreauffcb2e42014-07-10 16:29:08 +020012034lower
12035 Convert a string sample to lower case. This can only be placed after a string
12036 sample fetch function or after a transformation keyword returning a string
12037 type. The result is of type string.
12038
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020012039ltime(<format>[,<offset>])
12040 Converts an integer supposed to contain a date since epoch to a string
12041 representing this date in local time using a format defined by the <format>
12042 string using strftime(3). The purpose is to allow any date format to be used
12043 in logs. An optional <offset> in seconds may be applied to the input date
12044 (positive or negative). See the strftime() man page for the format supported
12045 by your operating system. See also the utime converter.
12046
12047 Example :
12048
12049 # Emit two colons, one with the local time and another with ip:port
12050 # Eg: 20140710162350 127.0.0.1:57325
12051 log-format %[date,ltime(%Y%m%d%H%M%S)]\ %ci:%cp
12052
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012053map(<map_file>[,<default_value>])
12054map_<match_type>(<map_file>[,<default_value>])
12055map_<match_type>_<output_type>(<map_file>[,<default_value>])
12056 Search the input value from <map_file> using the <match_type> matching method,
12057 and return the associated value converted to the type <output_type>. If the
12058 input value cannot be found in the <map_file>, the converter returns the
12059 <default_value>. If the <default_value> is not set, the converter fails and
12060 acts as if no input value could be fetched. If the <match_type> is not set, it
12061 defaults to "str". Likewise, if the <output_type> is not set, it defaults to
12062 "str". For convenience, the "map" keyword is an alias for "map_str" and maps a
12063 string to another string.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012064
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012065 It is important to avoid overlapping between the keys : IP addresses and
12066 strings are stored in trees, so the first of the finest match will be used.
12067 Other keys are stored in lists, so the first matching occurrence will be used.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012068
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012069 The following array contains the list of all map functions avalaible sorted by
12070 input type, match type and output type.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012071
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012072 input type | match method | output type str | output type int | output type ip
12073 -----------+--------------+-----------------+-----------------+---------------
12074 str | str | map_str | map_str_int | map_str_ip
12075 -----------+--------------+-----------------+-----------------+---------------
Willy Tarreau787a4c02014-05-10 07:55:30 +020012076 str | beg | map_beg | map_beg_int | map_end_ip
12077 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012078 str | sub | map_sub | map_sub_int | map_sub_ip
12079 -----------+--------------+-----------------+-----------------+---------------
12080 str | dir | map_dir | map_dir_int | map_dir_ip
12081 -----------+--------------+-----------------+-----------------+---------------
12082 str | dom | map_dom | map_dom_int | map_dom_ip
12083 -----------+--------------+-----------------+-----------------+---------------
12084 str | end | map_end | map_end_int | map_end_ip
12085 -----------+--------------+-----------------+-----------------+---------------
Thierry Fournier8feaa662016-02-10 22:55:20 +010012086 | | map_reg | |
12087 str | reg +-----------------+ map_reg_int | map_reg_ip
12088 | | map_regm | |
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012089 -----------+--------------+-----------------+-----------------+---------------
12090 int | int | map_int | map_int_int | map_int_ip
12091 -----------+--------------+-----------------+-----------------+---------------
12092 ip | ip | map_ip | map_ip_int | map_ip_ip
12093 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012094
Thierry Fournier8feaa662016-02-10 22:55:20 +010012095 The special map called "map_regm" expect matching zone in the regular
12096 expression and modify the output replacing back reference (like "\1") by
12097 the corresponding match text.
12098
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012099 The file contains one key + value per line. Lines which start with '#' are
12100 ignored, just like empty lines. Leading tabs and spaces are stripped. The key
12101 is then the first "word" (series of non-space/tabs characters), and the value
12102 is what follows this series of space/tab till the end of the line excluding
12103 trailing spaces/tabs.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012104
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012105 Example :
12106
12107 # this is a comment and is ignored
12108 2.22.246.0/23 United Kingdom \n
12109 <-><-----------><--><------------><---->
12110 | | | | `- trailing spaces ignored
12111 | | | `---------- value
12112 | | `-------------------- middle spaces ignored
12113 | `---------------------------- key
12114 `------------------------------------ leading spaces ignored
12115
Willy Tarreau97707872015-01-27 15:12:13 +010012116mod(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012117 Divides the input value of type signed integer by <value>, and returns the
12118 remainder as an signed integer. If <value> is null, then zero is returned.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012119 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010012120 starts with an indication about its scope. The scopes allowed are:
12121 "sess" : the variable is shared with the whole session
12122 "txn" : the variable is shared with the transaction (request and response)
12123 "req" : the variable is shared only during request processing
12124 "res" : the variable is shared only during response processing
12125 This prefix is followed by a name. The separator is a '.'. The name may only
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012126 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012127
12128mul(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012129 Multiplies the input value of type signed integer by <value>, and returns
Thierry FOURNIER00c005c2015-07-08 01:10:21 +020012130 the product as an signed integer. In case of overflow, the largest possible
12131 value for the sign is returned so that the operation doesn't wrap around.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012132 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010012133 starts with an indication about its scope. The scopes allowed are:
12134 "sess" : the variable is shared with the whole session
12135 "txn" : the variable is shared with the transaction (request and response)
12136 "req" : the variable is shared only during request processing
12137 "res" : the variable is shared only during response processing
12138 This prefix is followed by a name. The separator is a '.'. The name may only
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012139 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012140
12141neg
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012142 Takes the input value of type signed integer, computes the opposite value,
12143 and returns the remainder as an signed integer. 0 is identity. This operator
12144 is provided for reversed subtracts : in order to subtract the input from a
12145 constant, simply perform a "neg,add(value)".
Willy Tarreau97707872015-01-27 15:12:13 +010012146
12147not
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012148 Returns a boolean FALSE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010012149 non-null, otherwise returns TRUE. Used in conjunction with and(), it can be
12150 used to report true/false for bit testing on input values (eg: verify the
12151 absence of a flag).
12152
12153odd
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012154 Returns a boolean TRUE if the input value of type signed integer is odd
Willy Tarreau97707872015-01-27 15:12:13 +010012155 otherwise returns FALSE. It is functionally equivalent to "and(1),bool".
12156
12157or(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012158 Performs a bitwise "OR" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012159 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010012160 numeric value or a variable name. The name of the variable starts with an
12161 indication about its scope. The scopes allowed are:
12162 "sess" : the variable is shared with the whole session
12163 "txn" : the variable is shared with the transaction (request and response)
12164 "req" : the variable is shared only during request processing
12165 "res" : the variable is shared only during response processing
12166 This prefix is followed by a name. The separator is a '.'. The name may only
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012167 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012168
Willy Tarreauc4dc3502015-01-23 20:39:28 +010012169regsub(<regex>,<subst>[,<flags>])
Willy Tarreau7eda8492015-01-20 19:47:06 +010012170 Applies a regex-based substitution to the input string. It does the same
12171 operation as the well-known "sed" utility with "s/<regex>/<subst>/". By
12172 default it will replace in the input string the first occurrence of the
12173 largest part matching the regular expression <regex> with the substitution
12174 string <subst>. It is possible to replace all occurrences instead by adding
12175 the flag "g" in the third argument <flags>. It is also possible to make the
12176 regex case insensitive by adding the flag "i" in <flags>. Since <flags> is a
12177 string, it is made up from the concatenation of all desired flags. Thus if
12178 both "i" and "g" are desired, using "gi" or "ig" will have the same effect.
12179 It is important to note that due to the current limitations of the
Baptiste Assmann66025d82016-03-06 23:36:48 +010012180 configuration parser, some characters such as closing parenthesis, closing
12181 square brackets or comma are not possible to use in the arguments. The first
12182 use of this converter is to replace certain characters or sequence of
12183 characters with other ones.
Willy Tarreau7eda8492015-01-20 19:47:06 +010012184
12185 Example :
12186
12187 # de-duplicate "/" in header "x-path".
12188 # input: x-path: /////a///b/c/xzxyz/
12189 # output: x-path: /a/b/c/xzxyz/
12190 http-request set-header x-path %[hdr(x-path),regsub(/+,/,g)]
12191
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020012192capture-req(<id>)
12193 Capture the string entry in the request slot <id> and returns the entry as
12194 is. If the slot doesn't exist, the capture fails silently.
12195
12196 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020012197 "http-response capture", "capture.req.hdr" and
12198 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020012199
12200capture-res(<id>)
12201 Capture the string entry in the response slot <id> and returns the entry as
12202 is. If the slot doesn't exist, the capture fails silently.
12203
12204 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020012205 "http-response capture", "capture.req.hdr" and
12206 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020012207
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012208sdbm([<avalanche>])
12209 Hashes a binary input sample into an unsigned 32-bit quantity using the SDBM
12210 hash function. Optionally, it is possible to apply a full avalanche hash
12211 function to the output if the optional <avalanche> argument equals 1. This
12212 converter uses the same functions as used by the various hash-based load
12213 balancing algorithms, so it will provide exactly the same results. It is
12214 mostly intended for debugging, but can be used as a stick-table entry to
12215 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010012216 32-bit hash is trivial to break. See also "crc32", "djb2", "wt6" and the
12217 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012218
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012219set-var(<var name>)
Daniel Schneller0b547052016-03-21 20:46:57 +010012220 Sets a variable with the input content and returns the content on the output as
12221 is. The variable keeps the value and the associated input type. The name of the
12222 variable starts with an indication about its scope. The scopes allowed are:
12223 "sess" : the variable is shared with the whole session
12224 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012225 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010012226 "req" : the variable is shared only during request processing,
12227 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012228 This prefix is followed by a name. The separator is a '.'. The name may only
12229 contain characters 'a-z', 'A-Z', '0-9' and '_'.
12230
Willy Tarreau97707872015-01-27 15:12:13 +010012231sub(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012232 Subtracts <value> from the input value of type signed integer, and returns
12233 the result as an signed integer. Note: in order to subtract the input from
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012234 a constant, simply perform a "neg,add(value)". <value> can be a numeric value
Daniel Schneller0b547052016-03-21 20:46:57 +010012235 or a variable name. The name of the variable starts with an indication about
12236 its scope. The scopes allowed are:
12237 "sess" : the variable is shared with the whole session
12238 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012239 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010012240 "req" : the variable is shared only during request processing,
12241 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012242 This prefix is followed by a name. The separator is a '.'. The name may only
12243 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012244
Willy Tarreaud9f316a2014-07-10 14:03:38 +020012245table_bytes_in_rate(<table>)
12246 Uses the string representation of the input sample to perform a look up in
12247 the specified table. If the key is not found in the table, integer value zero
12248 is returned. Otherwise the converter returns the average client-to-server
12249 bytes rate associated with the input sample in the designated table, measured
12250 in amount of bytes over the period configured in the table. See also the
12251 sc_bytes_in_rate sample fetch keyword.
12252
12253
12254table_bytes_out_rate(<table>)
12255 Uses the string representation of the input sample to perform a look up in
12256 the specified table. If the key is not found in the table, integer value zero
12257 is returned. Otherwise the converter returns the average server-to-client
12258 bytes rate associated with the input sample in the designated table, measured
12259 in amount of bytes over the period configured in the table. See also the
12260 sc_bytes_out_rate sample fetch keyword.
12261
12262table_conn_cnt(<table>)
12263 Uses the string representation of the input sample to perform a look up in
12264 the specified table. If the key is not found in the table, integer value zero
12265 is returned. Otherwise the converter returns the cumulated amount of incoming
12266 connections associated with the input sample in the designated table. See
12267 also the sc_conn_cnt sample fetch keyword.
12268
12269table_conn_cur(<table>)
12270 Uses the string representation of the input sample to perform a look up in
12271 the specified table. If the key is not found in the table, integer value zero
12272 is returned. Otherwise the converter returns the current amount of concurrent
12273 tracked connections associated with the input sample in the designated table.
12274 See also the sc_conn_cur sample fetch keyword.
12275
12276table_conn_rate(<table>)
12277 Uses the string representation of the input sample to perform a look up in
12278 the specified table. If the key is not found in the table, integer value zero
12279 is returned. Otherwise the converter returns the average incoming connection
12280 rate associated with the input sample in the designated table. See also the
12281 sc_conn_rate sample fetch keyword.
12282
Thierry FOURNIER236657b2015-08-19 08:25:14 +020012283table_gpt0(<table>)
12284 Uses the string representation of the input sample to perform a look up in
12285 the specified table. If the key is not found in the table, boolean value zero
12286 is returned. Otherwise the converter returns the current value of the first
12287 general purpose tag associated with the input sample in the designated table.
12288 See also the sc_get_gpt0 sample fetch keyword.
12289
Willy Tarreaud9f316a2014-07-10 14:03:38 +020012290table_gpc0(<table>)
12291 Uses the string representation of the input sample to perform a look up in
12292 the specified table. If the key is not found in the table, integer value zero
12293 is returned. Otherwise the converter returns the current value of the first
12294 general purpose counter associated with the input sample in the designated
12295 table. See also the sc_get_gpc0 sample fetch keyword.
12296
12297table_gpc0_rate(<table>)
12298 Uses the string representation of the input sample to perform a look up in
12299 the specified table. If the key is not found in the table, integer value zero
12300 is returned. Otherwise the converter returns the frequency which the gpc0
12301 counter was incremented over the configured period in the table, associated
12302 with the input sample in the designated table. See also the sc_get_gpc0_rate
12303 sample fetch keyword.
12304
12305table_http_err_cnt(<table>)
12306 Uses the string representation of the input sample to perform a look up in
12307 the specified table. If the key is not found in the table, integer value zero
12308 is returned. Otherwise the converter returns the cumulated amount of HTTP
12309 errors associated with the input sample in the designated table. See also the
12310 sc_http_err_cnt sample fetch keyword.
12311
12312table_http_err_rate(<table>)
12313 Uses the string representation of the input sample to perform a look up in
12314 the specified table. If the key is not found in the table, integer value zero
12315 is returned. Otherwise the average rate of HTTP errors associated with the
12316 input sample in the designated table, measured in amount of errors over the
12317 period configured in the table. See also the sc_http_err_rate sample fetch
12318 keyword.
12319
12320table_http_req_cnt(<table>)
12321 Uses the string representation of the input sample to perform a look up in
12322 the specified table. If the key is not found in the table, integer value zero
12323 is returned. Otherwise the converter returns the cumulated amount of HTTP
12324 requests associated with the input sample in the designated table. See also
12325 the sc_http_req_cnt sample fetch keyword.
12326
12327table_http_req_rate(<table>)
12328 Uses the string representation of the input sample to perform a look up in
12329 the specified table. If the key is not found in the table, integer value zero
12330 is returned. Otherwise the average rate of HTTP requests associated with the
12331 input sample in the designated table, measured in amount of requests over the
12332 period configured in the table. See also the sc_http_req_rate sample fetch
12333 keyword.
12334
12335table_kbytes_in(<table>)
12336 Uses the string representation of the input sample to perform a look up in
12337 the specified table. If the key is not found in the table, integer value zero
12338 is returned. Otherwise the converter returns the cumulated amount of client-
12339 to-server data associated with the input sample in the designated table,
12340 measured in kilobytes. The test is currently performed on 32-bit integers,
12341 which limits values to 4 terabytes. See also the sc_kbytes_in sample fetch
12342 keyword.
12343
12344table_kbytes_out(<table>)
12345 Uses the string representation of the input sample to perform a look up in
12346 the specified table. If the key is not found in the table, integer value zero
12347 is returned. Otherwise the converter returns the cumulated amount of server-
12348 to-client data associated with the input sample in the designated table,
12349 measured in kilobytes. The test is currently performed on 32-bit integers,
12350 which limits values to 4 terabytes. See also the sc_kbytes_out sample fetch
12351 keyword.
12352
12353table_server_id(<table>)
12354 Uses the string representation of the input sample to perform a look up in
12355 the specified table. If the key is not found in the table, integer value zero
12356 is returned. Otherwise the converter returns the server ID associated with
12357 the input sample in the designated table. A server ID is associated to a
12358 sample by a "stick" rule when a connection to a server succeeds. A server ID
12359 zero means that no server is associated with this key.
12360
12361table_sess_cnt(<table>)
12362 Uses the string representation of the input sample to perform a look up in
12363 the specified table. If the key is not found in the table, integer value zero
12364 is returned. Otherwise the converter returns the cumulated amount of incoming
12365 sessions associated with the input sample in the designated table. Note that
12366 a session here refers to an incoming connection being accepted by the
12367 "tcp-request connection" rulesets. See also the sc_sess_cnt sample fetch
12368 keyword.
12369
12370table_sess_rate(<table>)
12371 Uses the string representation of the input sample to perform a look up in
12372 the specified table. If the key is not found in the table, integer value zero
12373 is returned. Otherwise the converter returns the average incoming session
12374 rate associated with the input sample in the designated table. Note that a
12375 session here refers to an incoming connection being accepted by the
12376 "tcp-request connection" rulesets. See also the sc_sess_rate sample fetch
12377 keyword.
12378
12379table_trackers(<table>)
12380 Uses the string representation of the input sample to perform a look up in
12381 the specified table. If the key is not found in the table, integer value zero
12382 is returned. Otherwise the converter returns the current amount of concurrent
12383 connections tracking the same key as the input sample in the designated
12384 table. It differs from table_conn_cur in that it does not rely on any stored
12385 information but on the table's reference count (the "use" value which is
12386 returned by "show table" on the CLI). This may sometimes be more suited for
12387 layer7 tracking. It can be used to tell a server how many concurrent
12388 connections there are from a given address for example. See also the
12389 sc_trackers sample fetch keyword.
12390
Willy Tarreauffcb2e42014-07-10 16:29:08 +020012391upper
12392 Convert a string sample to upper case. This can only be placed after a string
12393 sample fetch function or after a transformation keyword returning a string
12394 type. The result is of type string.
12395
Thierry FOURNIER82ff3c92015-05-07 15:46:20 +020012396url_dec
12397 Takes an url-encoded string provided as input and returns the decoded
12398 version as output. The input and the output are of type string.
12399
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020012400utime(<format>[,<offset>])
12401 Converts an integer supposed to contain a date since epoch to a string
12402 representing this date in UTC time using a format defined by the <format>
12403 string using strftime(3). The purpose is to allow any date format to be used
12404 in logs. An optional <offset> in seconds may be applied to the input date
12405 (positive or negative). See the strftime() man page for the format supported
12406 by your operating system. See also the ltime converter.
12407
12408 Example :
12409
12410 # Emit two colons, one with the UTC time and another with ip:port
12411 # Eg: 20140710162350 127.0.0.1:57325
12412 log-format %[date,utime(%Y%m%d%H%M%S)]\ %ci:%cp
12413
Emeric Brunc9a0f6d2014-11-25 14:09:01 +010012414word(<index>,<delimiters>)
12415 Extracts the nth word considering given delimiters from an input string.
12416 Indexes start at 1 and delimiters are a string formatted list of chars.
12417
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012418wt6([<avalanche>])
12419 Hashes a binary input sample into an unsigned 32-bit quantity using the WT6
12420 hash function. Optionally, it is possible to apply a full avalanche hash
12421 function to the output if the optional <avalanche> argument equals 1. This
12422 converter uses the same functions as used by the various hash-based load
12423 balancing algorithms, so it will provide exactly the same results. It is
12424 mostly intended for debugging, but can be used as a stick-table entry to
12425 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010012426 32-bit hash is trivial to break. See also "crc32", "djb2", "sdbm", and the
12427 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012428
Willy Tarreau97707872015-01-27 15:12:13 +010012429xor(<value>)
12430 Performs a bitwise "XOR" (exclusive OR) between <value> and the input value
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012431 of type signed integer, and returns the result as an signed integer.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012432 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010012433 starts with an indication about its scope. The scopes allowed are:
12434 "sess" : the variable is shared with the whole session
12435 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012436 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010012437 "req" : the variable is shared only during request processing,
12438 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012439 This prefix is followed by a name. The separator is a '.'. The name may only
12440 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012441
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012442
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200124437.3.2. Fetching samples from internal states
Willy Tarreau74ca5042013-06-11 23:12:07 +020012444--------------------------------------------
12445
12446A first set of sample fetch methods applies to internal information which does
12447not even relate to any client information. These ones are sometimes used with
12448"monitor-fail" directives to report an internal status to external watchers.
12449The sample fetch methods described in this section are usable anywhere.
12450
12451always_false : boolean
12452 Always returns the boolean "false" value. It may be used with ACLs as a
12453 temporary replacement for another one when adjusting configurations.
12454
12455always_true : boolean
12456 Always returns the boolean "true" value. It may be used with ACLs as a
12457 temporary replacement for another one when adjusting configurations.
12458
12459avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010012460 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +020012461 divided by the number of active servers. The current backend is used if no
12462 backend is specified. This is very similar to "queue" except that the size of
12463 the farm is considered, in order to give a more accurate measurement of the
12464 time it may take for a new connection to be processed. The main usage is with
12465 ACL to return a sorry page to new users when it becomes certain they will get
12466 a degraded service, or to pass to the backend servers in a header so that
12467 they decide to work in degraded mode or to disable some functions to speed up
12468 the processing a bit. Note that in the event there would not be any active
12469 server anymore, twice the number of queued connections would be considered as
12470 the measured value. This is a fair estimate, as we expect one server to get
12471 back soon anyway, but we still prefer to send new traffic to another backend
12472 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
12473 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +010012474
Willy Tarreau74ca5042013-06-11 23:12:07 +020012475be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +020012476 Applies to the number of currently established connections on the backend,
12477 possibly including the connection being evaluated. If no backend name is
12478 specified, the current one is used. But it is also possible to check another
12479 backend. It can be used to use a specific farm when the nominal one is full.
12480 See also the "fe_conn", "queue" and "be_sess_rate" criteria.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012481
Willy Tarreau74ca5042013-06-11 23:12:07 +020012482be_sess_rate([<backend>]) : integer
12483 Returns an integer value corresponding to the sessions creation rate on the
12484 backend, in number of new sessions per second. This is used with ACLs to
12485 switch to an alternate backend when an expensive or fragile one reaches too
12486 high a session rate, or to limit abuse of service (eg. prevent sucking of an
12487 online dictionary). It can also be useful to add this element to logs using a
12488 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012489
12490 Example :
12491 # Redirect to an error page if the dictionary is requested too often
12492 backend dynamic
12493 mode http
12494 acl being_scanned be_sess_rate gt 100
12495 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +010012496
Thierry FOURNIERcc103292015-06-06 19:30:17 +020012497bin(<hexa>) : bin
12498 Returns a binary chain. The input is the hexadecimal representation
12499 of the string.
