blob: ec851d80d66c9668ab0beaff39cbf3db5cf3bc5a [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 Tarreaud5d890b2016-11-09 23:18:17 +01007 2016/11/09
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
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +02001099.3. Stream Processing Offload Engine (SPOE)
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200110
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200111
1121. Quick reminder about HTTP
113----------------------------
114
115When haproxy is running in HTTP mode, both the request and the response are
116fully analyzed and indexed, thus it becomes possible to build matching criteria
117on almost anything found in the contents.
118
119However, it is important to understand how HTTP requests and responses are
120formed, and how HAProxy decomposes them. It will then become easier to write
121correct rules and to debug existing configurations.
122
123
1241.1. The HTTP transaction model
125-------------------------------
126
127The HTTP protocol is transaction-driven. This means that each request will lead
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100128to one and only one response. Traditionally, a TCP connection is established
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200129from the client to the server, a request is sent by the client on the
130connection, the server responds and the connection is closed. A new request
131will involve a new connection :
132
133 [CON1] [REQ1] ... [RESP1] [CLO1] [CON2] [REQ2] ... [RESP2] [CLO2] ...
134
135In this mode, called the "HTTP close" mode, there are as many connection
136establishments as there are HTTP transactions. Since the connection is closed
137by the server after the response, the client does not need to know the content
138length.
139
140Due to the transactional nature of the protocol, it was possible to improve it
141to avoid closing a connection between two subsequent transactions. In this mode
142however, it is mandatory that the server indicates the content length for each
143response so that the client does not wait indefinitely. For this, a special
144header is used: "Content-length". This mode is called the "keep-alive" mode :
145
146 [CON] [REQ1] ... [RESP1] [REQ2] ... [RESP2] [CLO] ...
147
148Its advantages are a reduced latency between transactions, and less processing
149power required on the server side. It is generally better than the close mode,
150but not always because the clients often limit their concurrent connections to
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200151a smaller value.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200152
153A last improvement in the communications is the pipelining mode. It still uses
154keep-alive, but the client does not wait for the first response to send the
155second request. This is useful for fetching large number of images composing a
156page :
157
158 [CON] [REQ1] [REQ2] ... [RESP1] [RESP2] [CLO] ...
159
160This can obviously have a tremendous benefit on performance because the network
161latency is eliminated between subsequent requests. Many HTTP agents do not
162correctly support pipelining since there is no way to associate a response with
163the corresponding request in HTTP. For this reason, it is mandatory for the
Cyril Bonté78caf842010-03-10 22:41:43 +0100164server to reply in the exact same order as the requests were received.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200165
Willy Tarreau70dffda2014-01-30 03:07:23 +0100166By default HAProxy operates in keep-alive mode with regards to persistent
167connections: for each connection it processes each request and response, and
168leaves the connection idle on both sides between the end of a response and the
169start of a new request.
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200170
Willy Tarreau70dffda2014-01-30 03:07:23 +0100171HAProxy supports 5 connection modes :
172 - keep alive : all requests and responses are processed (default)
173 - tunnel : only the first request and response are processed,
174 everything else is forwarded with no analysis.
175 - passive close : tunnel with "Connection: close" added in both directions.
176 - server close : the server-facing connection is closed after the response.
177 - forced close : the connection is actively closed after end of response.
178
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200179
1801.2. HTTP request
181-----------------
182
183First, let's consider this HTTP request :
184
185 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100186 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200187 1 GET /serv/login.php?lang=en&profile=2 HTTP/1.1
188 2 Host: www.mydomain.com
189 3 User-agent: my small browser
190 4 Accept: image/jpeg, image/gif
191 5 Accept: image/png
192
193
1941.2.1. The Request line
195-----------------------
196
197Line 1 is the "request line". It is always composed of 3 fields :
198
199 - a METHOD : GET
200 - a URI : /serv/login.php?lang=en&profile=2
201 - a version tag : HTTP/1.1
202
203All of them are delimited by what the standard calls LWS (linear white spaces),
204which are commonly spaces, but can also be tabs or line feeds/carriage returns
205followed by spaces/tabs. The method itself cannot contain any colon (':') and
206is limited to alphabetic letters. All those various combinations make it
207desirable that HAProxy performs the splitting itself rather than leaving it to
208the user to write a complex or inaccurate regular expression.