12500
12501bool(<bool>) : bool
12502 Returns a boolean value. <bool> can be 'true', 'false', '1' or '0'.
12503 'false' and '0' are the same. 'true' and '1' are the same.
12504
Willy Tarreau74ca5042013-06-11 23:12:07 +020012505connslots([<backend>]) : integer
12506 Returns an integer value corresponding to the number of connection slots
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012507 still available in the backend, by totaling the maximum amount of
Willy Tarreau74ca5042013-06-11 23:12:07 +020012508 connections on all servers and the maximum queue size. This is probably only
12509 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -050012510
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012511 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +020012512 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012513 usage; see "use_backend" keyword) can be redirected to a different backend.
12514
Willy Tarreau55165fe2009-05-10 12:02:55 +020012515 'connslots' = number of available server connection slots, + number of
12516 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012517
Willy Tarreaua36af912009-10-10 12:02:45 +020012518 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +020012519 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +020012520 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +020012521 you want to be able to differentiate between different backends, and their
12522 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +020012523 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +020012524 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012525
Willy Tarreau55165fe2009-05-10 12:02:55 +020012526 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
12527 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +020012528 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +020012529 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012530
Willy Tarreau6236d3a2013-07-25 14:28:25 +020012531date([<offset>]) : integer
12532 Returns the current date as the epoch (number of seconds since 01/01/1970).
12533 If an offset value is specified, then it is a number of seconds that is added
12534 to the current date before returning the value. This is particularly useful
12535 to compute relative dates, as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +020012536 It is useful combined with the http_date converter.
12537
12538 Example :
12539
12540 # set an expires header to now+1 hour in every response
12541 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +020012542
Willy Tarreau595ec542013-06-12 21:34:28 +020012543env(<name>) : string
12544 Returns a string containing the value of environment variable <name>. As a
12545 reminder, environment variables are per-process and are sampled when the
12546 process starts. This can be useful to pass some information to a next hop
12547 server, or with ACLs to take specific action when the process is started a
12548 certain way.
12549
12550 Examples :
12551 # Pass the Via header to next hop with the local hostname in it
12552 http-request add-header Via 1.1\ %[env(HOSTNAME)]
12553
12554 # reject cookie-less requests when the STOP environment variable is set
12555 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
12556
Willy Tarreau74ca5042013-06-11 23:12:07 +020012557fe_conn([<frontend>]) : integer
12558 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +010012559 possibly including the connection being evaluated. If no frontend name is
12560 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020012561 frontend. It can be used to return a sorry page before hard-blocking, or to
12562 use a specific backend to drain new requests when the farm is considered
12563 full. This is mostly used with ACLs but can also be used to pass some
12564 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
12565 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +020012566
Willy Tarreau74ca5042013-06-11 23:12:07 +020012567fe_sess_rate([<frontend>]) : integer
12568 Returns an integer value corresponding to the sessions creation rate on the
12569 frontend, in number of new sessions per second. This is used with ACLs to
12570 limit the incoming session rate to an acceptable range in order to prevent
12571 abuse of service at the earliest moment, for example when combined with other
12572 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
12573 down below the limit. It can also be useful to add this element to logs using
12574 a log-format directive. See also the "rate-limit sessions" directive for use
12575 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010012576
12577 Example :
12578 # This frontend limits incoming mails to 10/s with a max of 100
12579 # concurrent connections. We accept any connection below 10/s, and
12580 # force excess clients to wait for 100 ms. Since clients are limited to
12581 # 100 max, there cannot be more than 10 incoming mails per second.
12582 frontend mail
12583 bind :25
12584 mode tcp
12585 maxconn 100
12586 acl too_fast fe_sess_rate ge 10
12587 tcp-request inspect-delay 100ms
12588 tcp-request content accept if ! too_fast
12589 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +010012590
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012591int(<integer>) : signed integer
12592 Returns a signed integer.
12593
Thierry FOURNIERcc103292015-06-06 19:30:17 +020012594ipv4(<ipv4>) : ipv4
12595 Returns an ipv4.
12596
12597ipv6(<ipv6>) : ipv6
12598 Returns an ipv6.
12599
12600meth(<method>) : method
12601 Returns a method.
12602
Willy Tarreau0f30d262014-11-24 16:02:05 +010012603nbproc : integer
12604 Returns an integer value corresponding to the number of processes that were
12605 started (it equals the global "nbproc" setting). This is useful for logging
12606 and debugging purposes.
12607
Willy Tarreau74ca5042013-06-11 23:12:07 +020012608nbsrv([<backend>]) : integer
12609 Returns an integer value corresponding to the number of usable servers of
12610 either the current backend or the named backend. This is mostly used with
12611 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +010012612 switch to an alternate backend when the number of servers is too low to
12613 to handle some load. It is useful to report a failure when combined with
12614 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +010012615
Willy Tarreau0f30d262014-11-24 16:02:05 +010012616proc : integer
12617 Returns an integer value corresponding to the position of the process calling
12618 the function, between 1 and global.nbproc. This is useful for logging and
12619 debugging purposes.
12620
Willy Tarreau74ca5042013-06-11 23:12:07 +020012621queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010012622 Returns the total number of queued connections of the designated backend,
12623 including all the connections in server queues. If no backend name is
12624 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020012625 one. This is useful with ACLs or to pass statistics to backend servers. This
12626 can be used to take actions when queuing goes above a known level, generally
12627 indicating a surge of traffic or a massive slowdown on the servers. One
12628 possible action could be to reject new users but still accept old ones. See
12629 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
12630
Willy Tarreau84310e22014-02-14 11:59:04 +010012631rand([<range>]) : integer
12632 Returns a random integer value within a range of <range> possible values,
12633 starting at zero. If the range is not specified, it defaults to 2^32, which
12634 gives numbers between 0 and 4294967295. It can be useful to pass some values
12635 needed to take some routing decisions for example, or just for debugging
12636 purposes. This random must not be used for security purposes.
12637
Willy Tarreau74ca5042013-06-11 23:12:07 +020012638srv_conn([<backend>/]<server>) : integer
12639 Returns an integer value corresponding to the number of currently established
12640 connections on the designated server, possibly including the connection being
12641 evaluated. If <backend> is omitted, then the server is looked up in the
12642 current backend. It can be used to use a specific farm when one server is
12643 full, or to inform the server about our view of the number of active
12644 connections with it. See also the "fe_conn", "be_conn" and "queue" fetch
12645 methods.
12646
12647srv_is_up([<backend>/]<server>) : boolean
12648 Returns true when the designated server is UP, and false when it is either
12649 DOWN or in maintenance mode. If <backend> is omitted, then the server is
12650 looked up in the current backend. It is mainly used to take action based on
12651 an external status reported via a health check (eg: a geographical site's
12652 availability). Another possible use which is more of a hack consists in
12653 using dummy servers as boolean variables that can be enabled or disabled from
12654 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
12655
12656srv_sess_rate([<backend>/]<server>) : integer
12657 Returns an integer corresponding to the sessions creation rate on the
12658 designated server, in number of new sessions per second. If <backend> is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012659 omitted, then the server is looked up in the current backend. This is mostly
Willy Tarreau74ca5042013-06-11 23:12:07 +020012660 used with ACLs but can make sense with logs too. This is used to switch to an
12661 alternate backend when an expensive or fragile one reaches too high a session
12662 rate, or to limit abuse of service (eg. prevent latent requests from
12663 overloading servers).
12664
12665 Example :
12666 # Redirect to a separate back
12667 acl srv1_full srv_sess_rate(be1/srv1) gt 50
12668 acl srv2_full srv_sess_rate(be1/srv2) gt 50
12669 use_backend be2 if srv1_full or srv2_full
12670
Willy Tarreau0f30d262014-11-24 16:02:05 +010012671stopping : boolean
12672 Returns TRUE if the process calling the function is currently stopping. This
12673 can be useful for logging, or for relaxing certain checks or helping close
12674 certain connections upon graceful shutdown.
12675
Thierry FOURNIERcc103292015-06-06 19:30:17 +020012676str(<string>) : string
12677 Returns a string.
12678
Willy Tarreau74ca5042013-06-11 23:12:07 +020012679table_avl([<table>]) : integer
12680 Returns the total number of available entries in the current proxy's
12681 stick-table or in the designated stick-table. See also table_cnt.
12682
12683table_cnt([<table>]) : integer
12684 Returns the total number of entries currently in use in the current proxy's
12685 stick-table or in the designated stick-table. See also src_conn_cnt and
12686 table_avl for other entry counting methods.
12687
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012688var(<var-name>) : undefined
12689 Returns a variable with the stored type. If the variable is not set, the
Daniel Schneller0b547052016-03-21 20:46:57 +010012690 sample fetch fails. The name of the variable starts with an indication
12691 about its scope. The scopes allowed are:
12692 "sess" : the variable is shared with the whole session
12693 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012694 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010012695 "req" : the variable is shared only during request processing,
12696 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012697 This prefix is followed by a name. The separator is a '.'. The name may only
12698 contain characters 'a-z', 'A-Z', '0-9' and '_'.
12699
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200127007.3.3. Fetching samples at Layer 4
Willy Tarreau74ca5042013-06-11 23:12:07 +020012701----------------------------------
12702
12703The layer 4 usually describes just the transport layer which in haproxy is
12704closest to the connection, where no content is yet made available. The fetch
12705methods described here are usable as low as the "tcp-request connection" rule
12706sets unless they require some future information. Those generally include
12707TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012708the incoming connection. For retrieving a value from a sticky counters, the
12709counter number can be explicitly set as 0, 1, or 2 using the pre-defined
12710"sc0_", "sc1_", or "sc2_" prefix, or it can be specified as the first integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012711argument when using the "sc_" prefix. An optional table may be specified with
12712the "sc*" form, in which case the currently tracked key will be looked up into
12713this alternate table instead of the table currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012714
12715be_id : integer
12716 Returns an integer containing the current backend's id. It can be used in
12717 frontends with responses to check which backend processed the request.
12718
12719dst : ip
12720 This is the destination IPv4 address of the connection on the client side,
12721 which is the address the client connected to. It can be useful when running
12722 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
12723 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
12724 RFC 4291.
12725
12726dst_conn : integer
12727 Returns an integer value corresponding to the number of currently established
12728 connections on the same socket including the one being evaluated. It is
12729 normally used with ACLs but can as well be used to pass the information to
12730 servers in an HTTP header or in logs. It can be used to either return a sorry
12731 page before hard-blocking, or to use a specific backend to drain new requests
12732 when the socket is considered saturated. This offers the ability to assign
12733 different limits to different listening ports or addresses. See also the
12734 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012735
Willy Tarreau16e01562016-08-09 16:46:18 +020012736dst_is_local : boolean
12737 Returns true if the destination address of the incoming connection is local
12738 to the system, or false if the address doesn't exist on the system, meaning
12739 that it was intercepted in transparent mode. It can be useful to apply
12740 certain rules by default to forwarded traffic and other rules to the traffic
12741 targetting the real address of the machine. For example the stats page could
12742 be delivered only on this address, or SSH access could be locally redirected.
12743 Please note that the check involves a few system calls, so it's better to do
12744 it only once per connection.
12745
Willy Tarreau74ca5042013-06-11 23:12:07 +020012746dst_port : integer
12747 Returns an integer value corresponding to the destination TCP port of the
12748 connection on the client side, which is the port the client connected to.
12749 This might be used when running in transparent mode, when assigning dynamic
12750 ports to some clients for a whole application session, to stick all users to
12751 a same server, or to pass the destination port information to a server using
12752 an HTTP header.
12753
Thierry Fournier / OZON.IO6310bef2016-07-24 20:16:50 +020012754fc_rtt(<unit>) : integer
12755 Returns the Round Trip Time (RTT) measured by the kernel for the client
12756 connection. <unit> is facultative, by default the unit is milliseconds. <unit>
12757 can be set to "ms" for milliseconds or "us" for microseconds. If the server
12758 connection is not established, if the connection is not TCP or if the
12759 operating system does not support TCP_INFO, for example Linux kernels before
12760 2.4, the sample fetch fails.
12761
12762fc_rttvar(<unit>) : integer
12763 Returns the Round Trip Time (RTT) variance measured by the kernel for the
12764 client connection. <unit> is facultative, by default the unit is milliseconds.
12765 <unit> can be set to "ms" for milliseconds or "us" for microseconds. If the
12766 server connection is not established, if the connection is not TCP or if the
12767 operating system does not support TCP_INFO, for example Linux kernels before
12768 2.4, the sample fetch fails.
12769
Joe Williams30fcd392016-08-10 07:06:44 -070012770fc_unacked(<unit>) : integer
12771 Returns the unacked counter measured by the kernel for the client connection.
12772 If the server connection is not established, if the connection is not TCP or
12773 if the operating system does not support TCP_INFO, for example Linux kernels
12774 before 2.4, the sample fetch fails.
12775
12776fc_sacked(<unit>) : integer
12777 Returns the sacked counter measured by the kernel for the client connection.
12778 If the server connection is not established, if the connection is not TCP or
12779 if the operating system does not support TCP_INFO, for example Linux kernels
12780 before 2.4, the sample fetch fails.
12781
12782fc_retrans(<unit>) : integer
12783 Returns the retransmits counter measured by the kernel for the client
12784 connection. If the server connection is not established, if the connection is
12785 not TCP or if the operating system does not support TCP_INFO, for example
12786 Linux kernels before 2.4, the sample fetch fails.
12787
12788fc_fackets(<unit>) : integer
12789 Returns the fack counter measured by the kernel for the client
12790 connection. If the server connection is not established, if the connection is
12791 not TCP or if the operating system does not support TCP_INFO, for example
12792 Linux kernels before 2.4, the sample fetch fails.
12793
12794fc_lost(<unit>) : integer
12795 Returns the lost counter measured by the kernel for the client
12796 connection. If the server connection is not established, if the connection is
12797 not TCP or if the operating system does not support TCP_INFO, for example
12798 Linux kernels before 2.4, the sample fetch fails.
12799
12800fc_reordering(<unit>) : integer
12801 Returns the reordering counter measured by the kernel for the client
12802 connection. If the server connection is not established, if the connection is
12803 not TCP or if the operating system does not support TCP_INFO, for example
12804 Linux kernels before 2.4, the sample fetch fails.
12805
Willy Tarreau74ca5042013-06-11 23:12:07 +020012806fe_id : integer
12807 Returns an integer containing the current frontend's id. It can be used in
12808 backends to check from which backend it was called, or to stick all users
12809 coming via a same frontend to the same server.
12810
Cyril Bonté62ba8702014-04-22 23:52:25 +020012811sc_bytes_in_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012812sc0_bytes_in_rate([<table>]) : integer
12813sc1_bytes_in_rate([<table>]) : integer
12814sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012815 Returns the average client-to-server bytes rate from the currently tracked
12816 counters, measured in amount of bytes over the period configured in the
12817 table. See also src_bytes_in_rate.
12818
Cyril Bonté62ba8702014-04-22 23:52:25 +020012819sc_bytes_out_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012820sc0_bytes_out_rate([<table>]) : integer
12821sc1_bytes_out_rate([<table>]) : integer
12822sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012823 Returns the average server-to-client bytes rate from the currently tracked
12824 counters, measured in amount of bytes over the period configured in the
12825 table. See also src_bytes_out_rate.
12826
Cyril Bonté62ba8702014-04-22 23:52:25 +020012827sc_clr_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012828sc0_clr_gpc0([<table>]) : integer
12829sc1_clr_gpc0([<table>]) : integer
12830sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +020012831 Clears the first General Purpose Counter associated to the currently tracked
12832 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +010012833 stored value is zero, so first invocation will always return zero. This is
12834 typically used as a second ACL in an expression in order to mark a connection
12835 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020012836
12837 # block if 5 consecutive requests continue to come faster than 10 sess
12838 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012839 acl abuse sc0_http_req_rate gt 10
12840 acl kill sc0_inc_gpc0 gt 5
12841 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020012842 tcp-request connection accept if !abuse save
12843 tcp-request connection reject if abuse kill
12844
Cyril Bonté62ba8702014-04-22 23:52:25 +020012845sc_conn_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012846sc0_conn_cnt([<table>]) : integer
12847sc1_conn_cnt([<table>]) : integer
12848sc2_conn_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012849 Returns the cumulated number of incoming connections from currently tracked
12850 counters. See also src_conn_cnt.
12851
Cyril Bonté62ba8702014-04-22 23:52:25 +020012852sc_conn_cur(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012853sc0_conn_cur([<table>]) : integer
12854sc1_conn_cur([<table>]) : integer
12855sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012856 Returns the current amount of concurrent connections tracking the same
12857 tracked counters. This number is automatically incremented when tracking
12858 begins and decremented when tracking stops. See also src_conn_cur.
12859
Cyril Bonté62ba8702014-04-22 23:52:25 +020012860sc_conn_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012861sc0_conn_rate([<table>]) : integer
12862sc1_conn_rate([<table>]) : integer
12863sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012864 Returns the average connection rate from the currently tracked counters,
12865 measured in amount of connections over the period configured in the table.
12866 See also src_conn_rate.
12867
Cyril Bonté62ba8702014-04-22 23:52:25 +020012868sc_get_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012869sc0_get_gpc0([<table>]) : integer
12870sc1_get_gpc0([<table>]) : integer
12871sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012872 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012873 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020012874
Thierry FOURNIER236657b2015-08-19 08:25:14 +020012875sc_get_gpt0(<ctr>[,<table>]) : integer
12876sc0_get_gpt0([<table>]) : integer
12877sc1_get_gpt0([<table>]) : integer
12878sc2_get_gpt0([<table>]) : integer
12879 Returns the value of the first General Purpose Tag associated to the
12880 currently tracked counters. See also src_get_gpt0.
12881
Cyril Bonté62ba8702014-04-22 23:52:25 +020012882sc_gpc0_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012883sc0_gpc0_rate([<table>]) : integer
12884sc1_gpc0_rate([<table>]) : integer
12885sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020012886 Returns the average increment rate of the first General Purpose Counter
12887 associated to the currently tracked counters. It reports the frequency
12888 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012889 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
12890 that the "gpc0_rate" counter must be stored in the stick-table for a value to
12891 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +020012892
Cyril Bonté62ba8702014-04-22 23:52:25 +020012893sc_http_err_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012894sc0_http_err_cnt([<table>]) : integer
12895sc1_http_err_cnt([<table>]) : integer
12896sc2_http_err_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012897 Returns the cumulated number of HTTP errors from the currently tracked
12898 counters. This includes the both request errors and 4xx error responses.
12899 See also src_http_err_cnt.
12900
Cyril Bonté62ba8702014-04-22 23:52:25 +020012901sc_http_err_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012902sc0_http_err_rate([<table>]) : integer
12903sc1_http_err_rate([<table>]) : integer
12904sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012905 Returns the average rate of HTTP errors from the currently tracked counters,
12906 measured in amount of errors over the period configured in the table. This
12907 includes the both request errors and 4xx error responses. See also
12908 src_http_err_rate.
12909
Cyril Bonté62ba8702014-04-22 23:52:25 +020012910sc_http_req_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012911sc0_http_req_cnt([<table>]) : integer
12912sc1_http_req_cnt([<table>]) : integer
12913sc2_http_req_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012914 Returns the cumulated number of HTTP requests from the currently tracked
12915 counters. This includes every started request, valid or not. See also
12916 src_http_req_cnt.
12917
Cyril Bonté62ba8702014-04-22 23:52:25 +020012918sc_http_req_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012919sc0_http_req_rate([<table>]) : integer
12920sc1_http_req_rate([<table>]) : integer
12921sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012922 Returns the average rate of HTTP requests from the currently tracked
12923 counters, measured in amount of requests over the period configured in
12924 the table. This includes every started request, valid or not. See also
12925 src_http_req_rate.
12926
Cyril Bonté62ba8702014-04-22 23:52:25 +020012927sc_inc_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012928sc0_inc_gpc0([<table>]) : integer
12929sc1_inc_gpc0([<table>]) : integer
12930sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012931 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +010012932 tracked counters, and returns its new value. Before the first invocation,
12933 the stored value is zero, so first invocation will increase it to 1 and will
12934 return 1. This is typically used as a second ACL in an expression in order
12935 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +020012936
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012937 acl abuse sc0_http_req_rate gt 10
12938 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020012939 tcp-request connection reject if abuse kill
12940
Cyril Bonté62ba8702014-04-22 23:52:25 +020012941sc_kbytes_in(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012942sc0_kbytes_in([<table>]) : integer
12943sc1_kbytes_in([<table>]) : integer
12944sc2_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020012945 Returns the total amount of client-to-server data from the currently tracked
12946 counters, measured in kilobytes. The test is currently performed on 32-bit
12947 integers, which limits values to 4 terabytes. See also src_kbytes_in.
Willy Tarreaue9656522010-08-17 15:40:09 +020012948
Cyril Bonté62ba8702014-04-22 23:52:25 +020012949sc_kbytes_out(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012950sc0_kbytes_out([<table>]) : integer
12951sc1_kbytes_out([<table>]) : integer
12952sc2_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020012953 Returns the total amount of server-to-client data from the currently tracked
12954 counters, measured in kilobytes. The test is currently performed on 32-bit
12955 integers, which limits values to 4 terabytes. See also src_kbytes_out.
Willy Tarreaue9656522010-08-17 15:40:09 +020012956
Cyril Bonté62ba8702014-04-22 23:52:25 +020012957sc_sess_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012958sc0_sess_cnt([<table>]) : integer
12959sc1_sess_cnt([<table>]) : integer
12960sc2_sess_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012961 Returns the cumulated number of incoming connections that were transformed
12962 into sessions, which means that they were accepted by a "tcp-request
12963 connection" rule, from the currently tracked counters. A backend may count
12964 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -040012965 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +020012966 with the client. See also src_sess_cnt.
12967
Cyril Bonté62ba8702014-04-22 23:52:25 +020012968sc_sess_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012969sc0_sess_rate([<table>]) : integer
12970sc1_sess_rate([<table>]) : integer
12971sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012972 Returns the average session rate from the currently tracked counters,
12973 measured in amount of sessions over the period configured in the table. A
12974 session is a connection that got past the early "tcp-request connection"
12975 rules. A backend may count more sessions than connections because each
12976 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -040012977 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +020012978
Cyril Bonté62ba8702014-04-22 23:52:25 +020012979sc_tracked(<ctr>[,<table>]) : boolean
Willy Tarreau0f791d42013-07-23 19:56:43 +020012980sc0_tracked([<table>]) : boolean
12981sc1_tracked([<table>]) : boolean
12982sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +020012983 Returns true if the designated session counter is currently being tracked by
12984 the current session. This can be useful when deciding whether or not we want
12985 to set some values in a header passed to the server.
12986
Cyril Bonté62ba8702014-04-22 23:52:25 +020012987sc_trackers(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012988sc0_trackers([<table>]) : integer
12989sc1_trackers([<table>]) : integer
12990sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +010012991 Returns the current amount of concurrent connections tracking the same
12992 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012993 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +010012994 that it does not rely on any stored information but on the table's reference
12995 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020012996 may sometimes be more suited for layer7 tracking. It can be used to tell a
12997 server how many concurrent connections there are from a given address for
12998 example.
Willy Tarreau2406db42012-12-09 12:16:43 +010012999
Willy Tarreau74ca5042013-06-11 23:12:07 +020013000so_id : integer
13001 Returns an integer containing the current listening socket's id. It is useful
13002 in frontends involving many "bind" lines, or to stick all users coming via a
13003 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013004
Willy Tarreau74ca5042013-06-11 23:12:07 +020013005src : ip
13006 This is the source IPv4 address of the client of the session. It is of type
13007 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
13008 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
13009 TCP-level source address which is used, and not the address of a client
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010013010 behind a proxy. However if the "accept-proxy" or "accept-netscaler-cip" bind
13011 directive is used, it can be the address of a client behind another
13012 PROXY-protocol compatible component for all rule sets except
13013 "tcp-request connection" which sees the real address.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013014
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013015 Example:
13016 # add an HTTP header in requests with the originating address' country
13017 http-request set-header X-Country %[src,map_ip(geoip.lst)]
13018
Willy Tarreau74ca5042013-06-11 23:12:07 +020013019src_bytes_in_rate([<table>]) : integer
13020 Returns the average bytes rate from the incoming connection's source address
13021 in the current proxy's stick-table or in the designated stick-table, measured
13022 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013023 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013024
Willy Tarreau74ca5042013-06-11 23:12:07 +020013025src_bytes_out_rate([<table>]) : integer
13026 Returns the average bytes rate to the incoming connection's source address in
13027 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +020013028 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013029 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013030
Willy Tarreau74ca5042013-06-11 23:12:07 +020013031src_clr_gpc0([<table>]) : integer
13032 Clears the first General Purpose Counter associated to the incoming
13033 connection's source address in the current proxy's stick-table or in the
13034 designated stick-table, and returns its previous value. If the address is not
13035 found, an entry is created and 0 is returned. This is typically used as a
13036 second ACL in an expression in order to mark a connection when a first ACL
13037 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020013038
13039 # block if 5 consecutive requests continue to come faster than 10 sess
13040 # per second, and reset the counter as soon as the traffic slows down.
13041 acl abuse src_http_req_rate gt 10
13042 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +010013043 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020013044 tcp-request connection accept if !abuse save
13045 tcp-request connection reject if abuse kill
13046
Willy Tarreau74ca5042013-06-11 23:12:07 +020013047src_conn_cnt([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020013048 Returns the cumulated number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +020013049 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020013050 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013051 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013052
Willy Tarreau74ca5042013-06-11 23:12:07 +020013053src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020013054 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +020013055 current incoming connection's source address in the current proxy's
13056 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013057 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013058
Willy Tarreau74ca5042013-06-11 23:12:07 +020013059src_conn_rate([<table>]) : integer
13060 Returns the average connection rate from the incoming connection's source
13061 address in the current proxy's stick-table or in the designated stick-table,
13062 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013063 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013064
Willy Tarreau74ca5042013-06-11 23:12:07 +020013065src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020013066 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +020013067 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020013068 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013069 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013070
Thierry FOURNIER236657b2015-08-19 08:25:14 +020013071src_get_gpt0([<table>]) : integer
13072 Returns the value of the first General Purpose Tag associated to the
13073 incoming connection's source address in the current proxy's stick-table or in
13074 the designated stick-table. If the address is not found, zero is returned.
13075 See also sc/sc0/sc1/sc2_get_gpt0.
13076
Willy Tarreau74ca5042013-06-11 23:12:07 +020013077src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020013078 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +020013079 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +020013080 stick-table or in the designated stick-table. It reports the frequency
13081 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013082 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
13083 that the "gpc0_rate" counter must be stored in the stick-table for a value to
13084 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020013085
Willy Tarreau74ca5042013-06-11 23:12:07 +020013086src_http_err_cnt([<table>]) : integer
13087 Returns the cumulated number of HTTP errors from the incoming connection's
13088 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +020013089 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013090 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +020013091 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013092
Willy Tarreau74ca5042013-06-11 23:12:07 +020013093src_http_err_rate([<table>]) : integer
13094 Returns the average rate of HTTP errors from the incoming connection's source
13095 address in the current proxy's stick-table or in the designated stick-table,
13096 measured in amount of errors over the period configured in the table. This
13097 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013098 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013099
Willy Tarreau74ca5042013-06-11 23:12:07 +020013100src_http_req_cnt([<table>]) : integer
13101 Returns the cumulated number of HTTP requests from the incoming connection's
13102 source address in the current proxy's stick-table or in the designated stick-
13103 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013104 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013105
Willy Tarreau74ca5042013-06-11 23:12:07 +020013106src_http_req_rate([<table>]) : integer
13107 Returns the average rate of HTTP requests from the incoming connection's
13108 source address in the current proxy's stick-table or in the designated stick-
13109 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +020013110 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013111 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013112
Willy Tarreau74ca5042013-06-11 23:12:07 +020013113src_inc_gpc0([<table>]) : integer
13114 Increments the first General Purpose Counter associated to the incoming
13115 connection's source address in the current proxy's stick-table or in the
13116 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020013117 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013118 This is typically used as a second ACL in an expression in order to mark a
13119 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +020013120
13121 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +010013122 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020013123 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +020013124
Willy Tarreau16e01562016-08-09 16:46:18 +020013125src_is_local : boolean
13126 Returns true if the source address of the incoming connection is local to the
13127 system, or false if the address doesn't exist on the system, meaning that it
13128 comes from a remote machine. Note that UNIX addresses are considered local.
13129 It can be useful to apply certain access restrictions based on where the
13130 client comes from (eg: require auth or https for remote machines). Please
13131 note that the check involves a few system calls, so it's better to do it only
13132 once per connection.
13133
Willy Tarreau74ca5042013-06-11 23:12:07 +020013134src_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020013135 Returns the total amount of data received from the incoming connection's
13136 source address in the current proxy's stick-table or in the designated
13137 stick-table, measured in kilobytes. If the address is not found, zero is
13138 returned. The test is currently performed on 32-bit integers, which limits
13139 values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013140
Willy Tarreau74ca5042013-06-11 23:12:07 +020013141src_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020013142 Returns the total amount of data sent to the incoming connection's source
13143 address in the current proxy's stick-table or in the designated stick-table,
13144 measured in kilobytes. If the address is not found, zero is returned. The
13145 test is currently performed on 32-bit integers, which limits values to 4
13146 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020013147
Willy Tarreau74ca5042013-06-11 23:12:07 +020013148src_port : integer
13149 Returns an integer value corresponding to the TCP source port of the
13150 connection on the client side, which is the port the client connected from.
13151 Usage of this function is very limited as modern protocols do not care much
13152 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010013153
Willy Tarreau74ca5042013-06-11 23:12:07 +020013154src_sess_cnt([<table>]) : integer
13155 Returns the cumulated number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +020013156 connection's source IPv4 address in the current proxy's stick-table or in the
13157 designated stick-table, that were transformed into sessions, which means that
13158 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013159 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013160
Willy Tarreau74ca5042013-06-11 23:12:07 +020013161src_sess_rate([<table>]) : integer
13162 Returns the average session rate from the incoming connection's source
13163 address in the current proxy's stick-table or in the designated stick-table,
13164 measured in amount of sessions over the period configured in the table. A
13165 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013166 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013167
Willy Tarreau74ca5042013-06-11 23:12:07 +020013168src_updt_conn_cnt([<table>]) : integer
13169 Creates or updates the entry associated to the incoming connection's source
13170 address in the current proxy's stick-table or in the designated stick-table.
13171 This table must be configured to store the "conn_cnt" data type, otherwise
13172 the match will be ignored. The current count is incremented by one, and the
13173 expiration timer refreshed. The updated count is returned, so this match
13174 can't return zero. This was used to reject service abusers based on their
13175 source address. Note: it is recommended to use the more complete "track-sc*"
13176 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020013177
13178 Example :
13179 # This frontend limits incoming SSH connections to 3 per 10 second for
13180 # each source address, and rejects excess connections until a 10 second
13181 # silence is observed. At most 20 addresses are tracked.
13182 listen ssh
13183 bind :22
13184 mode tcp
13185 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +020013186 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +020013187 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +020013188 server local 127.0.0.1:22
13189
Willy Tarreau74ca5042013-06-11 23:12:07 +020013190srv_id : integer
13191 Returns an integer containing the server's id when processing the response.
13192 While it's almost only used with ACLs, it may be used for logging or
13193 debugging.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +020013194
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200131957.3.4. Fetching samples at Layer 5
Willy Tarreau74ca5042013-06-11 23:12:07 +020013196----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +020013197
Willy Tarreau74ca5042013-06-11 23:12:07 +020013198The layer 5 usually describes just the session layer which in haproxy is
13199closest to the session once all the connection handshakes are finished, but
13200when no content is yet made available. The fetch methods described here are
13201usable as low as the "tcp-request content" rule sets unless they require some
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013202future information. Those generally include the results of SSL negotiations.
Willy Tarreauc735a072011-03-29 00:57:02 +020013203
Emeric Brun645ae792014-04-30 14:21:06 +020013204ssl_bc : boolean
13205 Returns true when the back connection was made via an SSL/TLS transport
13206 layer and is locally deciphered. This means the outgoing connection was made
13207 other a server with the "ssl" option.
13208
13209ssl_bc_alg_keysize : integer
13210 Returns the symmetric cipher key size supported in bits when the outgoing
13211 connection was made over an SSL/TLS transport layer.
13212
13213ssl_bc_cipher : string
13214 Returns the name of the used cipher when the outgoing connection was made
13215 over an SSL/TLS transport layer.
13216
13217ssl_bc_protocol : string
13218 Returns the name of the used protocol when the outgoing connection was made
13219 over an SSL/TLS transport layer.
13220
Emeric Brunb73a9b02014-04-30 18:49:19 +020013221ssl_bc_unique_id : binary
Emeric Brun645ae792014-04-30 14:21:06 +020013222 When the outgoing connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020013223 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
13224 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
Emeric Brun645ae792014-04-30 14:21:06 +020013225
13226ssl_bc_session_id : binary
13227 Returns the SSL ID of the back connection when the outgoing connection was
13228 made over an SSL/TLS transport layer. It is useful to log if we want to know
13229 if session was reused or not.
13230
13231ssl_bc_use_keysize : integer
13232 Returns the symmetric cipher key size used in bits when the outgoing
13233 connection was made over an SSL/TLS transport layer.
13234
Willy Tarreau74ca5042013-06-11 23:12:07 +020013235ssl_c_ca_err : integer
13236 When the incoming connection was made over an SSL/TLS transport layer,
13237 returns the ID of the first error detected during verification of the client
13238 certificate at depth > 0, or 0 if no error was encountered during this
13239 verification process. Please refer to your SSL library's documentation to
13240 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +020013241
Willy Tarreau74ca5042013-06-11 23:12:07 +020013242ssl_c_ca_err_depth : integer
13243 When the incoming connection was made over an SSL/TLS transport layer,
13244 returns the depth in the CA chain of the first error detected during the
13245 verification of the client certificate. If no error is encountered, 0 is
13246 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +010013247
Emeric Brun43e79582014-10-29 19:03:26 +010013248ssl_c_der : binary
13249 Returns the DER formatted certificate presented by the client when the
13250 incoming connection was made over an SSL/TLS transport layer. When used for
13251 an ACL, the value(s) to match against can be passed in hexadecimal form.
13252
Willy Tarreau74ca5042013-06-11 23:12:07 +020013253ssl_c_err : integer
13254 When the incoming connection was made over an SSL/TLS transport layer,
13255 returns the ID of the first error detected during verification at depth 0, or
13256 0 if no error was encountered during this verification process. Please refer
13257 to your SSL library's documentation to find the exhaustive list of error
13258 codes.
Willy Tarreau62644772008-07-16 18:36:06 +020013259
Willy Tarreau74ca5042013-06-11 23:12:07 +020013260ssl_c_i_dn([<entry>[,<occ>]]) : string
13261 When the incoming connection was made over an SSL/TLS transport layer,
13262 returns the full distinguished name of the issuer of the certificate
13263 presented by the client when no <entry> is specified, or the value of the
13264 first given entry found from the beginning of the DN. If a positive/negative
13265 occurrence number is specified as the optional second argument, it returns
13266 the value of the nth given entry value from the beginning/end of the DN.
13267 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
13268 "ssl_c_i_dn(CN)" retrieves the common name.
Willy Tarreau62644772008-07-16 18:36:06 +020013269
Willy Tarreau74ca5042013-06-11 23:12:07 +020013270ssl_c_key_alg : string
13271 Returns the name of the algorithm used to generate the key of the certificate
13272 presented by the client when the incoming connection was made over an SSL/TLS
13273 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +020013274
Willy Tarreau74ca5042013-06-11 23:12:07 +020013275ssl_c_notafter : string
13276 Returns the end date presented by the client as a formatted string
13277 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
13278 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +020013279
Willy Tarreau74ca5042013-06-11 23:12:07 +020013280ssl_c_notbefore : string
13281 Returns the start date presented by the client as a formatted string
13282 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
13283 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +010013284
Willy Tarreau74ca5042013-06-11 23:12:07 +020013285ssl_c_s_dn([<entry>[,<occ>]]) : string
13286 When the incoming connection was made over an SSL/TLS transport layer,
13287 returns the full distinguished name of the subject of the certificate
13288 presented by the client when no <entry> is specified, or the value of the
13289 first given entry found from the beginning of the DN. If a positive/negative
13290 occurrence number is specified as the optional second argument, it returns
13291 the value of the nth given entry value from the beginning/end of the DN.
13292 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
13293 "ssl_c_s_dn(CN)" retrieves the common name.
Willy Tarreaub6672b52011-12-12 17:23:41 +010013294
Willy Tarreau74ca5042013-06-11 23:12:07 +020013295ssl_c_serial : binary
13296 Returns the serial of the certificate presented by the client when the
13297 incoming connection was made over an SSL/TLS transport layer. When used for
13298 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +020013299
Willy Tarreau74ca5042013-06-11 23:12:07 +020013300ssl_c_sha1 : binary
13301 Returns the SHA-1 fingerprint of the certificate presented by the client when
13302 the incoming connection was made over an SSL/TLS transport layer. This can be
13303 used to stick a client to a server, or to pass this information to a server.
Willy Tarreau2d0caa32014-07-02 19:01:22 +020013304 Note that the output is binary, so if you want to pass that signature to the
13305 server, you need to encode it in hex or base64, such as in the example below:
13306
13307 http-request set-header X-SSL-Client-SHA1 %[ssl_c_sha1,hex]
Emeric Brun2525b6b2012-10-18 15:59:43 +020013308
Willy Tarreau74ca5042013-06-11 23:12:07 +020013309ssl_c_sig_alg : string
13310 Returns the name of the algorithm used to sign the certificate presented by
13311 the client when the incoming connection was made over an SSL/TLS transport
13312 layer.
Emeric Brun87855892012-10-17 17:39:35 +020013313
Willy Tarreau74ca5042013-06-11 23:12:07 +020013314ssl_c_used : boolean
13315 Returns true if current SSL session uses a client certificate even if current
13316 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +020013317
Willy Tarreau74ca5042013-06-11 23:12:07 +020013318ssl_c_verify : integer
13319 Returns the verify result error ID when the incoming connection was made over
13320 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
13321 refer to your SSL library's documentation for an exhaustive list of error
13322 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +020013323
Willy Tarreau74ca5042013-06-11 23:12:07 +020013324ssl_c_version : integer
13325 Returns the version of the certificate presented by the client when the
13326 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +020013327
Emeric Brun43e79582014-10-29 19:03:26 +010013328ssl_f_der : binary
13329 Returns the DER formatted certificate presented by the frontend when the
13330 incoming connection was made over an SSL/TLS transport layer. When used for
13331 an ACL, the value(s) to match against can be passed in hexadecimal form.
13332
Willy Tarreau74ca5042013-06-11 23:12:07 +020013333ssl_f_i_dn([<entry>[,<occ>]]) : string
13334 When the incoming connection was made over an SSL/TLS transport layer,
13335 returns the full distinguished name of the issuer of the certificate
13336 presented by the frontend when no <entry> is specified, or the value of the
13337 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +020013338 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +020013339 the value of the nth given entry value from the beginning/end of the DN.
13340 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
13341 "ssl_f_i_dn(CN)" retrieves the common name.
Emeric Brun87855892012-10-17 17:39:35 +020013342
Willy Tarreau74ca5042013-06-11 23:12:07 +020013343ssl_f_key_alg : string
13344 Returns the name of the algorithm used to generate the key of the certificate
13345 presented by the frontend when the incoming connection was made over an
13346 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020013347
Willy Tarreau74ca5042013-06-11 23:12:07 +020013348ssl_f_notafter : string
13349 Returns the end date presented by the frontend as a formatted string
13350 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
13351 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +020013352
Willy Tarreau74ca5042013-06-11 23:12:07 +020013353ssl_f_notbefore : string
13354 Returns the start date presented by the frontend as a formatted string
13355 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
13356 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +020013357
Willy Tarreau74ca5042013-06-11 23:12:07 +020013358ssl_f_s_dn([<entry>[,<occ>]]) : string
13359 When the incoming connection was made over an SSL/TLS transport layer,
13360 returns the full distinguished name of the subject of the certificate
13361 presented by the frontend when no <entry> is specified, or the value of the
13362 first given entry found from the beginning of the DN. If a positive/negative
13363 occurrence number is specified as the optional second argument, it returns
13364 the value of the nth given entry value from the beginning/end of the DN.
13365 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
13366 "ssl_f_s_dn(CN)" retrieves the common name.
Emeric Brunce5ad802012-10-22 14:11:22 +020013367
Willy Tarreau74ca5042013-06-11 23:12:07 +020013368ssl_f_serial : binary
13369 Returns the serial of the certificate presented by the frontend when the
13370 incoming connection was made over an SSL/TLS transport layer. When used for
13371 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +020013372
Emeric Brun55f4fa82014-04-30 17:11:25 +020013373ssl_f_sha1 : binary
13374 Returns the SHA-1 fingerprint of the certificate presented by the frontend
13375 when the incoming connection was made over an SSL/TLS transport layer. This
13376 can be used to know which certificate was chosen using SNI.
13377
Willy Tarreau74ca5042013-06-11 23:12:07 +020013378ssl_f_sig_alg : string
13379 Returns the name of the algorithm used to sign the certificate presented by
13380 the frontend when the incoming connection was made over an SSL/TLS transport
13381 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020013382
Willy Tarreau74ca5042013-06-11 23:12:07 +020013383ssl_f_version : integer
13384 Returns the version of the certificate presented by the frontend when the
13385 incoming connection was made over an SSL/TLS transport layer.
13386
13387ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020013388 Returns true when the front connection was made via an SSL/TLS transport
13389 layer and is locally deciphered. This means it has matched a socket declared
13390 with a "bind" line having the "ssl" option.
13391
Willy Tarreau74ca5042013-06-11 23:12:07 +020013392 Example :
13393 # This passes "X-Proto: https" to servers when client connects over SSL
13394 listen http-https
13395 bind :80
13396 bind :443 ssl crt /etc/haproxy.pem
13397 http-request add-header X-Proto https if { ssl_fc }
13398
13399ssl_fc_alg_keysize : integer
13400 Returns the symmetric cipher key size supported in bits when the incoming
13401 connection was made over an SSL/TLS transport layer.
13402
13403ssl_fc_alpn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013404 This extracts the Application Layer Protocol Negotiation field from an
Willy Tarreau74ca5042013-06-11 23:12:07 +020013405 incoming connection made via a TLS transport layer and locally deciphered by
13406 haproxy. The result is a string containing the protocol name advertised by
13407 the client. The SSL library must have been built with support for TLS
13408 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
13409 not advertised unless the "alpn" keyword on the "bind" line specifies a
13410 protocol list. Also, nothing forces the client to pick a protocol from this
13411 list, any other one may be requested. The TLS ALPN extension is meant to
13412 replace the TLS NPN extension. See also "ssl_fc_npn".
13413
Willy Tarreau74ca5042013-06-11 23:12:07 +020013414ssl_fc_cipher : string
13415 Returns the name of the used cipher when the incoming connection was made
13416 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +020013417
Willy Tarreau74ca5042013-06-11 23:12:07 +020013418ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020013419 Returns true if a client certificate is present in an incoming connection over
13420 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +010013421 Note: on SSL session resumption with Session ID or TLS ticket, client
13422 certificate is not present in the current connection but may be retrieved
13423 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
13424 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +020013425
Willy Tarreau74ca5042013-06-11 23:12:07 +020013426ssl_fc_has_sni : boolean
13427 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020013428 in an incoming connection was made over an SSL/TLS transport layer. Returns
13429 true when the incoming connection presents a TLS SNI field. This requires
13430 that the SSL library is build with support for TLS extensions enabled (check
13431 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +020013432
Nenad Merdanovic1516fe32016-05-17 03:31:21 +020013433ssl_fc_is_resumed : boolean
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020013434 Returns true if the SSL/TLS session has been resumed through the use of
13435 SSL session cache or TLS tickets.
13436
Willy Tarreau74ca5042013-06-11 23:12:07 +020013437ssl_fc_npn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013438 This extracts the Next Protocol Negotiation field from an incoming connection
Willy Tarreau74ca5042013-06-11 23:12:07 +020013439 made via a TLS transport layer and locally deciphered by haproxy. The result
13440 is a string containing the protocol name advertised by the client. The SSL
13441 library must have been built with support for TLS extensions enabled (check
13442 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
13443 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
13444 forces the client to pick a protocol from this list, any other one may be
13445 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +020013446
Willy Tarreau74ca5042013-06-11 23:12:07 +020013447ssl_fc_protocol : string
13448 Returns the name of the used protocol when the incoming connection was made
13449 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +020013450
Emeric Brunb73a9b02014-04-30 18:49:19 +020013451ssl_fc_unique_id : binary
David Sc1ad52e2014-04-08 18:48:47 -040013452 When the incoming connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020013453 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
13454 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
David Sc1ad52e2014-04-08 18:48:47 -040013455
Willy Tarreau74ca5042013-06-11 23:12:07 +020013456ssl_fc_session_id : binary
13457 Returns the SSL ID of the front connection when the incoming connection was
13458 made over an SSL/TLS transport layer. It is useful to stick a given client to
13459 a server. It is important to note that some browsers refresh their session ID
13460 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +020013461
Willy Tarreau74ca5042013-06-11 23:12:07 +020013462ssl_fc_sni : string
13463 This extracts the Server Name Indication TLS extension (SNI) field from an
13464 incoming connection made via an SSL/TLS transport layer and locally
13465 deciphered by haproxy. The result (when present) typically is a string
13466 matching the HTTPS host name (253 chars or less). The SSL library must have
13467 been built with support for TLS extensions enabled (check haproxy -vv).