209
210The URI itself can have several forms :
211
212 - A "relative URI" :
213
214 /serv/login.php?lang=en&profile=2
215
216 It is a complete URL without the host part. This is generally what is
217 received by servers, reverse proxies and transparent proxies.
218
219 - An "absolute URI", also called a "URL" :
220
221 http://192.168.0.12:8080/serv/login.php?lang=en&profile=2
222
223 It is composed of a "scheme" (the protocol name followed by '://'), a host
224 name or address, optionally a colon (':') followed by a port number, then
225 a relative URI beginning at the first slash ('/') after the address part.
226 This is generally what proxies receive, but a server supporting HTTP/1.1
227 must accept this form too.
228
229 - a star ('*') : this form is only accepted in association with the OPTIONS
230 method and is not relayable. It is used to inquiry a next hop's
231 capabilities.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100232
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200233 - an address:port combination : 192.168.0.12:80
234 This is used with the CONNECT method, which is used to establish TCP
235 tunnels through HTTP proxies, generally for HTTPS, but sometimes for
236 other protocols too.
237
238In a relative URI, two sub-parts are identified. The part before the question
239mark is called the "path". It is typically the relative path to static objects
240on the server. The part after the question mark is called the "query string".
241It is mostly used with GET requests sent to dynamic scripts and is very
242specific to the language, framework or application in use.
243
244
2451.2.2. The request headers
246--------------------------
247
248The headers start at the second line. They are composed of a name at the
249beginning of the line, immediately followed by a colon (':'). Traditionally,
250an LWS is added after the colon but that's not required. Then come the values.
251Multiple identical headers may be folded into one single line, delimiting the
252values with commas, provided that their order is respected. This is commonly
253encountered in the "Cookie:" field. A header may span over multiple lines if
254the subsequent lines begin with an LWS. In the example in 1.2, lines 4 and 5
255define a total of 3 values for the "Accept:" header.
256
257Contrary to a common mis-conception, header names are not case-sensitive, and
258their values are not either if they refer to other header names (such as the
259"Connection:" header).
260
261The end of the headers is indicated by the first empty line. People often say
262that it's a double line feed, which is not exact, even if a double line feed
263is one valid form of empty line.
264
265Fortunately, HAProxy takes care of all these complex combinations when indexing
266headers, checking values and counting them, so there is no reason to worry
267about the way they could be written, but it is important not to accuse an
268application of being buggy if it does unusual, valid things.
269
270Important note:
271 As suggested by RFC2616, HAProxy normalizes headers by replacing line breaks
272 in the middle of headers by LWS in order to join multi-line headers. This
273 is necessary for proper analysis and helps less capable HTTP parsers to work
274 correctly and not to be fooled by such complex constructs.
275
276
2771.3. HTTP response
278------------------
279
280An HTTP response looks very much like an HTTP request. Both are called HTTP
281messages. Let's consider this HTTP response :
282
283 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100284 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200285 1 HTTP/1.1 200 OK
286 2 Content-length: 350
287 3 Content-Type: text/html
288
Willy Tarreau816b9792009-09-15 21:25:21 +0200289As a special case, HTTP supports so called "Informational responses" as status
290codes 1xx. These messages are special in that they don't convey any part of the
291response, they're just used as sort of a signaling message to ask a client to
Willy Tarreau5843d1a2010-02-01 15:13:32 +0100292continue to post its request for instance. In the case of a status 100 response
293the requested information will be carried by the next non-100 response message
294following the informational one. This implies that multiple responses may be
295sent to a single request, and that this only works when keep-alive is enabled
296(1xx messages are HTTP/1.1 only). HAProxy handles these messages and is able to
297correctly forward and skip them, and only process the next non-100 response. As
298such, these messages are neither logged nor transformed, unless explicitly
299state otherwise. Status 101 messages indicate that the protocol is changing
300over the same connection and that haproxy must switch to tunnel mode, just as
301if a CONNECT had occurred. Then the Upgrade header would contain additional
302information about the type of protocol the connection is switching to.