13468
13469 This fetch is different from "req_ssl_sni" above in that it applies to the
13470 connection being deciphered by haproxy and not to SSL contents being blindly
13471 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +020013472 requires that the SSL library is build with support for TLS extensions
13473 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +020013474
Willy Tarreau74ca5042013-06-11 23:12:07 +020013475 ACL derivatives :
Willy Tarreau74ca5042013-06-11 23:12:07 +020013476 ssl_fc_sni_end : suffix match
13477 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +020013478
Willy Tarreau74ca5042013-06-11 23:12:07 +020013479ssl_fc_use_keysize : integer
13480 Returns the symmetric cipher key size used in bits when the incoming
13481 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020013482
Willy Tarreaub6fb4202008-07-20 11:18:28 +020013483
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200134847.3.5. Fetching samples from buffer contents (Layer 6)
Willy Tarreau74ca5042013-06-11 23:12:07 +020013485------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +020013486
Willy Tarreau74ca5042013-06-11 23:12:07 +020013487Fetching samples from buffer contents is a bit different from the previous
13488sample fetches above because the sampled data are ephemeral. These data can
13489only be used when they're available and will be lost when they're forwarded.
13490For this reason, samples fetched from buffer contents during a request cannot
13491be used in a response for example. Even while the data are being fetched, they
13492can change. Sometimes it is necessary to set some delays or combine multiple
13493sample fetch methods to ensure that the expected data are complete and usable,
13494for example through TCP request content inspection. Please see the "tcp-request
13495content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +020013496
Willy Tarreau74ca5042013-06-11 23:12:07 +020013497payload(<offset>,<length>) : binary (deprecated)
13498 This is an alias for "req.payload" when used in the context of a request (eg:
13499 "stick on", "stick match"), and for "res.payload" when used in the context of
13500 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010013501
Willy Tarreau74ca5042013-06-11 23:12:07 +020013502payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
13503 This is an alias for "req.payload_lv" when used in the context of a request
13504 (eg: "stick on", "stick match"), and for "res.payload_lv" when used in the
13505 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010013506
Willy Tarreau74ca5042013-06-11 23:12:07 +020013507req.len : integer
13508req_len : integer (deprecated)
13509 Returns an integer value corresponding to the number of bytes present in the
13510 request buffer. This is mostly used in ACL. It is important to understand
13511 that this test does not return false as long as the buffer is changing. This
13512 means that a check with equality to zero will almost always immediately match
13513 at the beginning of the session, while a test for more data will wait for
13514 that data to come in and return false only when haproxy is certain that no
13515 more data will come in. This test was designed to be used with TCP request
13516 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013517
Willy Tarreau74ca5042013-06-11 23:12:07 +020013518req.payload(<offset>,<length>) : binary
13519 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020013520 in the request buffer. As a special case, if the <length> argument is zero,
13521 the the whole buffer from <offset> to the end is extracted. This can be used
13522 with ACLs in order to check for the presence of some content in a buffer at
13523 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013524
Willy Tarreau74ca5042013-06-11 23:12:07 +020013525 ACL alternatives :
13526 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013527
Willy Tarreau74ca5042013-06-11 23:12:07 +020013528req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
13529 This extracts a binary block whose size is specified at <offset1> for <length>
13530 bytes, and which starts at <offset2> if specified or just after the length in
13531 the request buffer. The <offset2> parameter also supports relative offsets if
13532 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013533
Willy Tarreau74ca5042013-06-11 23:12:07 +020013534 ACL alternatives :
13535 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013536
Willy Tarreau74ca5042013-06-11 23:12:07 +020013537 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013538
Willy Tarreau74ca5042013-06-11 23:12:07 +020013539req.proto_http : boolean
13540req_proto_http : boolean (deprecated)
13541 Returns true when data in the request buffer look like HTTP and correctly
13542 parses as such. It is the same parser as the common HTTP request parser which
13543 is used so there should be no surprises. The test does not match until the
13544 request is complete, failed or timed out. This test may be used to report the
13545 protocol in TCP logs, but the biggest use is to block TCP request analysis
13546 until a complete HTTP request is present in the buffer, for example to track
13547 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013548
Willy Tarreau74ca5042013-06-11 23:12:07 +020013549 Example:
13550 # track request counts per "base" (concatenation of Host+URL)
13551 tcp-request inspect-delay 10s
13552 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020013553 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013554
Willy Tarreau74ca5042013-06-11 23:12:07 +020013555req.rdp_cookie([<name>]) : string
13556rdp_cookie([<name>]) : string (deprecated)
13557 When the request buffer looks like the RDP protocol, extracts the RDP cookie
13558 <name>, or any cookie if unspecified. The parser only checks for the first
13559 cookie, as illustrated in the RDP protocol specification. The cookie name is
13560 case insensitive. Generally the "MSTS" cookie name will be used, as it can
13561 contain the user name of the client connecting to the server if properly
13562 configured on the client. The "MSTSHASH" cookie is often used as well for
13563 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013564
Willy Tarreau74ca5042013-06-11 23:12:07 +020013565 This differs from "balance rdp-cookie" in that any balancing algorithm may be
13566 used and thus the distribution of clients to backend servers is not linked to
13567 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
13568 such as "balance roundrobin" or "balance leastconn" will lead to a more even
13569 distribution of clients to backend servers than the hash used by "balance
13570 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013571
Willy Tarreau74ca5042013-06-11 23:12:07 +020013572 ACL derivatives :
13573 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013574
Willy Tarreau74ca5042013-06-11 23:12:07 +020013575 Example :
13576 listen tse-farm
13577 bind 0.0.0.0:3389
13578 # wait up to 5s for an RDP cookie in the request
13579 tcp-request inspect-delay 5s
13580 tcp-request content accept if RDP_COOKIE
13581 # apply RDP cookie persistence
13582 persist rdp-cookie
13583 # Persist based on the mstshash cookie
13584 # This is only useful makes sense if
13585 # balance rdp-cookie is not used
13586 stick-table type string size 204800
13587 stick on req.rdp_cookie(mstshash)
13588 server srv1 1.1.1.1:3389
13589 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013590
Willy Tarreau74ca5042013-06-11 23:12:07 +020013591 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
13592 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013593
Willy Tarreau74ca5042013-06-11 23:12:07 +020013594req.rdp_cookie_cnt([name]) : integer
13595rdp_cookie_cnt([name]) : integer (deprecated)
13596 Tries to parse the request buffer as RDP protocol, then returns an integer
13597 corresponding to the number of RDP cookies found. If an optional cookie name
13598 is passed, only cookies matching this name are considered. This is mostly
13599 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013600
Willy Tarreau74ca5042013-06-11 23:12:07 +020013601 ACL derivatives :
13602 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013603
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020013604req.ssl_ec_ext : boolean
13605 Returns a boolean identifying if client sent the Supported Elliptic Curves
13606 Extension as defined in RFC4492, section 5.1. within the SSL ClientHello
Cyril Bonté307ee1e2015-09-28 23:16:06 +020013607 message. This can be used to present ECC compatible clients with EC
13608 certificate and to use RSA for all others, on the same IP address. Note that
13609 this only applies to raw contents found in the request buffer and not to
13610 contents deciphered via an SSL data layer, so this will not work with "bind"
13611 lines having the "ssl" option.
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020013612
Willy Tarreau74ca5042013-06-11 23:12:07 +020013613req.ssl_hello_type : integer
13614req_ssl_hello_type : integer (deprecated)
13615 Returns an integer value containing the type of the SSL hello message found
13616 in the request buffer if the buffer contains data that parse as a complete
13617 SSL (v3 or superior) client hello message. Note that this only applies to raw
13618 contents found in the request buffer and not to contents deciphered via an
13619 SSL data layer, so this will not work with "bind" lines having the "ssl"
13620 option. This is mostly used in ACL to detect presence of an SSL hello message
13621 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013622
Willy Tarreau74ca5042013-06-11 23:12:07 +020013623req.ssl_sni : string
13624req_ssl_sni : string (deprecated)
13625 Returns a string containing the value of the Server Name TLS extension sent
13626 by a client in a TLS stream passing through the request buffer if the buffer
13627 contains data that parse as a complete SSL (v3 or superior) client hello
13628 message. Note that this only applies to raw contents found in the request
13629 buffer and not to contents deciphered via an SSL data layer, so this will not
13630 work with "bind" lines having the "ssl" option. SNI normally contains the
13631 name of the host the client tries to connect to (for recent browsers). SNI is
13632 useful for allowing or denying access to certain hosts when SSL/TLS is used
13633 by the client. This test was designed to be used with TCP request content
13634 inspection. If content switching is needed, it is recommended to first wait
13635 for a complete client hello (type 1), like in the example below. See also
13636 "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013637
Willy Tarreau74ca5042013-06-11 23:12:07 +020013638 ACL derivatives :
13639 req_ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013640
Willy Tarreau74ca5042013-06-11 23:12:07 +020013641 Examples :
13642 # Wait for a client hello for at most 5 seconds
13643 tcp-request inspect-delay 5s
13644 tcp-request content accept if { req_ssl_hello_type 1 }
13645 use_backend bk_allow if { req_ssl_sni -f allowed_sites }
13646 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013647
Pradeep Jindalbb2acf52015-09-29 10:12:57 +053013648req.ssl_st_ext : integer
13649 Returns 0 if the client didn't send a SessionTicket TLS Extension (RFC5077)
13650 Returns 1 if the client sent SessionTicket TLS Extension
13651 Returns 2 if the client also sent non-zero length TLS SessionTicket
13652 Note that this only applies to raw contents found in the request buffer and
13653 not to contents deciphered via an SSL data layer, so this will not work with
13654 "bind" lines having the "ssl" option. This can for example be used to detect
13655 whether the client sent a SessionTicket or not and stick it accordingly, if
13656 no SessionTicket then stick on SessionID or don't stick as there's no server
13657 side state is there when SessionTickets are in use.
13658
Willy Tarreau74ca5042013-06-11 23:12:07 +020013659req.ssl_ver : integer
13660req_ssl_ver : integer (deprecated)
13661 Returns an integer value containing the version of the SSL/TLS protocol of a
13662 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
13663 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
13664 composed of the major version multiplied by 65536, added to the minor
13665 version. Note that this only applies to raw contents found in the request
13666 buffer and not to contents deciphered via an SSL data layer, so this will not
13667 work with "bind" lines having the "ssl" option. The ACL version of the test
13668 matches against a decimal notation in the form MAJOR.MINOR (eg: 3.1). This
13669 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013670
Willy Tarreau74ca5042013-06-11 23:12:07 +020013671 ACL derivatives :
13672 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +010013673
Willy Tarreau47e8eba2013-09-11 23:28:46 +020013674res.len : integer
13675 Returns an integer value corresponding to the number of bytes present in the
13676 response buffer. This is mostly used in ACL. It is important to understand
13677 that this test does not return false as long as the buffer is changing. This
13678 means that a check with equality to zero will almost always immediately match
13679 at the beginning of the session, while a test for more data will wait for
13680 that data to come in and return false only when haproxy is certain that no
13681 more data will come in. This test was designed to be used with TCP response
13682 content inspection.
13683
Willy Tarreau74ca5042013-06-11 23:12:07 +020013684res.payload(<offset>,<length>) : binary
13685 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020013686 in the response buffer. As a special case, if the <length> argument is zero,
13687 the the whole buffer from <offset> to the end is extracted. This can be used
13688 with ACLs in order to check for the presence of some content in a buffer at
13689 any location.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013690
Willy Tarreau74ca5042013-06-11 23:12:07 +020013691res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
13692 This extracts a binary block whose size is specified at <offset1> for <length>
13693 bytes, and which starts at <offset2> if specified or just after the length in
13694 the response buffer. The <offset2> parameter also supports relative offsets
13695 if prepended with a '+' or '-' sign.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013696
Willy Tarreau74ca5042013-06-11 23:12:07 +020013697 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013698
Willy Tarreau971f7b62015-09-29 14:06:59 +020013699res.ssl_hello_type : integer
13700rep_ssl_hello_type : integer (deprecated)
13701 Returns an integer value containing the type of the SSL hello message found
13702 in the response buffer if the buffer contains data that parses as a complete
13703 SSL (v3 or superior) hello message. Note that this only applies to raw
13704 contents found in the response buffer and not to contents deciphered via an
13705 SSL data layer, so this will not work with "server" lines having the "ssl"
13706 option. This is mostly used in ACL to detect presence of an SSL hello message
13707 that is supposed to contain an SSL session ID usable for stickiness.
13708
Willy Tarreau74ca5042013-06-11 23:12:07 +020013709wait_end : boolean
13710 This fetch either returns true when the inspection period is over, or does
13711 not fetch. It is only used in ACLs, in conjunction with content analysis to
13712 avoid returning a wrong verdict early. It may also be used to delay some
13713 actions, such as a delayed reject for some special addresses. Since it either
13714 stops the rules evaluation or immediately returns true, it is recommended to
13715 use this acl as the last one in a rule. Please note that the default ACL
13716 "WAIT_END" is always usable without prior declaration. This test was designed
13717 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013718
Willy Tarreau74ca5042013-06-11 23:12:07 +020013719 Examples :
13720 # delay every incoming request by 2 seconds
13721 tcp-request inspect-delay 2s
13722 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +010013723
Willy Tarreau74ca5042013-06-11 23:12:07 +020013724 # don't immediately tell bad guys they are rejected
13725 tcp-request inspect-delay 10s
13726 acl goodguys src 10.0.0.0/24
13727 acl badguys src 10.0.1.0/24
13728 tcp-request content accept if goodguys
13729 tcp-request content reject if badguys WAIT_END
13730 tcp-request content reject
13731
13732
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200137337.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +020013734--------------------------------------
13735
13736It is possible to fetch samples from HTTP contents, requests and responses.
13737This application layer is also called layer 7. It is only possible to fetch the
13738data in this section when a full HTTP request or response has been parsed from
13739its respective request or response buffer. This is always the case with all
13740HTTP specific rules and for sections running with "mode http". When using TCP
13741content inspection, it may be necessary to support an inspection delay in order
13742to let the request or response come in first. These fetches may require a bit
13743more CPU resources than the layer 4 ones, but not much since the request and
13744response are indexed.
13745
13746base : string
13747 This returns the concatenation of the first Host header and the path part of
13748 the request, which starts at the first slash and ends before the question
13749 mark. It can be useful in virtual hosted environments to detect URL abuses as
13750 well as to improve shared caches efficiency. Using this with a limited size
13751 stick table also allows one to collect statistics about most commonly
13752 requested objects by host/path. With ACLs it can allow simple content
13753 switching rules involving the host and the path at the same time, such as
13754 "www.example.com/favicon.ico". See also "path" and "uri".
13755
13756 ACL derivatives :
13757 base : exact string match
13758 base_beg : prefix match
13759 base_dir : subdir match
13760 base_dom : domain match
13761 base_end : suffix match
13762 base_len : length match
13763 base_reg : regex match
13764 base_sub : substring match
13765
13766base32 : integer
13767 This returns a 32-bit hash of the value returned by the "base" fetch method
13768 above. This is useful to track per-URL activity on high traffic sites without
13769 having to store all URLs. Instead a shorter hash is stored, saving a lot of
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013770 memory. The output type is an unsigned integer. The hash function used is
13771 SDBM with full avalanche on the output. Technically, base32 is exactly equal
13772 to "base,sdbm(1)".
Willy Tarreau74ca5042013-06-11 23:12:07 +020013773
13774base32+src : binary
13775 This returns the concatenation of the base32 fetch above and the src fetch
13776 below. The resulting type is of type binary, with a size of 8 or 20 bytes
13777 depending on the source address family. This can be used to track per-IP,
13778 per-URL counters.
13779
William Lallemand65ad6e12014-01-31 15:08:02 +010013780capture.req.hdr(<idx>) : string
13781 This extracts the content of the header captured by the "capture request
13782 header", idx is the position of the capture keyword in the configuration.
13783 The first entry is an index of 0. See also: "capture request header".
13784
13785capture.req.method : string
13786 This extracts the METHOD of an HTTP request. It can be used in both request
13787 and response. Unlike "method", it can be used in both request and response
13788 because it's allocated.
13789
13790capture.req.uri : string
13791 This extracts the request's URI, which starts at the first slash and ends
13792 before the first space in the request (without the host part). Unlike "path"
13793 and "url", it can be used in both request and response because it's
13794 allocated.
13795
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020013796capture.req.ver : string
13797 This extracts the request's HTTP version and returns either "HTTP/1.0" or
13798 "HTTP/1.1". Unlike "req.ver", it can be used in both request, response, and
13799 logs because it relies on a persistent flag.
13800
William Lallemand65ad6e12014-01-31 15:08:02 +010013801capture.res.hdr(<idx>) : string
13802 This extracts the content of the header captured by the "capture response
13803 header", idx is the position of the capture keyword in the configuration.
13804 The first entry is an index of 0.
13805 See also: "capture response header"
13806
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020013807capture.res.ver : string
13808 This extracts the response's HTTP version and returns either "HTTP/1.0" or
13809 "HTTP/1.1". Unlike "res.ver", it can be used in logs because it relies on a
13810 persistent flag.
13811
Willy Tarreaua5910cc2015-05-02 00:46:08 +020013812req.body : binary
13813 This returns the HTTP request's available body as a block of data. It
13814 requires that the request body has been buffered made available using
13815 "option http-buffer-request". In case of chunked-encoded body, currently only
13816 the first chunk is analyzed.
13817
Thierry FOURNIER9826c772015-05-20 15:50:54 +020013818req.body_param([<name>) : string
13819 This fetch assumes that the body of the POST request is url-encoded. The user
13820 can check if the "content-type" contains the value
13821 "application/x-www-form-urlencoded". This extracts the first occurrence of the
13822 parameter <name> in the body, which ends before '&'. The parameter name is
13823 case-sensitive. If no name is given, any parameter will match, and the first
13824 one will be returned. The result is a string corresponding to the value of the
13825 parameter <name> as presented in the request body (no URL decoding is
13826 performed). Note that the ACL version of this fetch iterates over multiple
13827 parameters and will iteratively report all parameters values if no name is
13828 given.
13829
Willy Tarreaua5910cc2015-05-02 00:46:08 +020013830req.body_len : integer
13831 This returns the length of the HTTP request's available body in bytes. It may
13832 be lower than the advertised length if the body is larger than the buffer. It
13833 requires that the request body has been buffered made available using
13834 "option http-buffer-request".
13835
13836req.body_size : integer
13837 This returns the advertised length of the HTTP request's body in bytes. It
13838 will represent the advertised Content-Length header, or the size of the first
13839 chunk in case of chunked encoding. In order to parse the chunks, it requires
13840 that the request body has been buffered made available using
13841 "option http-buffer-request".
13842
Willy Tarreau74ca5042013-06-11 23:12:07 +020013843req.cook([<name>]) : string
13844cook([<name>]) : string (deprecated)
13845 This extracts the last occurrence of the cookie name <name> on a "Cookie"
13846 header line from the request, and returns its value as string. If no name is
13847 specified, the first cookie value is returned. When used with ACLs, all
13848 matching cookies are evaluated. Spaces around the name and the value are
13849 ignored as requested by the Cookie header specification (RFC6265). The cookie
13850 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
13851 well return an empty value if it is present. Use the "found" match to detect
13852 presence. Use the res.cook() variant for response cookies sent by the server.
13853
13854 ACL derivatives :
13855 cook([<name>]) : exact string match
13856 cook_beg([<name>]) : prefix match
13857 cook_dir([<name>]) : subdir match
13858 cook_dom([<name>]) : domain match
13859 cook_end([<name>]) : suffix match
13860 cook_len([<name>]) : length match
13861 cook_reg([<name>]) : regex match
13862 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +010013863
Willy Tarreau74ca5042013-06-11 23:12:07 +020013864req.cook_cnt([<name>]) : integer
13865cook_cnt([<name>]) : integer (deprecated)
13866 Returns an integer value representing the number of occurrences of the cookie
13867 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013868
Willy Tarreau74ca5042013-06-11 23:12:07 +020013869req.cook_val([<name>]) : integer
13870cook_val([<name>]) : integer (deprecated)
13871 This extracts the last occurrence of the cookie name <name> on a "Cookie"
13872 header line from the request, and converts its value to an integer which is
13873 returned. If no name is specified, the first cookie value is returned. When
13874 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +020013875
Willy Tarreau74ca5042013-06-11 23:12:07 +020013876cookie([<name>]) : string (deprecated)
13877 This extracts the last occurrence of the cookie name <name> on a "Cookie"
13878 header line from the request, or a "Set-Cookie" header from the response, and
13879 returns its value as a string. A typical use is to get multiple clients
13880 sharing a same profile use the same server. This can be similar to what
Willy Tarreau294d0f02015-08-10 19:40:12 +020013881 "appsession" did with the "request-learn" statement, but with support for
Willy Tarreau74ca5042013-06-11 23:12:07 +020013882 multi-peer synchronization and state keeping across restarts. If no name is
13883 specified, the first cookie value is returned. This fetch should not be used
13884 anymore and should be replaced by req.cook() or res.cook() instead as it
13885 ambiguously uses the direction based on the context where it is used.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013886
Willy Tarreau74ca5042013-06-11 23:12:07 +020013887hdr([<name>[,<occ>]]) : string
13888 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
13889 used on responses. Please refer to these respective fetches for more details.
13890 In case of doubt about the fetch direction, please use the explicit ones.
13891 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013892 unambiguously apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013893
Willy Tarreau74ca5042013-06-11 23:12:07 +020013894req.fhdr(<name>[,<occ>]) : string
13895 This extracts the last occurrence of header <name> in an HTTP request. When
13896 used from an ACL, all occurrences are iterated over until a match is found.
13897 Optionally, a specific occurrence might be specified as a position number.
13898 Positive values indicate a position from the first occurrence, with 1 being
13899 the first one. Negative values indicate positions relative to the last one,
13900 with -1 being the last one. It differs from req.hdr() in that any commas
13901 present in the value are returned and are not used as delimiters. This is
13902 sometimes useful with headers such as User-Agent.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013903
Willy Tarreau74ca5042013-06-11 23:12:07 +020013904req.fhdr_cnt([<name>]) : integer
13905 Returns an integer value representing the number of occurrences of request
13906 header field name <name>, or the total number of header fields if <name> is
13907 not specified. Contrary to its req.hdr_cnt() cousin, this function returns
13908 the number of full line headers and does not stop on commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013909
Willy Tarreau74ca5042013-06-11 23:12:07 +020013910req.hdr([<name>[,<occ>]]) : string
13911 This extracts the last occurrence of header <name> in an HTTP request. When
13912 used from an ACL, all occurrences are iterated over until a match is found.
13913 Optionally, a specific occurrence might be specified as a position number.
13914 Positive values indicate a position from the first occurrence, with 1 being
13915 the first one. Negative values indicate positions relative to the last one,
13916 with -1 being the last one. A typical use is with the X-Forwarded-For header
13917 once converted to IP, associated with an IP stick-table. The function
13918 considers any comma as a delimiter for distinct values. If full-line headers
13919 are desired instead, use req.fhdr(). Please carefully check RFC2616 to know
13920 how certain headers are supposed to be parsed. Also, some of them are case
13921 insensitive (eg: Connection).
Willy Tarreaud63335a2010-02-26 12:56:52 +010013922
Willy Tarreau74ca5042013-06-11 23:12:07 +020013923 ACL derivatives :
13924 hdr([<name>[,<occ>]]) : exact string match
13925 hdr_beg([<name>[,<occ>]]) : prefix match
13926 hdr_dir([<name>[,<occ>]]) : subdir match
13927 hdr_dom([<name>[,<occ>]]) : domain match
13928 hdr_end([<name>[,<occ>]]) : suffix match
13929 hdr_len([<name>[,<occ>]]) : length match
13930 hdr_reg([<name>[,<occ>]]) : regex match
13931 hdr_sub([<name>[,<occ>]]) : substring match
13932
13933req.hdr_cnt([<name>]) : integer
13934hdr_cnt([<header>]) : integer (deprecated)
13935 Returns an integer value representing the number of occurrences of request
13936 header field name <name>, or the total number of header field values if
13937 <name> is not specified. It is important to remember that one header line may
13938 count as several headers if it has several values. The function considers any
13939 comma as a delimiter for distinct values. If full-line headers are desired
13940 instead, req.fhdr_cnt() should be used instead. With ACLs, it can be used to
13941 detect presence, absence or abuse of a specific header, as well as to block
13942 request smuggling attacks by rejecting requests which contain more than one
13943 of certain headers. See "req.hdr" for more information on header matching.
13944
13945req.hdr_ip([<name>[,<occ>]]) : ip
13946hdr_ip([<name>[,<occ>]]) : ip (deprecated)
13947 This extracts the last occurrence of header <name> in an HTTP request,
13948 converts it to an IPv4 or IPv6 address and returns this address. When used
13949 with ACLs, all occurrences are checked, and if <name> is omitted, every value
13950 of every header is checked. Optionally, a specific occurrence might be
13951 specified as a position number. Positive values indicate a position from the
13952 first occurrence, with 1 being the first one. Negative values indicate
13953 positions relative to the last one, with -1 being the last one. A typical use
13954 is with the X-Forwarded-For and X-Client-IP headers.
13955
13956req.hdr_val([<name>[,<occ>]]) : integer
13957hdr_val([<name>[,<occ>]]) : integer (deprecated)
13958 This extracts the last occurrence of header <name> in an HTTP request, and
13959 converts it to an integer value. When used with ACLs, all occurrences are
13960 checked, and if <name> is omitted, every value of every header is checked.
13961 Optionally, a specific occurrence might be specified as a position number.
13962 Positive values indicate a position from the first occurrence, with 1 being
13963 the first one. Negative values indicate positions relative to the last one,
13964 with -1 being the last one. A typical use is with the X-Forwarded-For header.
13965
13966http_auth(<userlist>) : boolean
13967 Returns a boolean indicating whether the authentication data received from
13968 the client match a username & password stored in the specified userlist. This
13969 fetch function is not really useful outside of ACLs. Currently only http
13970 basic auth is supported.
13971
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010013972http_auth_group(<userlist>) : string
13973 Returns a string corresponding to the user name found in the authentication
13974 data received from the client if both the user name and password are valid
13975 according to the specified userlist. The main purpose is to use it in ACLs
13976 where it is then checked whether the user belongs to any group within a list.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013977 This fetch function is not really useful outside of ACLs. Currently only http
13978 basic auth is supported.
13979
13980 ACL derivatives :
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010013981 http_auth_group(<userlist>) : group ...
13982 Returns true when the user extracted from the request and whose password is
13983 valid according to the specified userlist belongs to at least one of the
13984 groups.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013985
13986http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020013987 Returns true when the request being processed is the first one of the
13988 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020013989 from some requests when a request is not the first one, or to help grouping
13990 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020013991
Willy Tarreau74ca5042013-06-11 23:12:07 +020013992method : integer + string
13993 Returns an integer value corresponding to the method in the HTTP request. For
13994 example, "GET" equals 1 (check sources to establish the matching). Value 9
13995 means "other method" and may be converted to a string extracted from the
13996 stream. This should not be used directly as a sample, this is only meant to
13997 be used from ACLs, which transparently convert methods from patterns to these
13998 integer + string values. Some predefined ACL already check for most common
13999 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014000
Willy Tarreau74ca5042013-06-11 23:12:07 +020014001 ACL derivatives :
14002 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020014003
Willy Tarreau74ca5042013-06-11 23:12:07 +020014004 Example :
14005 # only accept GET and HEAD requests
14006 acl valid_method method GET HEAD
14007 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020014008
Willy Tarreau74ca5042013-06-11 23:12:07 +020014009path : string
14010 This extracts the request's URL path, which starts at the first slash and
14011 ends before the question mark (without the host part). A typical use is with
14012 prefetch-capable caches, and with portals which need to aggregate multiple
14013 information from databases and keep them in caches. Note that with outgoing
14014 caches, it would be wiser to use "url" instead. With ACLs, it's typically
14015 used to match exact file names (eg: "/login.php"), or directory parts using
14016 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014017
Willy Tarreau74ca5042013-06-11 23:12:07 +020014018 ACL derivatives :
14019 path : exact string match
14020 path_beg : prefix match
14021 path_dir : subdir match
14022 path_dom : domain match
14023 path_end : suffix match
14024 path_len : length match
14025 path_reg : regex match
14026 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020014027
Willy Tarreau49ad95c2015-01-19 15:06:26 +010014028query : string
14029 This extracts the request's query string, which starts after the first
14030 question mark. If no question mark is present, this fetch returns nothing. If
14031 a question mark is present but nothing follows, it returns an empty string.
14032 This means it's possible to easily know whether a query string is present
14033 using the "found" matching method. This fetch is the completemnt of "path"
14034 which stops before the question mark.
14035
Willy Tarreaueb27ec72015-02-20 13:55:29 +010014036req.hdr_names([<delim>]) : string
14037 This builds a string made from the concatenation of all header names as they
14038 appear in the request when the rule is evaluated. The default delimiter is
14039 the comma (',') but it may be overridden as an optional argument <delim>. In
14040 this case, only the first character of <delim> is considered.
14041
Willy Tarreau74ca5042013-06-11 23:12:07 +020014042req.ver : string
14043req_ver : string (deprecated)
14044 Returns the version string from the HTTP request, for example "1.1". This can
14045 be useful for logs, but is mostly there for ACL. Some predefined ACL already
14046 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014047
Willy Tarreau74ca5042013-06-11 23:12:07 +020014048 ACL derivatives :
14049 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020014050
Willy Tarreau74ca5042013-06-11 23:12:07 +020014051res.comp : boolean
14052 Returns the boolean "true" value if the response has been compressed by
14053 HAProxy, otherwise returns boolean "false". This may be used to add
14054 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014055
Willy Tarreau74ca5042013-06-11 23:12:07 +020014056res.comp_algo : string
14057 Returns a string containing the name of the algorithm used if the response
14058 was compressed by HAProxy, for example : "deflate". This may be used to add
14059 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014060
Willy Tarreau74ca5042013-06-11 23:12:07 +020014061res.cook([<name>]) : string
14062scook([<name>]) : string (deprecated)
14063 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
14064 header line from the response, and returns its value as string. If no name is
14065 specified, the first cookie value is returned.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020014066
Willy Tarreau74ca5042013-06-11 23:12:07 +020014067 ACL derivatives :
14068 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020014069
Willy Tarreau74ca5042013-06-11 23:12:07 +020014070res.cook_cnt([<name>]) : integer
14071scook_cnt([<name>]) : integer (deprecated)
14072 Returns an integer value representing the number of occurrences of the cookie
14073 <name> in the response, or all cookies if <name> is not specified. This is
14074 mostly useful when combined with ACLs to detect suspicious responses.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014075
Willy Tarreau74ca5042013-06-11 23:12:07 +020014076res.cook_val([<name>]) : integer
14077scook_val([<name>]) : integer (deprecated)
14078 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
14079 header line from the response, and converts its value to an integer which is
14080 returned. If no name is specified, the first cookie value is returned.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014081
Willy Tarreau74ca5042013-06-11 23:12:07 +020014082res.fhdr([<name>[,<occ>]]) : string
14083 This extracts the last occurrence of header <name> in an HTTP response, or of
14084 the last header if no <name> is specified. Optionally, a specific occurrence
14085 might be specified as a position number. Positive values indicate a position
14086 from the first occurrence, with 1 being the first one. Negative values
14087 indicate positions relative to the last one, with -1 being the last one. It
14088 differs from res.hdr() in that any commas present in the value are returned
14089 and are not used as delimiters. If this is not desired, the res.hdr() fetch
14090 should be used instead. This is sometimes useful with headers such as Date or
14091 Expires.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014092
Willy Tarreau74ca5042013-06-11 23:12:07 +020014093res.fhdr_cnt([<name>]) : integer
14094 Returns an integer value representing the number of occurrences of response
14095 header field name <name>, or the total number of header fields if <name> is
14096 not specified. Contrary to its res.hdr_cnt() cousin, this function returns
14097 the number of full line headers and does not stop on commas. If this is not
14098 desired, the res.hdr_cnt() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014099
Willy Tarreau74ca5042013-06-11 23:12:07 +020014100res.hdr([<name>[,<occ>]]) : string
14101shdr([<name>[,<occ>]]) : string (deprecated)
14102 This extracts the last occurrence of header <name> in an HTTP response, or of
14103 the last header if no <name> is specified. Optionally, a specific occurrence
14104 might be specified as a position number. Positive values indicate a position
14105 from the first occurrence, with 1 being the first one. Negative values
14106 indicate positions relative to the last one, with -1 being the last one. This
14107 can be useful to learn some data into a stick-table. The function considers
14108 any comma as a delimiter for distinct values. If this is not desired, the
14109 res.fhdr() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014110
Willy Tarreau74ca5042013-06-11 23:12:07 +020014111 ACL derivatives :
14112 shdr([<name>[,<occ>]]) : exact string match
14113 shdr_beg([<name>[,<occ>]]) : prefix match
14114 shdr_dir([<name>[,<occ>]]) : subdir match
14115 shdr_dom([<name>[,<occ>]]) : domain match
14116 shdr_end([<name>[,<occ>]]) : suffix match
14117 shdr_len([<name>[,<occ>]]) : length match
14118 shdr_reg([<name>[,<occ>]]) : regex match
14119 shdr_sub([<name>[,<occ>]]) : substring match
14120
14121res.hdr_cnt([<name>]) : integer
14122shdr_cnt([<name>]) : integer (deprecated)
14123 Returns an integer value representing the number of occurrences of response
14124 header field name <name>, or the total number of header fields if <name> is
14125 not specified. The function considers any comma as a delimiter for distinct
14126 values. If this is not desired, the res.fhdr_cnt() fetch should be used
14127 instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014128
Willy Tarreau74ca5042013-06-11 23:12:07 +020014129res.hdr_ip([<name>[,<occ>]]) : ip
14130shdr_ip([<name>[,<occ>]]) : ip (deprecated)
14131 This extracts the last occurrence of header <name> in an HTTP response,
14132 convert it to an IPv4 or IPv6 address and returns this address. Optionally, a
14133 specific occurrence might be specified as a position number. Positive values
14134 indicate a position from the first occurrence, with 1 being the first one.
14135 Negative values indicate positions relative to the last one, with -1 being
14136 the last one. This can be useful to learn some data into a stick table.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014137
Willy Tarreaueb27ec72015-02-20 13:55:29 +010014138res.hdr_names([<delim>]) : string
14139 This builds a string made from the concatenation of all header names as they
14140 appear in the response when the rule is evaluated. The default delimiter is
14141 the comma (',') but it may be overridden as an optional argument <delim>. In
14142 this case, only the first character of <delim> is considered.
14143
Willy Tarreau74ca5042013-06-11 23:12:07 +020014144res.hdr_val([<name>[,<occ>]]) : integer
14145shdr_val([<name>[,<occ>]]) : integer (deprecated)
14146 This extracts the last occurrence of header <name> in an HTTP response, and
14147 converts it to an integer value. Optionally, a specific occurrence might be
14148 specified as a position number. Positive values indicate a position from the
14149 first occurrence, with 1 being the first one. Negative values indicate
14150 positions relative to the last one, with -1 being the last one. This can be
14151 useful to learn some data into a stick table.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010014152
Willy Tarreau74ca5042013-06-11 23:12:07 +020014153res.ver : string
14154resp_ver : string (deprecated)
14155 Returns the version string from the HTTP response, for example "1.1". This
14156 can be useful for logs, but is mostly there for ACL.
Willy Tarreau0e698542011-09-16 08:32:32 +020014157
Willy Tarreau74ca5042013-06-11 23:12:07 +020014158 ACL derivatives :
14159 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010014160
Willy Tarreau74ca5042013-06-11 23:12:07 +020014161set-cookie([<name>]) : string (deprecated)
14162 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
14163 header line from the response and uses the corresponding value to match. This
Willy Tarreau294d0f02015-08-10 19:40:12 +020014164 can be comparable to what "appsession" did with default options, but with
Willy Tarreau74ca5042013-06-11 23:12:07 +020014165 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010014166
Willy Tarreau74ca5042013-06-11 23:12:07 +020014167 This fetch function is deprecated and has been superseded by the "res.cook"
14168 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010014169
Willy Tarreau74ca5042013-06-11 23:12:07 +020014170status : integer
14171 Returns an integer containing the HTTP status code in the HTTP response, for
14172 example, 302. It is mostly used within ACLs and integer ranges, for example,
14173 to remove any Location header if the response is not a 3xx.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014174
Thierry Fournier0e00dca2016-04-07 15:47:40 +020014175unique-id : string
14176 Returns the unique-id attached to the request. The directive
14177 "unique-id-format" must be set. If it is not set, the unique-id sample fetch
14178 fails. Note that the unique-id is usually used with HTTP requests, however this
14179 sample fetch can be used with other protocols. Obviously, if it is used with
14180 other protocols than HTTP, the unique-id-format directive must not contain
14181 HTTP parts. See: unique-id-format and unique-id-header
14182
Willy Tarreau74ca5042013-06-11 23:12:07 +020014183url : string
14184 This extracts the request's URL as presented in the request. A typical use is
14185 with prefetch-capable caches, and with portals which need to aggregate
14186 multiple information from databases and keep them in caches. With ACLs, using
14187 "path" is preferred over using "url", because clients may send a full URL as
14188 is normally done with proxies. The only real use is to match "*" which does
14189 not match in "path", and for which there is already a predefined ACL. See
14190 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014191
Willy Tarreau74ca5042013-06-11 23:12:07 +020014192 ACL derivatives :
14193 url : exact string match
14194 url_beg : prefix match
14195 url_dir : subdir match
14196 url_dom : domain match
14197 url_end : suffix match
14198 url_len : length match
14199 url_reg : regex match
14200 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014201
Willy Tarreau74ca5042013-06-11 23:12:07 +020014202url_ip : ip
14203 This extracts the IP address from the request's URL when the host part is
14204 presented as an IP address. Its use is very limited. For instance, a
14205 monitoring system might use this field as an alternative for the source IP in
14206 order to test what path a given source address would follow, or to force an
14207 entry in a table for a given source address. With ACLs it can be used to
14208 restrict access to certain systems through a proxy, for example when combined
14209 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014210
Willy Tarreau74ca5042013-06-11 23:12:07 +020014211url_port : integer
14212 This extracts the port part from the request's URL. Note that if the port is
14213 not specified in the request, port 80 is assumed. With ACLs it can be used to
14214 restrict access to certain systems through a proxy, for example when combined
14215 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014216
Willy Tarreau1ede1da2015-05-07 16:06:18 +020014217urlp([<name>[,<delim>]]) : string
14218url_param([<name>[,<delim>]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020014219 This extracts the first occurrence of the parameter <name> in the query
14220 string, which begins after either '?' or <delim>, and which ends before '&',
Willy Tarreau1ede1da2015-05-07 16:06:18 +020014221 ';' or <delim>. The parameter name is case-sensitive. If no name is given,
14222 any parameter will match, and the first one will be returned. The result is
14223 a string corresponding to the value of the parameter <name> as presented in
14224 the request (no URL decoding is performed). This can be used for session
Willy Tarreau74ca5042013-06-11 23:12:07 +020014225 stickiness based on a client ID, to extract an application cookie passed as a
14226 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
Willy Tarreau1ede1da2015-05-07 16:06:18 +020014227 this fetch iterates over multiple parameters and will iteratively report all
14228 parameters values if no name is given
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014229
Willy Tarreau74ca5042013-06-11 23:12:07 +020014230 ACL derivatives :
14231 urlp(<name>[,<delim>]) : exact string match
14232 urlp_beg(<name>[,<delim>]) : prefix match
14233 urlp_dir(<name>[,<delim>]) : subdir match
14234 urlp_dom(<name>[,<delim>]) : domain match
14235 urlp_end(<name>[,<delim>]) : suffix match
14236 urlp_len(<name>[,<delim>]) : length match
14237 urlp_reg(<name>[,<delim>]) : regex match
14238 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014239
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014240
Willy Tarreau74ca5042013-06-11 23:12:07 +020014241 Example :
14242 # match http://example.com/foo?PHPSESSIONID=some_id
14243 stick on urlp(PHPSESSIONID)
14244 # match http://example.com/foo;JSESSIONID=some_id
14245 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014246
Willy Tarreau1ede1da2015-05-07 16:06:18 +020014247urlp_val([<name>[,<delim>])] : integer
Willy Tarreau74ca5042013-06-11 23:12:07 +020014248 See "urlp" above. This one extracts the URL parameter <name> in the request
14249 and converts it to an integer value. This can be used for session stickiness
14250 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020014251
Dragan Dosen0070cd52016-06-16 12:19:49 +020014252url32 : integer
14253 This returns a 32-bit hash of the value obtained by concatenating the first
14254 Host header and the whole URL including parameters (not only the path part of
14255 the request, as in the "base32" fetch above). This is useful to track per-URL
14256 activity. A shorter hash is stored, saving a lot of memory. The output type
14257 is an unsigned integer.
14258
14259url32+src : binary
14260 This returns the concatenation of the "url32" fetch and the "src" fetch. The
14261 resulting type is of type binary, with a size of 8 or 20 bytes depending on
14262 the source address family. This can be used to track per-IP, per-URL counters.
14263
Willy Tarreau198a7442008-01-17 12:05:32 +010014264
Willy Tarreau74ca5042013-06-11 23:12:07 +0200142657.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014266---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010014267
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014268Some predefined ACLs are hard-coded so that they do not have to be declared in
14269every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020014270order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010014271
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014272ACL name Equivalent to Usage
14273---------------+-----------------------------+---------------------------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014274FALSE always_false never match
Willy Tarreau2492d5b2009-07-11 00:06:00 +020014275HTTP req_proto_http match if protocol is valid HTTP
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014276HTTP_1.0 req_ver 1.0 match HTTP version 1.0
14277HTTP_1.1 req_ver 1.1 match HTTP version 1.1
Willy Tarreaud63335a2010-02-26 12:56:52 +010014278HTTP_CONTENT hdr_val(content-length) gt 0 match an existing content-length
14279HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
14280HTTP_URL_SLASH url_beg / match URL beginning with "/"
14281HTTP_URL_STAR url * match URL equal to "*"
14282LOCALHOST src 127.0.0.1/8 match connection from local host
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014283METH_CONNECT method CONNECT match HTTP CONNECT method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020014284METH_DELETE method DELETE match HTTP DELETE method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014285METH_GET method GET HEAD match HTTP GET or HEAD method
14286METH_HEAD method HEAD match HTTP HEAD method
14287METH_OPTIONS method OPTIONS match HTTP OPTIONS method
14288METH_POST method POST match HTTP POST method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020014289METH_PUT method PUT match HTTP PUT method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014290METH_TRACE method TRACE match HTTP TRACE method
Emeric Brunbede3d02009-06-30 17:54:00 +020014291RDP_COOKIE req_rdp_cookie_cnt gt 0 match presence of an RDP cookie
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014292REQ_CONTENT req_len gt 0 match data in the request buffer
Willy Tarreaud63335a2010-02-26 12:56:52 +010014293TRUE always_true always match
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014294WAIT_END wait_end wait for end of content analysis
14295---------------+-----------------------------+---------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010014296
Willy Tarreaub937b7e2010-01-12 15:27:54 +010014297
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200142988. Logging
14299----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010014300
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014301One of HAProxy's strong points certainly lies is its precise logs. It probably
14302provides the finest level of information available for such a product, which is
14303very important for troubleshooting complex environments. Standard information
14304provided in logs include client ports, TCP/HTTP state timers, precise session
14305state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010014306to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014307headers.