Willy Tarreau816b9792009-09-15 21:25:21 +0200303
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200304
3051.3.1. The Response line
306------------------------
307
308Line 1 is the "response line". It is always composed of 3 fields :
309
310 - a version tag : HTTP/1.1
311 - a status code : 200
312 - a reason : OK
313
314The status code is always 3-digit. The first digit indicates a general status :
Willy Tarreau816b9792009-09-15 21:25:21 +0200315 - 1xx = informational message to be skipped (eg: 100, 101)
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200316 - 2xx = OK, content is following (eg: 200, 206)
317 - 3xx = OK, no content following (eg: 302, 304)
318 - 4xx = error caused by the client (eg: 401, 403, 404)
319 - 5xx = error caused by the server (eg: 500, 502, 503)
320
321Please refer to RFC2616 for the detailed meaning of all such codes. The
Willy Tarreaud72758d2010-01-12 10:42:19 +0100322"reason" field is just a hint, but is not parsed by clients. Anything can be
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200323found there, but it's a common practice to respect the well-established
324messages. It can be composed of one or multiple words, such as "OK", "Found",
325or "Authentication Required".
326
327Haproxy may emit the following status codes by itself :
328
329 Code When / reason
330 200 access to stats page, and when replying to monitoring requests
331 301 when performing a redirection, depending on the configured code
332 302 when performing a redirection, depending on the configured code
333 303 when performing a redirection, depending on the configured code
Willy Tarreaub67fdc42013-03-29 19:28:11 +0100334 307 when performing a redirection, depending on the configured code
335 308 when performing a redirection, depending on the configured code
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200336 400 for an invalid or too large request
337 401 when an authentication is required to perform the action (when
338 accessing the stats page)
339 403 when a request is forbidden by a "block" ACL or "reqdeny" filter
340 408 when the request timeout strikes before the request is complete
341 500 when haproxy encounters an unrecoverable internal error, such as a
342 memory allocation failure, which should never happen
343 502 when the server returns an empty, invalid or incomplete response, or
344 when an "rspdeny" filter blocks the response.
345 503 when no server was available to handle the request, or in response to
346 monitoring requests which match the "monitor fail" condition
347 504 when the response timeout strikes before the server responds
348
349The error 4xx and 5xx codes above may be customized (see "errorloc" in section
3504.2).
351
352
3531.3.2. The response headers
354---------------------------
355
356Response headers work exactly like request headers, and as such, HAProxy uses
357the same parsing function for both. Please refer to paragraph 1.2.2 for more
358details.
359
360
3612. Configuring HAProxy
362----------------------
363
3642.1. Configuration file format
365------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +0200366
367HAProxy's configuration process involves 3 major sources of parameters :
368
369 - the arguments from the command-line, which always take precedence
370 - the "global" section, which sets process-wide parameters
371 - the proxies sections which can take form of "defaults", "listen",
372 "frontend" and "backend".
373
Willy Tarreau0ba27502007-12-24 16:55:16 +0100374The configuration file syntax consists in lines beginning with a keyword
375referenced in this manual, optionally followed by one or several parameters
William Lallemandf9873ba2015-05-05 17:37:14 +0200376delimited by spaces.
Willy Tarreau0ba27502007-12-24 16:55:16 +0100377
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200378
William Lallemandf9873ba2015-05-05 17:37:14 +02003792.2. Quoting and escaping
380-------------------------
381
382HAProxy's configuration introduces a quoting and escaping system similar to
383many programming languages. The configuration file supports 3 types: escaping
384with a backslash, weak quoting with double quotes, and strong quoting with
385single quotes.