14308
14309In order to improve administrators reactivity, it offers a great transparency
14310about encountered problems, both internal and external, and it is possible to
14311send logs to different sources at the same time with different level filters :
14312
14313 - global process-level logs (system errors, start/stop, etc..)
14314 - per-instance system and internal errors (lack of resource, bugs, ...)
14315 - per-instance external troubles (servers up/down, max connections)
14316 - per-instance activity (client connections), either at the establishment or
14317 at the termination.
Jim Freeman9e8714b2015-05-26 09:16:34 -060014318 - per-request control of log-level, eg:
14319 http-request set-log-level silent if sensitive_request
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014320
14321The ability to distribute different levels of logs to different log servers
14322allow several production teams to interact and to fix their problems as soon
14323as possible. For example, the system team might monitor system-wide errors,
14324while the application team might be monitoring the up/down for their servers in
14325real time, and the security team might analyze the activity logs with one hour
14326delay.
14327
14328
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200143298.1. Log levels
14330---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014331
Simon Hormandf791f52011-05-29 15:01:10 +090014332TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014333source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090014334HTTP request, HTTP return code, number of bytes transmitted, conditions
14335in which the session ended, and even exchanged cookies values. For example
14336track a particular user's problems. All messages may be sent to up to two
14337syslog servers. Check the "log" keyword in section 4.2 for more information
14338about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014339
14340
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200143418.2. Log formats
14342----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014343
William Lallemand48940402012-01-30 16:47:22 +010014344HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090014345and will be detailed in the following sections. A few of them may vary
14346slightly with the configuration, due to indicators specific to certain
14347options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014348
14349 - the default format, which is very basic and very rarely used. It only
14350 provides very basic information about the incoming connection at the moment
14351 it is accepted : source IP:port, destination IP:port, and frontend-name.
14352 This mode will eventually disappear so it will not be described to great
14353 extents.
14354
14355 - the TCP format, which is more advanced. This format is enabled when "option
14356 tcplog" is set on the frontend. HAProxy will then usually wait for the
14357 connection to terminate before logging. This format provides much richer
14358 information, such as timers, connection counts, queue size, etc... This
14359 format is recommended for pure TCP proxies.
14360
14361 - the HTTP format, which is the most advanced for HTTP proxying. This format
14362 is enabled when "option httplog" is set on the frontend. It provides the
14363 same information as the TCP format with some HTTP-specific fields such as
14364 the request, the status code, and captures of headers and cookies. This
14365 format is recommended for HTTP proxies.
14366
Emeric Brun3a058f32009-06-30 18:26:00 +020014367 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
14368 fields arranged in the same order as the CLF format. In this mode, all
14369 timers, captures, flags, etc... appear one per field after the end of the
14370 common fields, in the same order they appear in the standard HTTP format.
14371
William Lallemand48940402012-01-30 16:47:22 +010014372 - the custom log format, allows you to make your own log line.
14373
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014374Next sections will go deeper into details for each of these formats. Format
14375specification will be performed on a "field" basis. Unless stated otherwise, a
14376field is a portion of text delimited by any number of spaces. Since syslog
14377servers are susceptible of inserting fields at the beginning of a line, it is
14378always assumed that the first field is the one containing the process name and
14379identifier.
14380
14381Note : Since log lines may be quite long, the log examples in sections below
14382 might be broken into multiple lines. The example log lines will be
14383 prefixed with 3 closing angle brackets ('>>>') and each time a log is
14384 broken into multiple lines, each non-final line will end with a
14385 backslash ('\') and the next line will start indented by two characters.
14386
14387
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200143888.2.1. Default log format
14389-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014390
14391This format is used when no specific option is set. The log is emitted as soon
14392as the connection is accepted. One should note that this currently is the only
14393format which logs the request's destination IP and ports.
14394
14395 Example :
14396 listen www
14397 mode http
14398 log global
14399 server srv1 127.0.0.1:8000
14400
14401 >>> Feb 6 12:12:09 localhost \
14402 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
14403 (www/HTTP)
14404
14405 Field Format Extract from the example above
14406 1 process_name '[' pid ']:' haproxy[14385]:
14407 2 'Connect from' Connect from
14408 3 source_ip ':' source_port 10.0.1.2:33312
14409 4 'to' to
14410 5 destination_ip ':' destination_port 10.0.3.31:8012
14411 6 '(' frontend_name '/' mode ')' (www/HTTP)
14412
14413Detailed fields description :
14414 - "source_ip" is the IP address of the client which initiated the connection.
14415 - "source_port" is the TCP port of the client which initiated the connection.
14416 - "destination_ip" is the IP address the client connected to.
14417 - "destination_port" is the TCP port the client connected to.
14418 - "frontend_name" is the name of the frontend (or listener) which received
14419 and processed the connection.
14420 - "mode is the mode the frontend is operating (TCP or HTTP).
14421
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014422In case of a UNIX socket, the source and destination addresses are marked as
14423"unix:" and the ports reflect the internal ID of the socket which accepted the
14424connection (the same ID as reported in the stats).
14425
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014426It is advised not to use this deprecated format for newer installations as it
14427will eventually disappear.
14428
14429
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200144308.2.2. TCP log format
14431---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014432
14433The TCP format is used when "option tcplog" is specified in the frontend, and
14434is the recommended format for pure TCP proxies. It provides a lot of precious
14435information for troubleshooting. Since this format includes timers and byte
14436counts, the log is normally emitted at the end of the session. It can be
14437emitted earlier if "option logasap" is specified, which makes sense in most
14438environments with long sessions such as remote terminals. Sessions which match
14439the "monitor" rules are never logged. It is also possible not to emit logs for
14440sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020014441specifying "option dontlognull" in the frontend. Successful connections will
14442not be logged if "option dontlog-normal" is specified in the frontend. A few
14443fields may slightly vary depending on some configuration options, those are
14444marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014445
14446 Example :
14447 frontend fnt
14448 mode tcp
14449 option tcplog
14450 log global
14451 default_backend bck
14452
14453 backend bck
14454 server srv1 127.0.0.1:8000
14455
14456 >>> Feb 6 12:12:56 localhost \
14457 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
14458 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
14459
14460 Field Format Extract from the example above
14461 1 process_name '[' pid ']:' haproxy[14387]:
14462 2 client_ip ':' client_port 10.0.1.2:33313
14463 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
14464 4 frontend_name fnt
14465 5 backend_name '/' server_name bck/srv1
14466 6 Tw '/' Tc '/' Tt* 0/0/5007
14467 7 bytes_read* 212
14468 8 termination_state --
14469 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
14470 10 srv_queue '/' backend_queue 0/0
14471
14472Detailed fields description :
14473 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014474 connection to haproxy. If the connection was accepted on a UNIX socket
14475 instead, the IP address would be replaced with the word "unix". Note that
14476 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010014477 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
14478 and the NetScaler Client IP insetion protocol is correctly used, then the
14479 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014480
14481 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014482 If the connection was accepted on a UNIX socket instead, the port would be
14483 replaced with the ID of the accepting socket, which is also reported in the
14484 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014485
14486 - "accept_date" is the exact date when the connection was received by haproxy
14487 (which might be very slightly different from the date observed on the
14488 network if there was some queuing in the system's backlog). This is usually
14489 the same date which may appear in any upstream firewall's log.
14490
14491 - "frontend_name" is the name of the frontend (or listener) which received
14492 and processed the connection.
14493
14494 - "backend_name" is the name of the backend (or listener) which was selected
14495 to manage the connection to the server. This will be the same as the
14496 frontend if no switching rule has been applied, which is common for TCP
14497 applications.
14498
14499 - "server_name" is the name of the last server to which the connection was
14500 sent, which might differ from the first one if there were connection errors
14501 and a redispatch occurred. Note that this server belongs to the backend
14502 which processed the request. If the connection was aborted before reaching
14503 a server, "<NOSRV>" is indicated instead of a server name.
14504
14505 - "Tw" is the total time in milliseconds spent waiting in the various queues.
14506 It can be "-1" if the connection was aborted before reaching the queue.
14507 See "Timers" below for more details.
14508
14509 - "Tc" is the total time in milliseconds spent waiting for the connection to
14510 establish to the final server, including retries. It can be "-1" if the
14511 connection was aborted before a connection could be established. See
14512 "Timers" below for more details.
14513
14514 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014515 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014516 "option logasap" was specified, then the time counting stops at the moment
14517 the log is emitted. In this case, a '+' sign is prepended before the value,
14518 indicating that the final one will be larger. See "Timers" below for more
14519 details.
14520
14521 - "bytes_read" is the total number of bytes transmitted from the server to
14522 the client when the log is emitted. If "option logasap" is specified, the
14523 this value will be prefixed with a '+' sign indicating that the final one
14524 may be larger. Please note that this value is a 64-bit counter, so log
14525 analysis tools must be able to handle it without overflowing.
14526
14527 - "termination_state" is the condition the session was in when the session
14528 ended. This indicates the session state, which side caused the end of
14529 session to happen, and for what reason (timeout, error, ...). The normal
14530 flags should be "--", indicating the session was closed by either end with
14531 no data remaining in buffers. See below "Session state at disconnection"
14532 for more details.
14533
14534 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014535 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014536 limits have been reached. For instance, if actconn is close to 512 when
14537 multiple connection errors occur, chances are high that the system limits
14538 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014539 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014540
14541 - "feconn" is the total number of concurrent connections on the frontend when
14542 the session was logged. It is useful to estimate the amount of resource
14543 required to sustain high loads, and to detect when the frontend's "maxconn"
14544 has been reached. Most often when this value increases by huge jumps, it is
14545 because there is congestion on the backend servers, but sometimes it can be
14546 caused by a denial of service attack.
14547
14548 - "beconn" is the total number of concurrent connections handled by the
14549 backend when the session was logged. It includes the total number of
14550 concurrent connections active on servers as well as the number of
14551 connections pending in queues. It is useful to estimate the amount of
14552 additional servers needed to support high loads for a given application.
14553 Most often when this value increases by huge jumps, it is because there is
14554 congestion on the backend servers, but sometimes it can be caused by a
14555 denial of service attack.
14556
14557 - "srv_conn" is the total number of concurrent connections still active on
14558 the server when the session was logged. It can never exceed the server's
14559 configured "maxconn" parameter. If this value is very often close or equal
14560 to the server's "maxconn", it means that traffic regulation is involved a
14561 lot, meaning that either the server's maxconn value is too low, or that
14562 there aren't enough servers to process the load with an optimal response
14563 time. When only one of the server's "srv_conn" is high, it usually means
14564 that this server has some trouble causing the connections to take longer to
14565 be processed than on other servers.
14566
14567 - "retries" is the number of connection retries experienced by this session
14568 when trying to connect to the server. It must normally be zero, unless a
14569 server is being stopped at the same moment the connection was attempted.
14570 Frequent retries generally indicate either a network problem between
14571 haproxy and the server, or a misconfigured system backlog on the server
14572 preventing new connections from being queued. This field may optionally be
14573 prefixed with a '+' sign, indicating that the session has experienced a
14574 redispatch after the maximal retry count has been reached on the initial
14575 server. In this case, the server name appearing in the log is the one the
14576 connection was redispatched to, and not the first one, though both may
14577 sometimes be the same in case of hashing for instance. So as a general rule
14578 of thumb, when a '+' is present in front of the retry count, this count
14579 should not be attributed to the logged server.
14580
14581 - "srv_queue" is the total number of requests which were processed before
14582 this one in the server queue. It is zero when the request has not gone
14583 through the server queue. It makes it possible to estimate the approximate
14584 server's response time by dividing the time spent in queue by the number of
14585 requests in the queue. It is worth noting that if a session experiences a
14586 redispatch and passes through two server queues, their positions will be
14587 cumulated. A request should not pass through both the server queue and the
14588 backend queue unless a redispatch occurs.
14589
14590 - "backend_queue" is the total number of requests which were processed before
14591 this one in the backend's global queue. It is zero when the request has not
14592 gone through the global queue. It makes it possible to estimate the average
14593 queue length, which easily translates into a number of missing servers when
14594 divided by a server's "maxconn" parameter. It is worth noting that if a
14595 session experiences a redispatch, it may pass twice in the backend's queue,
14596 and then both positions will be cumulated. A request should not pass
14597 through both the server queue and the backend queue unless a redispatch
14598 occurs.
14599
14600
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200146018.2.3. HTTP log format
14602----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014603
14604The HTTP format is the most complete and the best suited for HTTP proxies. It
14605is enabled by when "option httplog" is specified in the frontend. It provides
14606the same level of information as the TCP format with additional features which
14607are specific to the HTTP protocol. Just like the TCP format, the log is usually
14608emitted at the end of the session, unless "option logasap" is specified, which
14609generally only makes sense for download sites. A session which matches the
14610"monitor" rules will never logged. It is also possible not to log sessions for
14611which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020014612frontend. Successful connections will not be logged if "option dontlog-normal"
14613is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014614
14615Most fields are shared with the TCP log, some being different. A few fields may
14616slightly vary depending on some configuration options. Those ones are marked
14617with a star ('*') after the field name below.
14618
14619 Example :
14620 frontend http-in
14621 mode http
14622 option httplog
14623 log global
14624 default_backend bck
14625
14626 backend static
14627 server srv1 127.0.0.1:8000
14628
14629 >>> Feb 6 12:14:14 localhost \
14630 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
14631 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 +010014632 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014633
14634 Field Format Extract from the example above
14635 1 process_name '[' pid ']:' haproxy[14389]:
14636 2 client_ip ':' client_port 10.0.1.2:33317
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020014637 3 '[' request_date ']' [06/Feb/2009:12:14:14.655]
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014638 4 frontend_name http-in
14639 5 backend_name '/' server_name static/srv1
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020014640 6 TR '/' Tw '/' Tc '/' Tr '/' Ta* 10/0/30/69/109
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014641 7 status_code 200
14642 8 bytes_read* 2750
14643 9 captured_request_cookie -
14644 10 captured_response_cookie -
14645 11 termination_state ----
14646 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
14647 13 srv_queue '/' backend_queue 0/0
14648 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
14649 15 '{' captured_response_headers* '}' {}
14650 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010014651
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014652Detailed fields description :
14653 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014654 connection to haproxy. If the connection was accepted on a UNIX socket
14655 instead, the IP address would be replaced with the word "unix". Note that
14656 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010014657 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
14658 and the NetScaler Client IP insetion protocol is correctly used, then the
14659 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014660
14661 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014662 If the connection was accepted on a UNIX socket instead, the port would be
14663 replaced with the ID of the accepting socket, which is also reported in the
14664 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014665
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020014666 - "request_date" is the exact date when the first byte of the HTTP request
14667 was received by haproxy (log field %tr).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014668
14669 - "frontend_name" is the name of the frontend (or listener) which received
14670 and processed the connection.
14671
14672 - "backend_name" is the name of the backend (or listener) which was selected
14673 to manage the connection to the server. This will be the same as the
14674 frontend if no switching rule has been applied.
14675
14676 - "server_name" is the name of the last server to which the connection was
14677 sent, which might differ from the first one if there were connection errors
14678 and a redispatch occurred. Note that this server belongs to the backend
14679 which processed the request. If the request was aborted before reaching a
14680 server, "<NOSRV>" is indicated instead of a server name. If the request was
14681 intercepted by the stats subsystem, "<STATS>" is indicated instead.
14682
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020014683 - "TR" is the total time in milliseconds spent waiting for a full HTTP
14684 request from the client (not counting body) after the first byte was
14685 received. It can be "-1" if the connection was aborted before a complete
14686 request could be received or the a bad request was received. It should
14687 always be very small because a request generally fits in one single packet.
14688 Large times here generally indicate network issues between the client and
14689 haproxy or requests being typed by hand. See "Timers" below for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014690
14691 - "Tw" is the total time in milliseconds spent waiting in the various queues.
14692 It can be "-1" if the connection was aborted before reaching the queue.
14693 See "Timers" below for more details.
14694
14695 - "Tc" is the total time in milliseconds spent waiting for the connection to
14696 establish to the final server, including retries. It can be "-1" if the
14697 request was aborted before a connection could be established. See "Timers"
14698 below for more details.
14699
14700 - "Tr" is the total time in milliseconds spent waiting for the server to send
14701 a full HTTP response, not counting data. It can be "-1" if the request was
14702 aborted before a complete response could be received. It generally matches
14703 the server's processing time for the request, though it may be altered by
14704 the amount of data sent by the client to the server. Large times here on
14705 "GET" requests generally indicate an overloaded server. See "Timers" below
14706 for more details.
14707
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020014708 - "Ta" is the time the request remained active in haproxy, which is the total
14709 time in milliseconds elapsed between the first byte of the request was
14710 received and the last byte of response was sent. It covers all possible
14711 processing except the handshake (see Th) and idle time (see Ti). There is
14712 one exception, if "option logasap" was specified, then the time counting
14713 stops at the moment the log is emitted. In this case, a '+' sign is
14714 prepended before the value, indicating that the final one will be larger.
14715 See "Timers" below for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014716
14717 - "status_code" is the HTTP status code returned to the client. This status
14718 is generally set by the server, but it might also be set by haproxy when
14719 the server cannot be reached or when its response is blocked by haproxy.
14720
14721 - "bytes_read" is the total number of bytes transmitted to the client when
14722 the log is emitted. This does include HTTP headers. If "option logasap" is
14723 specified, the this value will be prefixed with a '+' sign indicating that
14724 the final one may be larger. Please note that this value is a 64-bit
14725 counter, so log analysis tools must be able to handle it without
14726 overflowing.
14727
14728 - "captured_request_cookie" is an optional "name=value" entry indicating that
14729 the client had this cookie in the request. The cookie name and its maximum
14730 length are defined by the "capture cookie" statement in the frontend
14731 configuration. The field is a single dash ('-') when the option is not
14732 set. Only one cookie may be captured, it is generally used to track session
14733 ID exchanges between a client and a server to detect session crossing
14734 between clients due to application bugs. For more details, please consult
14735 the section "Capturing HTTP headers and cookies" below.
14736
14737 - "captured_response_cookie" is an optional "name=value" entry indicating
14738 that the server has returned a cookie with its response. The cookie name
14739 and its maximum length are defined by the "capture cookie" statement in the
14740 frontend configuration. The field is a single dash ('-') when the option is
14741 not set. Only one cookie may be captured, it is generally used to track
14742 session ID exchanges between a client and a server to detect session
14743 crossing between clients due to application bugs. For more details, please
14744 consult the section "Capturing HTTP headers and cookies" below.
14745
14746 - "termination_state" is the condition the session was in when the session
14747 ended. This indicates the session state, which side caused the end of
14748 session to happen, for what reason (timeout, error, ...), just like in TCP
14749 logs, and information about persistence operations on cookies in the last
14750 two characters. The normal flags should begin with "--", indicating the
14751 session was closed by either end with no data remaining in buffers. See
14752 below "Session state at disconnection" for more details.
14753
14754 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014755 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014756 limits have been reached. For instance, if actconn is close to 512 or 1024
14757 when multiple connection errors occur, chances are high that the system
14758 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014759 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014760 system.
14761
14762 - "feconn" is the total number of concurrent connections on the frontend when
14763 the session was logged. It is useful to estimate the amount of resource
14764 required to sustain high loads, and to detect when the frontend's "maxconn"
14765 has been reached. Most often when this value increases by huge jumps, it is
14766 because there is congestion on the backend servers, but sometimes it can be
14767 caused by a denial of service attack.
14768
14769 - "beconn" is the total number of concurrent connections handled by the
14770 backend when the session was logged. It includes the total number of
14771 concurrent connections active on servers as well as the number of
14772 connections pending in queues. It is useful to estimate the amount of
14773 additional servers needed to support high loads for a given application.
14774 Most often when this value increases by huge jumps, it is because there is
14775 congestion on the backend servers, but sometimes it can be caused by a
14776 denial of service attack.
14777
14778 - "srv_conn" is the total number of concurrent connections still active on
14779 the server when the session was logged. It can never exceed the server's
14780 configured "maxconn" parameter. If this value is very often close or equal
14781 to the server's "maxconn", it means that traffic regulation is involved a
14782 lot, meaning that either the server's maxconn value is too low, or that
14783 there aren't enough servers to process the load with an optimal response
14784 time. When only one of the server's "srv_conn" is high, it usually means
14785 that this server has some trouble causing the requests to take longer to be
14786 processed than on other servers.
14787
14788 - "retries" is the number of connection retries experienced by this session
14789 when trying to connect to the server. It must normally be zero, unless a
14790 server is being stopped at the same moment the connection was attempted.
14791 Frequent retries generally indicate either a network problem between
14792 haproxy and the server, or a misconfigured system backlog on the server
14793 preventing new connections from being queued. This field may optionally be
14794 prefixed with a '+' sign, indicating that the session has experienced a
14795 redispatch after the maximal retry count has been reached on the initial
14796 server. In this case, the server name appearing in the log is the one the
14797 connection was redispatched to, and not the first one, though both may
14798 sometimes be the same in case of hashing for instance. So as a general rule
14799 of thumb, when a '+' is present in front of the retry count, this count
14800 should not be attributed to the logged server.
14801
14802 - "srv_queue" is the total number of requests which were processed before
14803 this one in the server queue. It is zero when the request has not gone
14804 through the server queue. It makes it possible to estimate the approximate
14805 server's response time by dividing the time spent in queue by the number of
14806 requests in the queue. It is worth noting that if a session experiences a
14807 redispatch and passes through two server queues, their positions will be
14808 cumulated. A request should not pass through both the server queue and the
14809 backend queue unless a redispatch occurs.
14810
14811 - "backend_queue" is the total number of requests which were processed before
14812 this one in the backend's global queue. It is zero when the request has not
14813 gone through the global queue. It makes it possible to estimate the average
14814 queue length, which easily translates into a number of missing servers when
14815 divided by a server's "maxconn" parameter. It is worth noting that if a
14816 session experiences a redispatch, it may pass twice in the backend's queue,
14817 and then both positions will be cumulated. A request should not pass
14818 through both the server queue and the backend queue unless a redispatch
14819 occurs.
14820
14821 - "captured_request_headers" is a list of headers captured in the request due
14822 to the presence of the "capture request header" statement in the frontend.