386
387If spaces have to be entered in strings, then they must be escaped by preceding
388them by a backslash ('\') or by quoting them. Backslashes also have to be
389escaped by doubling or strong quoting them.
390
391Escaping is achieved by preceding a special character by a backslash ('\'):
392
393 \ to mark a space and differentiate it from a delimiter
394 \# to mark a hash and differentiate it from a comment
395 \\ to use a backslash
396 \' to use a single quote and differentiate it from strong quoting
397 \" to use a double quote and differentiate it from weak quoting
398
399Weak quoting is achieved by using double quotes (""). Weak quoting prevents
400the interpretation of:
401
402 space as a parameter separator
403 ' single quote as a strong quoting delimiter
404 # hash as a comment start
405
William Lallemandb2f07452015-05-12 14:27:13 +0200406Weak quoting permits the interpretation of variables, if you want to use a non
407-interpreted dollar within a double quoted string, you should escape it with a
408backslash ("\$"), it does not work outside weak quoting.
409
410Interpretation of escaping and special characters are not prevented by weak
William Lallemandf9873ba2015-05-05 17:37:14 +0200411quoting.
412
413Strong quoting is achieved by using single quotes (''). Inside single quotes,
414nothing is interpreted, it's the efficient way to quote regexes.
415
416Quoted and escaped strings are replaced in memory by their interpreted
417equivalent, it allows you to perform concatenation.
418
419 Example:
420 # those are equivalents:
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 log-format "%{+Q}o %t"' %s'\ %{-Q}r
426
427 # those are equivalents:
428 reqrep "^([^\ :]*)\ /static/(.*)" \1\ /\2
429 reqrep "^([^ :]*)\ /static/(.*)" '\1 /\2'
430 reqrep "^([^ :]*)\ /static/(.*)" "\1 /\2"
431 reqrep "^([^ :]*)\ /static/(.*)" "\1\ /\2"
432
433
William Lallemandb2f07452015-05-12 14:27:13 +02004342.3. Environment variables
435--------------------------
436
437HAProxy's configuration supports environment variables. Those variables are
438interpreted only within double quotes. Variables are expanded during the
439configuration parsing. Variable names must be preceded by a dollar ("$") and
440optionally enclosed with braces ("{}") similarly to what is done in Bourne
441shell. Variable names can contain alphanumerical characters or the character
442underscore ("_") but should not start with a digit.
443
444 Example:
445
446 bind "fd@${FD_APP1}"
447
448 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
449
450 user "$HAPROXY_USER"
451
452
4532.4. Time format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200454----------------
455
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100456Some parameters involve values representing time, such as timeouts. These
Willy Tarreau0ba27502007-12-24 16:55:16 +0100457values are generally expressed in milliseconds (unless explicitly stated
458otherwise) but may be expressed in any other unit by suffixing the unit to the
459numeric value. It is important to consider this because it will not be repeated
460for every keyword. Supported units are :
461
462 - us : microseconds. 1 microsecond = 1/1000000 second
463 - ms : milliseconds. 1 millisecond = 1/1000 second. This is the default.
464 - s : seconds. 1s = 1000ms
465 - m : minutes. 1m = 60s = 60000ms
466 - h : hours. 1h = 60m = 3600s = 3600000ms
467 - d : days. 1d = 24h = 1440m = 86400s = 86400000ms
468
469
William Lallemandf9873ba2015-05-05 17:37:14 +02004702.4. Examples
Patrick Mezard35da19c2010-06-12 17:02:47 +0200471-------------
472
473 # Simple configuration for an HTTP proxy listening on port 80 on all
474 # interfaces and forwarding requests to a single backend "servers" with a
475 # single server "server1" listening on 127.0.0.1:8000
476 global
477 daemon
478 maxconn 256
479
480 defaults
481 mode http
482 timeout connect 5000ms
483 timeout client 50000ms
484 timeout server 50000ms
485
486 frontend http-in
487 bind *:80
488 default_backend servers
489
490 backend servers
491 server server1 127.0.0.1:8000 maxconn 32
492
493
494 # The same configuration defined with a single listen block. Shorter but
495 # less expressive, especially in HTTP mode.