14823 Multiple headers can be captured, they will be delimited by a vertical bar
14824 ('|'). When no capture is enabled, the braces do not appear, causing a
14825 shift of remaining fields. It is important to note that this field may
14826 contain spaces, and that using it requires a smarter log parser than when
14827 it's not used. Please consult the section "Capturing HTTP headers and
14828 cookies" below for more details.
14829
14830 - "captured_response_headers" is a list of headers captured in the response
14831 due to the presence of the "capture response header" statement in the
14832 frontend. Multiple headers can be captured, they will be delimited by a
14833 vertical bar ('|'). When no capture is enabled, the braces do not appear,
14834 causing a shift of remaining fields. It is important to note that this
14835 field may contain spaces, and that using it requires a smarter log parser
14836 than when it's not used. Please consult the section "Capturing HTTP headers
14837 and cookies" below for more details.
14838
14839 - "http_request" is the complete HTTP request line, including the method,
14840 request and HTTP version string. Non-printable characters are encoded (see
14841 below the section "Non-printable characters"). This is always the last
14842 field, and it is always delimited by quotes and is the only one which can
14843 contain quotes. If new fields are added to the log format, they will be
14844 added before this field. This field might be truncated if the request is
14845 huge and does not fit in the standard syslog buffer (1024 characters). This
14846 is the reason why this field must always remain the last one.
14847
14848
Cyril Bontédc4d9032012-04-08 21:57:39 +0200148498.2.4. Custom log format
14850------------------------
William Lallemand48940402012-01-30 16:47:22 +010014851
Willy Tarreau2beef582012-12-20 17:22:52 +010014852The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010014853mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010014854
14855HAproxy understands some log format variables. % precedes log format variables.
14856Variables can take arguments using braces ('{}'), and multiple arguments are
14857separated by commas within the braces. Flags may be added or removed by
14858prefixing them with a '+' or '-' sign.
14859
14860Special variable "%o" may be used to propagate its flags to all other
14861variables on the same format string. This is particularly handy with quoted
Dragan Dosen835b9212016-02-12 13:23:03 +010014862("Q") and escaped ("E") string formats.
William Lallemand48940402012-01-30 16:47:22 +010014863
Willy Tarreauc8368452012-12-21 00:09:23 +010014864If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreaube722a22014-06-13 16:31:59 +020014865as a sample expression rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010014866less common information such as the client's SSL certificate's DN, or to log
14867the key that would be used to store an entry into a stick table.
14868
William Lallemand48940402012-01-30 16:47:22 +010014869Note: spaces must be escaped. A space character is considered as a separator.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014870In order to emit a verbatim '%', it must be preceded by another '%' resulting
Willy Tarreau06d97f92013-12-02 17:45:48 +010014871in '%%'. HAProxy will automatically merge consecutive separators.
William Lallemand48940402012-01-30 16:47:22 +010014872
Dragan Dosen835b9212016-02-12 13:23:03 +010014873Note: when using the RFC5424 syslog message format, the characters '"',
14874'\' and ']' inside PARAM-VALUE should be escaped with '\' as prefix (see
14875https://tools.ietf.org/html/rfc5424#section-6.3.3 for more details). In
14876such cases, the use of the flag "E" should be considered.
14877
William Lallemand48940402012-01-30 16:47:22 +010014878Flags are :
14879 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014880 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
Dragan Dosen835b9212016-02-12 13:23:03 +010014881 * E: escape characters '"', '\' and ']' in a string with '\' as prefix
14882 (intended purpose is for the RFC5424 structured-data log formats)
William Lallemand48940402012-01-30 16:47:22 +010014883
14884 Example:
14885
14886 log-format %T\ %t\ Some\ Text
14887 log-format %{+Q}o\ %t\ %s\ %{-Q}r
14888
Dragan Dosen835b9212016-02-12 13:23:03 +010014889 log-format-sd %{+Q,+E}o\ [exampleSDID@1234\ header=%[capture.req.hdr(0)]]
14890
William Lallemand48940402012-01-30 16:47:22 +010014891At the moment, the default HTTP format is defined this way :
14892
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020014893 log-format "%ci:%cp [%tr] %ft %b/%s %TR/%Tw/%Tc/%Tr/%Ta %ST %B %CC \
14894 %CS %tsc %ac/%fc/%bc/%sc/%rc %sq/%bq %hr %hs %{+Q}r"
William Lallemand48940402012-01-30 16:47:22 +010014895
William Lallemandbddd4fd2012-02-27 11:23:10 +010014896the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010014897
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020014898 log-format "%{+Q}o %{-Q}ci - - [%trg] %r %ST %B \"\" \"\" %cp \
14899 %ms %ft %b %s %TR %Tw %Tc %Tr %Ta %tsc %ac %fc \
14900 %bc %sc %rc %sq %bq %CC %CS %hrl %hsl"
William Lallemand48940402012-01-30 16:47:22 +010014901
William Lallemandbddd4fd2012-02-27 11:23:10 +010014902and the default TCP format is defined this way :
14903
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020014904 log-format "%ci:%cp [%t] %ft %b/%s %Tw/%Tc/%Tt %B %ts \
14905 %ac/%fc/%bc/%sc/%rc %sq/%bq"
William Lallemandbddd4fd2012-02-27 11:23:10 +010014906
William Lallemand48940402012-01-30 16:47:22 +010014907Please refer to the table below for currently defined variables :
14908
William Lallemandbddd4fd2012-02-27 11:23:10 +010014909 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014910 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014911 +---+------+-----------------------------------------------+-------------+
14912 | | %o | special variable, apply flags on all next var | |
14913 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010014914 | | %B | bytes_read (from server to client) | numeric |
14915 | H | %CC | captured_request_cookie | string |
14916 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020014917 | | %H | hostname | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000014918 | H | %HM | HTTP method (ex: POST) | string |
14919 | H | %HP | HTTP request URI without query string (path) | string |
Andrew Hayworthe63ac872015-07-31 16:14:16 +000014920 | H | %HQ | HTTP request URI query string (ex: ?bar=baz) | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000014921 | H | %HU | HTTP request URI (ex: /foo?bar=baz) | string |
14922 | H | %HV | HTTP version (ex: HTTP/1.0) | string |
William Lallemanda73203e2012-03-12 12:48:57 +010014923 | | %ID | unique-id | string |
Willy Tarreau4bf99632014-06-13 12:21:40 +020014924 | | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020014925 | | %T | gmt_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020014926 | | %Ta | Active time of the request (from TR to end) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014927 | | %Tc | Tc | numeric |
Willy Tarreau27b639d2016-05-17 17:55:27 +020014928 | | %Td | Td = Tt - (Tq + Tw + Tc + Tr) | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080014929 | | %Tl | local_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020014930 | | %Th | connection handshake time (SSL, PROXY proto) | numeric |
14931 | H | %Ti | idle time before the HTTP request | numeric |
14932 | H | %Tq | Th + Ti + TR | numeric |
14933 | H | %TR | time to receive the full request from 1st byte| numeric |
14934 | H | %Tr | Tr (response time) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020014935 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014936 | | %Tt | Tt | numeric |
14937 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010014938 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014939 | | %ac | actconn | numeric |
14940 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010014941 | | %bc | beconn (backend concurrent connections) | numeric |
14942 | | %bi | backend_source_ip (connecting address) | IP |
14943 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014944 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010014945 | | %ci | client_ip (accepted address) | IP |
14946 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014947 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010014948 | | %fc | feconn (frontend concurrent connections) | numeric |
14949 | | %fi | frontend_ip (accepting address) | IP |
14950 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020014951 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreau7346acb2014-08-28 15:03:15 +020014952 | | %lc | frontend_log_counter | numeric |
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020014953 | | %hr | captured_request_headers default style | string |
14954 | | %hrl | captured_request_headers CLF style | string list |
14955 | | %hs | captured_response_headers default style | string |
14956 | | %hsl | captured_response_headers CLF style | string list |
Willy Tarreau812c88e2015-08-09 10:56:35 +020014957 | | %ms | accept date milliseconds (left-padded with 0) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020014958 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014959 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014960 | | %rc | retries | numeric |
Willy Tarreau1f0da242014-01-25 11:01:50 +010014961 | | %rt | request_counter (HTTP req or TCP session) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014962 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010014963 | | %sc | srv_conn (server concurrent connections) | numeric |
14964 | | %si | server_IP (target address) | IP |
14965 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014966 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014967 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
14968 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010014969 | | %t | date_time (with millisecond resolution) | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020014970 | H | %tr | date_time of HTTP request | date |
14971 | H | %trg | gmt_date_time of start of HTTP request | date |
14972 | H | %trl | locla_date_time of start of HTTP request | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014973 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014974 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014975 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010014976
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014977 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010014978
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010014979
149808.2.5. Error log format
14981-----------------------
14982
14983When an incoming connection fails due to an SSL handshake or an invalid PROXY
14984protocol header, haproxy will log the event using a shorter, fixed line format.
14985By default, logs are emitted at the LOG_INFO level, unless the option
14986"log-separate-errors" is set in the backend, in which case the LOG_ERR level
14987will be used. Connections on which no data are exchanged (eg: probes) are not
14988logged if the "dontlognull" option is set.
14989
14990The format looks like this :
14991
14992 >>> Dec 3 18:27:14 localhost \
14993 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
14994 Connection error during SSL handshake
14995
14996 Field Format Extract from the example above
14997 1 process_name '[' pid ']:' haproxy[6103]:
14998 2 client_ip ':' client_port 127.0.0.1:56059
14999 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
15000 4 frontend_name "/" bind_name ":" frt/f1:
15001 5 message Connection error during SSL handshake
15002
15003These fields just provide minimal information to help debugging connection
15004failures.
15005
15006
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200150078.3. Advanced logging options
15008-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015009
15010Some advanced logging options are often looked for but are not easy to find out
15011just by looking at the various options. Here is an entry point for the few
15012options which can enable better logging. Please refer to the keywords reference
15013for more information about their usage.
15014
15015
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200150168.3.1. Disabling logging of external tests
15017------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015018
15019It is quite common to have some monitoring tools perform health checks on
15020haproxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
15021commercial load-balancer, and sometimes it will simply be a more complete
15022monitoring system such as Nagios. When the tests are very frequent, users often
15023ask how to disable logging for those checks. There are three possibilities :
15024
15025 - if connections come from everywhere and are just TCP probes, it is often
15026 desired to simply disable logging of connections without data exchange, by
15027 setting "option dontlognull" in the frontend. It also disables logging of
15028 port scans, which may or may not be desired.
15029
15030 - if the connection come from a known source network, use "monitor-net" to
15031 declare this network as monitoring only. Any host in this network will then
15032 only be able to perform health checks, and their requests will not be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015033 logged. This is generally appropriate to designate a list of equipment
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015034 such as other load-balancers.
15035
15036 - if the tests are performed on a known URI, use "monitor-uri" to declare
15037 this URI as dedicated to monitoring. Any host sending this request will
15038 only get the result of a health-check, and the request will not be logged.
15039
15040
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200150418.3.2. Logging before waiting for the session to terminate
15042----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015043
15044The problem with logging at end of connection is that you have no clue about
15045what is happening during very long sessions, such as remote terminal sessions
15046or large file downloads. This problem can be worked around by specifying
15047"option logasap" in the frontend. Haproxy will then log as soon as possible,
15048just before data transfer begins. This means that in case of TCP, it will still
15049log the connection status to the server, and in case of HTTP, it will log just
15050after processing the server headers. In this case, the number of bytes reported
15051is the number of header bytes sent to the client. In order to avoid confusion
15052with normal logs, the total time field and the number of bytes are prefixed
15053with a '+' sign which means that real numbers are certainly larger.
15054
15055
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200150568.3.3. Raising log level upon errors
15057------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020015058
15059Sometimes it is more convenient to separate normal traffic from errors logs,
15060for instance in order to ease error monitoring from log files. When the option
15061"log-separate-errors" is used, connections which experience errors, timeouts,
15062retries, redispatches or HTTP status codes 5xx will see their syslog level
15063raised from "info" to "err". This will help a syslog daemon store the log in
15064a separate file. It is very important to keep the errors in the normal traffic
15065file too, so that log ordering is not altered. You should also be careful if
15066you already have configured your syslog daemon to store all logs higher than
15067"notice" in an "admin" file, because the "err" level is higher than "notice".
15068
15069
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200150708.3.4. Disabling logging of successful connections
15071--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020015072
15073Although this may sound strange at first, some large sites have to deal with
15074multiple thousands of logs per second and are experiencing difficulties keeping
15075them intact for a long time or detecting errors within them. If the option
15076"dontlog-normal" is set on the frontend, all normal connections will not be
15077logged. In this regard, a normal connection is defined as one without any
15078error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
15079and a response with a status 5xx is not considered normal and will be logged
15080too. Of course, doing is is really discouraged as it will remove most of the
15081useful information from the logs. Do this only if you have no other
15082alternative.
15083
15084
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200150858.4. Timing events
15086------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015087
15088Timers provide a great help in troubleshooting network problems. All values are
15089reported in milliseconds (ms). These timers should be used in conjunction with
15090the session termination flags. In TCP mode with "option tcplog" set on the
15091frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015092mode, 5 control points are reported under the form "TR/Tw/Tc/Tr/Ta". In
15093addition, three other measures are provided, "Th", "Ti", and "Tq".
15094
15095 - Th: total time to accept tcp connection and execute handshakes for low level
15096 protocols. Currently, these protocoles are proxy-protocol and SSL. This may
15097 only happen once during the whole connection's lifetime. A large time here
15098 may indicate that the client only pre-established the connection without
15099 speaking, that it is experiencing network issues preventing it from
15100 completing a handshake in a reasonable time (eg: MTU issues), or that an
15101 SSL handshake was very expensive to compute.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015102
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015103 - Ti: is the idle time before the HTTP request (HTTP mode only). This timer
15104 counts between the end of the handshakes and the first byte of the HTTP
15105 request. When dealing with a second request in keep-alive mode, it starts
15106 to count after the end of the transmission the previous response. Some
15107 browsers pre-establish connections to a server in order to reduce the
15108 latency of a future request, and keep them pending until they need it. This
15109 delay will be reported as the idle time. A value of -1 indicates that
15110 nothing was received on the connection.
15111
15112 - TR: total time to get the client request (HTTP mode only). It's the time
15113 elapsed between the first bytes received and the moment the proxy received
15114 the empty line marking the end of the HTTP headers. The value "-1"
15115 indicates that the end of headers has never been seen. This happens when
15116 the client closes prematurely or times out. This time is usually very short
15117 since most requests fit in a single packet. A large time may indicate a
15118 request typed by hand during a test.
15119
15120 - Tq: total time to get the client request from the accept date or since the
15121 emission of the last byte of the previous response (HTTP mode only). It's
15122 exactly equalt to Th + Ti + TR unless any of them is -1, in which case it
15123 returns -1 as well. This timer used to be very useful before the arrival of
15124 HTTP keep-alive and browsers' pre-connect feature. It's recommended to drop
15125 it in favor of TR nowadays, as the idle time adds a lot of noise to the
15126 reports.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015127
15128 - Tw: total time spent in the queues waiting for a connection slot. It
15129 accounts for backend queue as well as the server queues, and depends on the
15130 queue size, and the time needed for the server to complete previous
15131 requests. The value "-1" means that the request was killed before reaching
15132 the queue, which is generally what happens with invalid or denied requests.
15133
15134 - Tc: total time to establish the TCP connection to the server. It's the time
15135 elapsed between the moment the proxy sent the connection request, and the
15136 moment it was acknowledged by the server, or between the TCP SYN packet and
15137 the matching SYN/ACK packet in return. The value "-1" means that the
15138 connection never established.
15139
15140 - Tr: server response time (HTTP mode only). It's the time elapsed between
15141 the moment the TCP connection was established to the server and the moment
15142 the server sent its complete response headers. It purely shows its request
15143 processing time, without the network overhead due to the data transmission.
15144 It is worth noting that when the client has data to send to the server, for
15145 instance during a POST request, the time already runs, and this can distort
15146 apparent response time. For this reason, it's generally wise not to trust
15147 too much this field for POST requests initiated from clients behind an
15148 untrusted network. A value of "-1" here means that the last the response
15149 header (empty line) was never seen, most likely because the server timeout
15150 stroke before the server managed to process the request.
15151
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015152 - Ta: total active time for the HTTP request, between the moment the proxy
15153 received the first byte of the request header and the emission of the last
15154 byte of the response body. The exception is when the "logasap" option is
15155 specified. In this case, it only equals (TR+Tw+Tc+Tr), and is prefixed with
15156 a '+' sign. From this field, we can deduce "Td", the data transmission time,
15157 by subtracting other timers when valid :
15158
15159 Td = Ta - (TR + Tw + Tc + Tr)
15160
15161 Timers with "-1" values have to be excluded from this equation. Note that
15162 "Ta" can never be negative.
15163
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015164 - Tt: total session duration time, between the moment the proxy accepted it
15165 and the moment both ends were closed. The exception is when the "logasap"
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015166 option is specified. In this case, it only equals (Th+Ti+TR+Tw+Tc+Tr), and
15167 is prefixed with a '+' sign. From this field, we can deduce "Td", the data
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015168 transmission time, by subtracting other timers when valid :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015169
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015170 Td = Tt - (Th + Ti + TR + Tw + Tc + Tr)
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015171
15172 Timers with "-1" values have to be excluded from this equation. In TCP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015173 mode, "Ti", "Tq" and "Tr" have to be excluded too. Note that "Tt" can never
15174 be negative and that for HTTP, Tt is simply equal to (Th+Ti+Ta).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015175
15176These timers provide precious indications on trouble causes. Since the TCP
15177protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
15178that timers close to multiples of 3s are nearly always related to lost packets
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015179due to network problems (wires, negotiation, congestion). Moreover, if "Ta" or
15180"Tt" is close to a timeout value specified in the configuration, it often means
15181that a session has been aborted on timeout.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015182
15183Most common cases :
15184
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015185 - If "Th" or "Ti" are close to 3000, a packet has probably been lost between
15186 the client and the proxy. This is very rare on local networks but might
15187 happen when clients are on far remote networks and send large requests. It
15188 may happen that values larger than usual appear here without any network
15189 cause. Sometimes, during an attack or just after a resource starvation has
15190 ended, haproxy may accept thousands of connections in a few milliseconds.
15191 The time spent accepting these connections will inevitably slightly delay
15192 processing of other connections, and it can happen that request times in the
15193 order of a few tens of milliseconds are measured after a few thousands of
15194 new connections have been accepted at once. Using one of the keep-alive
15195 modes may display larger idle times since "Ti" measures the time spent
Patrick Mezard105faca2010-06-12 17:02:46 +020015196 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015197
15198 - If "Tc" is close to 3000, a packet has probably been lost between the
15199 server and the proxy during the server connection phase. This value should
15200 always be very low, such as 1 ms on local networks and less than a few tens
15201 of ms on remote networks.
15202
Willy Tarreau55165fe2009-05-10 12:02:55 +020015203 - If "Tr" is nearly always lower than 3000 except some rare values which seem
15204 to be the average majored by 3000, there are probably some packets lost
15205 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015206
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015207 - If "Ta" is large even for small byte counts, it generally is because
15208 neither the client nor the server decides to close the connection while
15209 haproxy is running in tunnel mode and both have agreed on a keep-alive
15210 connection mode. In order to solve this issue, it will be needed to specify
15211 one of the HTTP options to manipulate keep-alive or close options on either
15212 the frontend or the backend. Having the smallest possible 'Ta' or 'Tt' is
15213 important when connection regulation is used with the "maxconn" option on
15214 the servers, since no new connection will be sent to the server until
15215 another one is released.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015216
15217Other noticeable HTTP log cases ('xx' means any value to be ignored) :
15218
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015219 TR/Tw/Tc/Tr/+Ta The "option logasap" is present on the frontend and the log
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015220 was emitted before the data phase. All the timers are valid
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015221 except "Ta" which is shorter than reality.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015222
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015223 -1/xx/xx/xx/Ta The client was not able to send a complete request in time
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015224 or it aborted too early. Check the session termination flags
15225 then "timeout http-request" and "timeout client" settings.
15226
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015227 TR/-1/xx/xx/Ta It was not possible to process the request, maybe because
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015228 servers were out of order, because the request was invalid
15229 or forbidden by ACL rules. Check the session termination
15230 flags.
15231
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015232 TR/Tw/-1/xx/Ta The connection could not establish on the server. Either it
15233 actively refused it or it timed out after Ta-(TR+Tw) ms.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015234 Check the session termination flags, then check the
15235 "timeout connect" setting. Note that the tarpit action might
15236 return similar-looking patterns, with "Tw" equal to the time
15237 the client connection was maintained open.
15238
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015239 TR/Tw/Tc/-1/Ta The server has accepted the connection but did not return
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015240 a complete response in time, or it closed its connection
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015241 unexpectedly after Ta-(TR+Tw+Tc) ms. Check the session
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015242 termination flags, then check the "timeout server" setting.
15243
15244
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200152458.5. Session state at disconnection
15246-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015247
15248TCP and HTTP logs provide a session termination indicator in the
15249"termination_state" field, just before the number of active connections. It is
152502-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
15251each of which has a special meaning :
15252
15253 - On the first character, a code reporting the first event which caused the
15254 session to terminate :
15255
15256 C : the TCP session was unexpectedly aborted by the client.
15257
15258 S : the TCP session was unexpectedly aborted by the server, or the
15259 server explicitly refused it.
15260
15261 P : the session was prematurely aborted by the proxy, because of a
15262 connection limit enforcement, because a DENY filter was matched,
15263 because of a security check which detected and blocked a dangerous
15264 error in server response which might have caused information leak
Willy Tarreau570f2212013-06-10 16:42:09 +020015265 (eg: cacheable cookie).
15266
15267 L : the session was locally processed by haproxy and was not passed to
15268 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015269
15270 R : a resource on the proxy has been exhausted (memory, sockets, source
15271 ports, ...). Usually, this appears during the connection phase, and
15272 system logs should contain a copy of the precise error. If this
15273 happens, it must be considered as a very serious anomaly which
15274 should be fixed as soon as possible by any means.
15275
15276 I : an internal error was identified by the proxy during a self-check.
15277 This should NEVER happen, and you are encouraged to report any log
15278 containing this, because this would almost certainly be a bug. It
15279 would be wise to preventively restart the process after such an
15280 event too, in case it would be caused by memory corruption.
15281
Simon Horman752dc4a2011-06-21 14:34:59 +090015282 D : the session was killed by haproxy because the server was detected
15283 as down and was configured to kill all connections when going down.
15284
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070015285 U : the session was killed by haproxy on this backup server because an
15286 active server was detected as up and was configured to kill all
15287 backup connections when going up.
15288
Willy Tarreaua2a64e92011-09-07 23:01:56 +020015289 K : the session was actively killed by an admin operating on haproxy.
15290
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015291 c : the client-side timeout expired while waiting for the client to
15292 send or receive data.
15293
15294 s : the server-side timeout expired while waiting for the server to
15295 send or receive data.
15296
15297 - : normal session completion, both the client and the server closed
15298 with nothing left in the buffers.
15299
15300 - on the second character, the TCP or HTTP session state when it was closed :
15301
Willy Tarreauf7b30a92010-12-06 22:59:17 +010015302 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015303 (HTTP mode only). Nothing was sent to any server.
15304
15305 Q : the proxy was waiting in the QUEUE for a connection slot. This can
15306 only happen when servers have a 'maxconn' parameter set. It can
15307 also happen in the global queue after a redispatch consecutive to
15308 a failed attempt to connect to a dying server. If no redispatch is
15309 reported, then no connection attempt was made to any server.
15310
15311 C : the proxy was waiting for the CONNECTION to establish on the
15312 server. The server might at most have noticed a connection attempt.
15313
15314 H : the proxy was waiting for complete, valid response HEADERS from the
15315 server (HTTP only).
15316
15317 D : the session was in the DATA phase.
15318
15319 L : the proxy was still transmitting LAST data to the client while the
15320 server had already finished. This one is very rare as it can only
15321 happen when the client dies while receiving the last packets.
15322
15323 T : the request was tarpitted. It has been held open with the client
15324 during the whole "timeout tarpit" duration or until the client
15325 closed, both of which will be reported in the "Tw" timer.
15326
15327 - : normal session completion after end of data transfer.
15328
15329 - the third character tells whether the persistence cookie was provided by
15330 the client (only in HTTP mode) :
15331
15332 N : the client provided NO cookie. This is usually the case for new
15333 visitors, so counting the number of occurrences of this flag in the
15334 logs generally indicate a valid trend for the site frequentation.
15335
15336 I : the client provided an INVALID cookie matching no known server.
15337 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020015338 cookies between HTTP/HTTPS sites, persistence conditionally
15339 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015340
15341 D : the client provided a cookie designating a server which was DOWN,
15342 so either "option persist" was used and the client was sent to
15343 this server, or it was not set and the client was redispatched to
15344 another server.
15345
Willy Tarreau996a92c2010-10-13 19:30:47 +020015346 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015347 server.
15348
Willy Tarreau996a92c2010-10-13 19:30:47 +020015349 E : the client provided a valid cookie, but with a last date which was
15350 older than what is allowed by the "maxidle" cookie parameter, so
15351 the cookie is consider EXPIRED and is ignored. The request will be
15352 redispatched just as if there was no cookie.
15353
15354 O : the client provided a valid cookie, but with a first date which was
15355 older than what is allowed by the "maxlife" cookie parameter, so
15356 the cookie is consider too OLD and is ignored. The request will be
15357 redispatched just as if there was no cookie.
15358
Willy Tarreauc89ccb62012-04-05 21:18:22 +020015359 U : a cookie was present but was not used to select the server because
15360 some other server selection mechanism was used instead (typically a
15361 "use-server" rule).
15362
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015363 - : does not apply (no cookie set in configuration).
15364
15365 - the last character reports what operations were performed on the persistence
15366 cookie returned by the server (only in HTTP mode) :
15367
15368 N : NO cookie was provided by the server, and none was inserted either.
15369
15370 I : no cookie was provided by the server, and the proxy INSERTED one.
15371 Note that in "cookie insert" mode, if the server provides a cookie,
15372 it will still be overwritten and reported as "I" here.
15373
Willy Tarreau996a92c2010-10-13 19:30:47 +020015374 U : the proxy UPDATED the last date in the cookie that was presented by
15375 the client. This can only happen in insert mode with "maxidle". It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015376 happens every time there is activity at a different date than the
Willy Tarreau996a92c2010-10-13 19:30:47 +020015377 date indicated in the cookie. If any other change happens, such as
15378 a redispatch, then the cookie will be marked as inserted instead.
15379
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015380 P : a cookie was PROVIDED by the server and transmitted as-is.
15381
15382 R : the cookie provided by the server was REWRITTEN by the proxy, which
15383 happens in "cookie rewrite" or "cookie prefix" modes.
15384
15385 D : the cookie provided by the server was DELETED by the proxy.
15386
15387 - : does not apply (no cookie set in configuration).
15388
Willy Tarreau996a92c2010-10-13 19:30:47 +020015389The combination of the two first flags gives a lot of information about what
15390was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015391helpful to detect server saturation, network troubles, local system resource
15392starvation, attacks, etc...
15393
15394The most common termination flags combinations are indicated below. They are
15395alphabetically sorted, with the lowercase set just after the upper case for
15396easier finding and understanding.
15397
15398 Flags Reason
15399
15400 -- Normal termination.
15401
15402 CC The client aborted before the connection could be established to the
15403 server. This can happen when haproxy tries to connect to a recently
15404 dead (or unchecked) server, and the client aborts while haproxy is
15405 waiting for the server to respond or for "timeout connect" to expire.
15406
15407 CD The client unexpectedly aborted during data transfer. This can be
15408 caused by a browser crash, by an intermediate equipment between the
15409 client and haproxy which decided to actively break the connection,
15410 by network routing issues between the client and haproxy, or by a
15411 keep-alive session between the server and the client terminated first
15412 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010015413
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015414 cD The client did not send nor acknowledge any data for as long as the
15415 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020015416 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015417
15418 CH The client aborted while waiting for the server to start responding.
15419 It might be the server taking too long to respond or the client
15420 clicking the 'Stop' button too fast.
15421
15422 cH The "timeout client" stroke while waiting for client data during a
15423 POST request. This is sometimes caused by too large TCP MSS values
15424 for PPPoE networks which cannot transport full-sized packets. It can
15425 also happen when client timeout is smaller than server timeout and
15426 the server takes too long to respond.
15427
15428 CQ The client aborted while its session was queued, waiting for a server
15429 with enough empty slots to accept it. It might be that either all the
15430 servers were saturated or that the assigned server was taking too
15431 long a time to respond.
15432
15433 CR The client aborted before sending a full HTTP request. Most likely
15434 the request was typed by hand using a telnet client, and aborted
15435 too early. The HTTP status code is likely a 400 here. Sometimes this
15436 might also be caused by an IDS killing the connection between haproxy
Willy Tarreau0f228a02015-05-01 15:37:53 +020015437 and the client. "option http-ignore-probes" can be used to ignore
15438 connections without any data transfer.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015439
15440 cR The "timeout http-request" stroke before the client sent a full HTTP
15441 request. This is sometimes caused by too large TCP MSS values on the
15442 client side for PPPoE networks which cannot transport full-sized
15443 packets, or by clients sending requests by hand and not typing fast
15444 enough, or forgetting to enter the empty line at the end of the
Willy Tarreau2705a612014-05-23 17:38:34 +020015445 request. The HTTP status code is likely a 408 here. Note: recently,
Willy Tarreau0f228a02015-05-01 15:37:53 +020015446 some browsers started to implement a "pre-connect" feature consisting
15447 in speculatively connecting to some recently visited web sites just
15448 in case the user would like to visit them. This results in many
15449 connections being established to web sites, which end up in 408
15450 Request Timeout if the timeout strikes first, or 400 Bad Request when
15451 the browser decides to close them first. These ones pollute the log
15452 and feed the error counters. Some versions of some browsers have even
15453 been reported to display the error code. It is possible to work
15454 around the undesirable effects of this behaviour by adding "option
15455 http-ignore-probes" in the frontend, resulting in connections with
15456 zero data transfer to be totally ignored. This will definitely hide
15457 the errors of people experiencing connectivity issues though.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015458
15459 CT The client aborted while its session was tarpitted. It is important to
15460 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020015461 wrong tarpit rules have been written. If a lot of them happen, it
15462 might make sense to lower the "timeout tarpit" value to something
15463 closer to the average reported "Tw" timer, in order not to consume
15464 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015465
Willy Tarreau570f2212013-06-10 16:42:09 +020015466 LR The request was intercepted and locally handled by haproxy. Generally
15467 it means that this was a redirect or a stats request.
15468
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010015469 SC The server or an equipment between it and haproxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015470 the TCP connection (the proxy received a TCP RST or an ICMP message
15471 in return). Under some circumstances, it can also be the network
15472 stack telling the proxy that the server is unreachable (eg: no route,
15473 or no ARP response on local network). When this happens in HTTP mode,
15474 the status code is likely a 502 or 503 here.
15475
15476 sC The "timeout connect" stroke before a connection to the server could
15477 complete. When this happens in HTTP mode, the status code is likely a
15478 503 or 504 here.
15479
15480 SD The connection to the server died with an error during the data
15481 transfer. This usually means that haproxy has received an RST from
15482 the server or an ICMP message from an intermediate equipment while
15483 exchanging data with the server. This can be caused by a server crash
15484 or by a network issue on an intermediate equipment.
15485
15486 sD The server did not send nor acknowledge any data for as long as the
15487 "timeout server" setting during the data phase. This is often caused
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010015488 by too short timeouts on L4 equipments before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015489 load-balancers, ...), as well as keep-alive sessions maintained
15490 between the client and the server expiring first on haproxy.
15491
15492 SH The server aborted before sending its full HTTP response headers, or
15493 it crashed while processing the request. Since a server aborting at
15494 this moment is very rare, it would be wise to inspect its logs to
15495 control whether it crashed and why. The logged request may indicate a
15496 small set of faulty requests, demonstrating bugs in the application.
15497 Sometimes this might also be caused by an IDS killing the connection
15498 between haproxy and the server.
15499
15500 sH The "timeout server" stroke before the server could return its
15501 response headers. This is the most common anomaly, indicating too
15502 long transactions, probably caused by server or database saturation.
15503 The immediate workaround consists in increasing the "timeout server"
15504 setting, but it is important to keep in mind that the user experience
15505 will suffer from these long response times. The only long term
15506 solution is to fix the application.
15507
15508 sQ The session spent too much time in queue and has been expired. See
15509 the "timeout queue" and "timeout connect" settings to find out how to
15510 fix this if it happens too often. If it often happens massively in
15511 short periods, it may indicate general problems on the affected
15512 servers due to I/O or database congestion, or saturation caused by
15513 external attacks.
15514
15515 PC The proxy refused to establish a connection to the server because the
15516 process' socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020015517 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015518 so that it does not happen anymore. This status is very rare and
15519 might happen when the global "ulimit-n" parameter is forced by hand.
15520
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010015521 PD The proxy blocked an incorrectly formatted chunked encoded message in
15522 a request or a response, after the server has emitted its headers. In
15523 most cases, this will indicate an invalid message from the server to
Willy Tarreauf3a3e132013-08-31 08:16:26 +020015524 the client. Haproxy supports chunk sizes of up to 2GB - 1 (2147483647
15525 bytes). Any larger size will be considered as an error.
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010015526
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015527 PH The proxy blocked the server's response, because it was invalid,
15528 incomplete, dangerous (cache control), or matched a security filter.
15529 In any case, an HTTP 502 error is sent to the client. One possible
15530 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010015531 containing unauthorized characters. It is also possible but quite
15532 rare, that the proxy blocked a chunked-encoding request from the
15533 client due to an invalid syntax, before the server responded. In this
15534 case, an HTTP 400 error is sent to the client and reported in the
15535 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015536
15537 PR The proxy blocked the client's HTTP request, either because of an
15538 invalid HTTP syntax, in which case it returned an HTTP 400 error to
15539 the client, or because a deny filter matched, in which case it
15540 returned an HTTP 403 error.
15541
15542 PT The proxy blocked the client's request and has tarpitted its
15543 connection before returning it a 500 server error. Nothing was sent
15544 to the server. The connection was maintained open for as long as
15545 reported by the "Tw" timer field.
15546
15547 RC A local resource has been exhausted (memory, sockets, source ports)
15548 preventing the connection to the server from establishing. The error
15549 logs will tell precisely what was missing. This is very rare and can
15550 only be solved by proper system tuning.
15551
Willy Tarreau996a92c2010-10-13 19:30:47 +020015552The combination of the two last flags gives a lot of information about how
15553persistence was handled by the client, the server and by haproxy. This is very
15554important to troubleshoot disconnections, when users complain they have to
15555re-authenticate. The commonly encountered flags are :
15556
15557 -- Persistence cookie is not enabled.
15558
15559 NN No cookie was provided by the client, none was inserted in the
15560 response. For instance, this can be in insert mode with "postonly"
15561 set on a GET request.
15562
15563 II A cookie designating an invalid server was provided by the client,
15564 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015565 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020015566 value can be presented by a client when no other server knows it.
15567
15568 NI No cookie was provided by the client, one was inserted in the
15569 response. This typically happens for first requests from every user
15570 in "insert" mode, which makes it an easy way to count real users.
15571
15572 VN A cookie was provided by the client, none was inserted in the
15573 response. This happens for most responses for which the client has
15574 already got a cookie.
15575
15576 VU A cookie was provided by the client, with a last visit date which is
15577 not completely up-to-date, so an updated cookie was provided in
15578 response. This can also happen if there was no date at all, or if
15579 there was a date but the "maxidle" parameter was not set, so that the
15580 cookie can be switched to unlimited time.
15581
15582 EI A cookie was provided by the client, with a last visit date which is
15583 too old for the "maxidle" parameter, so the cookie was ignored and a
15584 new cookie was inserted in the response.
15585
15586 OI A cookie was provided by the client, with a first visit date which is
15587 too old for the "maxlife" parameter, so the cookie was ignored and a
15588 new cookie was inserted in the response.
15589
15590 DI The server designated by the cookie was down, a new server was
15591 selected and a new cookie was emitted in the response.
15592
15593 VI The server designated by the cookie was not marked dead but could not
15594 be reached. A redispatch happened and selected another one, which was
15595 then advertised in the response.
15596
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015597
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200155988.6. Non-printable characters
15599-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015600
15601In order not to cause trouble to log analysis tools or terminals during log
15602consulting, non-printable characters are not sent as-is into log files, but are
15603converted to the two-digits hexadecimal representation of their ASCII code,
15604prefixed by the character '#'. The only characters that can be logged without
15605being escaped are comprised between 32 and 126 (inclusive). Obviously, the
15606escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
15607is the same for the character '"' which becomes "#22", as well as '{', '|' and
15608'}' when logging headers.
15609
15610Note that the space character (' ') is not encoded in headers, which can cause
15611issues for tools relying on space count to locate fields. A typical header
15612containing spaces is "User-Agent".
15613
15614Last, it has been observed that some syslog daemons such as syslog-ng escape
15615the quote ('"') with a backslash ('\'). The reverse operation can safely be
15616performed since no quote may appear anywhere else in the logs.
15617
15618
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200156198.7. Capturing HTTP cookies
15620---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015621
15622Cookie capture simplifies the tracking a complete user session. This can be
15623achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015624section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015625cookie will simultaneously be checked in the request ("Cookie:" header) and in
15626the response ("Set-Cookie:" header). The respective values will be reported in
15627the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015628locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015629not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
15630user switches to a new session for example, because the server will reassign it
15631a new cookie. It is also possible to detect if a server unexpectedly sets a
15632wrong cookie to a client, leading to session crossing.
15633
15634 Examples :
15635 # capture the first cookie whose name starts with "ASPSESSION"
15636 capture cookie ASPSESSION len 32
15637
15638 # capture the first cookie whose name is exactly "vgnvisitor"
15639 capture cookie vgnvisitor= len 32
15640
15641
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200156428.8. Capturing HTTP headers
15643---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015644
15645Header captures are useful to track unique request identifiers set by an upper
15646proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
15647the response, one can search for information about the response length, how the
15648server asked the cache to behave, or an object location during a redirection.
15649
15650Header captures are performed using the "capture request header" and "capture
15651response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015652section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015653
15654It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010015655time. Non-existent headers are logged as empty strings, and if one header
15656appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015657are grouped within braces '{' and '}' in the same order as they were declared,
15658and delimited with a vertical bar '|' without any space. Response headers
15659follow the same representation, but are displayed after a space following the
15660request headers block. These blocks are displayed just before the HTTP request
15661in the logs.
15662
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020015663As a special case, it is possible to specify an HTTP header capture in a TCP
15664frontend. The purpose is to enable logging of headers which will be parsed in
15665an HTTP backend if the request is then switched to this HTTP backend.
15666
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015667 Example :
15668 # This instance chains to the outgoing proxy
15669 listen proxy-out
15670 mode http
15671 option httplog
15672 option logasap
15673 log global
15674 server cache1 192.168.1.1:3128
15675
15676 # log the name of the virtual server
15677 capture request header Host len 20
15678
15679 # log the amount of data uploaded during a POST
15680 capture request header Content-Length len 10
15681
15682 # log the beginning of the referrer
15683 capture request header Referer len 20
15684
15685 # server name (useful for outgoing proxies only)
15686 capture response header Server len 20
15687
15688 # logging the content-length is useful with "option logasap"
15689 capture response header Content-Length len 10
15690
15691 # log the expected cache behaviour on the response
15692 capture response header Cache-Control len 8
15693
15694 # the Via header will report the next proxy's name
15695 capture response header Via len 20
15696
15697 # log the URL location during a redirection
15698 capture response header Location len 20
15699
15700 >>> Aug 9 20:26:09 localhost \
15701 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
15702 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
15703 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
15704 "GET http://fr.adserver.yahoo.com/"
15705
15706 >>> Aug 9 20:30:46 localhost \
15707 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
15708 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
15709 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010015710 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015711
15712 >>> Aug 9 20:30:46 localhost \
15713 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
15714 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
15715 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
15716 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010015717 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015718
15719
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200157208.9. Examples of logs
15721---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015722
15723These are real-world examples of logs accompanied with an explanation. Some of
15724them have been made up by hand. The syslog part has been removed for better
15725reading. Their sole purpose is to explain how to decipher them.
15726
15727 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
15728 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
15729 "HEAD / HTTP/1.0"
15730
15731 => long request (6.5s) entered by hand through 'telnet'. The server replied
15732 in 147 ms, and the session ended normally ('----')
15733
15734 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
15735 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
15736 0/9 "HEAD / HTTP/1.0"
15737
15738 => Idem, but the request was queued in the global queue behind 9 other
15739 requests, and waited there for 1230 ms.
15740
15741 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
15742 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
15743 "GET /image.iso HTTP/1.0"
15744
15745 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010015746 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015747 14 ms, 243 bytes of headers were sent to the client, and total time from
15748 accept to first data byte is 30 ms.
15749
15750 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
15751 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
15752 "GET /cgi-bin/bug.cgi? HTTP/1.0"
15753
15754 => the proxy blocked a server response either because of an "rspdeny" or
15755 "rspideny" filter, or because the response was improperly formatted and
Willy Tarreau3c92c5f2011-08-28 09:45:47 +020015756 not HTTP-compliant, or because it blocked sensitive information which
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015757 risked being cached. In this case, the response is replaced with a "502
15758 bad gateway". The flags ("PH--") tell us that it was haproxy who decided
15759 to return the 502 and not the server.
15760
15761 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010015762 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 +010015763
15764 => the client never completed its request and aborted itself ("C---") after
15765 8.5s, while the proxy was waiting for the request headers ("-R--").
15766 Nothing was sent to any server.
15767
15768 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
15769 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
15770
15771 => The client never completed its request, which was aborted by the
15772 time-out ("c---") after 50s, while the proxy was waiting for the request
15773 headers ("-R--"). Nothing was sent to any server, but the proxy could
15774 send a 408 return code to the client.
15775
15776 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
15777 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
15778
15779 => This log was produced with "option tcplog". The client timed out after
15780 5 seconds ("c----").
15781
15782 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
15783 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010015784 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015785
15786 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015787 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015788 (config says 'retries 3'), and no redispatch (otherwise we would have
15789 seen "/+3"). Status code 503 was returned to the client. There were 115
15790 connections on this server, 202 connections on this proxy, and 205 on
15791 the global process. It is possible that the server refused the
15792 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010015793
Willy Tarreau52b2d222011-09-07 23:48:48 +020015794
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200157959. Supported filters
15796--------------------
15797
15798Here are listed officially supported filters with the list of parameters they
15799accept. Depending on compile options, some of these filters might be
15800unavailable. The list of available filters is reported in haproxy -vv.
15801
15802See also : "filter"
15803
158049.1. Trace
15805----------
15806
15807filter trace [name <name>] [random-parsing] [random-forwarding]
15808
15809 Arguments:
15810 <name> is an arbitrary name that will be reported in
15811 messages. If no name is provided, "TRACE" is used.
15812
15813 <random-parsing> enables the random parsing of data exchanged between
15814 the client and the server. By default, this filter
15815 parses all available data. With this parameter, it
15816 only parses a random amount of the available data.
15817
15818 <random-forwarding> enables the random forwading of parsed data. By
15819 default, this filter forwards all previously parsed
15820 data. With this parameter, it only forwards a random
15821 amount of the parsed data.
15822
15823This filter can be used as a base to develop new filters. It defines all
15824callbacks and print a message on the standard error stream (stderr) with useful
15825information for all of them. It may be useful to debug the activity of other
15826filters or, quite simply, HAProxy's activity.
15827
15828Using <random-parsing> and/or <random-forwarding> parameters is a good way to
15829tests the behavior of a filter that parses data exchanged between a client and
15830a server by adding some latencies in the processing.
15831
15832
158339.2. HTTP compression
15834---------------------
15835
15836filter compression
15837
15838The HTTP compression has been moved in a filter in HAProxy 1.7. "compression"
15839keyword must still be used to enable and configure the HTTP compression. And
15840when no other filter is used, it is enough. But it is mandatory to explicitly
15841use a filter line to enable the HTTP compression when two or more filters are
15842used for the same listener/frontend/backend. This is important to know the
15843filters evaluation order.
15844
15845See also : "compression"
15846
15847
Willy Tarreau0ba27502007-12-24 16:55:16 +010015848/*
15849 * Local variables:
15850 * fill-column: 79
15851 * End:
15852 */