496 global
497 daemon
498 maxconn 256
499
500 defaults
501 mode http
502 timeout connect 5000ms
503 timeout client 50000ms
504 timeout server 50000ms
505
506 listen http-in
507 bind *:80
508 server server1 127.0.0.1:8000 maxconn 32
509
510
511Assuming haproxy is in $PATH, test these configurations in a shell with:
512
Willy Tarreauccb289d2010-12-11 20:19:38 +0100513 $ sudo haproxy -f configuration.conf -c
Patrick Mezard35da19c2010-06-12 17:02:47 +0200514
515
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005163. Global parameters
Willy Tarreau6a06a402007-07-15 20:15:28 +0200517--------------------
518
519Parameters in the "global" section are process-wide and often OS-specific. They
520are generally set once for all and do not need being changed once correct. Some
521of them have command-line equivalents.
522
523The following keywords are supported in the "global" section :
524
525 * Process management and security
Emeric Brunc8e8d122012-10-02 18:42:10 +0200526 - ca-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200527 - chroot
Emeric Brunc8e8d122012-10-02 18:42:10 +0200528 - crt-base
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200529 - cpu-map
Willy Tarreau6a06a402007-07-15 20:15:28 +0200530 - daemon
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200531 - description
532 - deviceatlas-json-file
533 - deviceatlas-log-level
534 - deviceatlas-separator
535 - deviceatlas-properties-cookie
Simon Horman98637e52014-06-20 12:30:16 +0900536 - external-check
Willy Tarreau6a06a402007-07-15 20:15:28 +0200537 - gid
538 - group
539 - log
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200540 - log-tag
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100541 - log-send-hostname
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200542 - lua-load
Willy Tarreau6a06a402007-07-15 20:15:28 +0200543 - nbproc
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200544 - node
Willy Tarreau6a06a402007-07-15 20:15:28 +0200545 - pidfile
Willy Tarreau1d549722016-02-16 12:41:57 +0100546 - presetenv
547 - resetenv
Willy Tarreau6a06a402007-07-15 20:15:28 +0200548 - uid
549 - ulimit-n
550 - user
Willy Tarreau1d549722016-02-16 12:41:57 +0100551 - setenv
Willy Tarreaufbee7132007-10-18 13:53:22 +0200552 - stats
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200553 - ssl-default-bind-ciphers
554 - ssl-default-bind-options
555 - ssl-default-server-ciphers
556 - ssl-default-server-options
557 - ssl-dh-param-file
Emeric Brun850efd52014-01-29 12:24:34 +0100558 - ssl-server-verify
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100559 - unix-bind
Willy Tarreau1d549722016-02-16 12:41:57 +0100560 - unsetenv
Thomas Holmesdb04f192015-05-18 13:21:39 +0100561 - 51degrees-data-file
562 - 51degrees-property-name-list
Dragan Dosen93b38d92015-06-29 16:43:25 +0200563 - 51degrees-property-separator
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200564 - 51degrees-cache-size
scientiamobiled0027ed2016-11-04 10:55:08 +0100565 - wurfl-data-file
566 - wurfl-information-list
567 - wurfl-information-list-separator
568 - wurfl-engine-mode
569 - wurfl-cache-size
570 - wurfl-useragent-priority
Willy Tarreaud72758d2010-01-12 10:42:19 +0100571
Willy Tarreau6a06a402007-07-15 20:15:28 +0200572 * Performance tuning
Willy Tarreau1746eec2014-04-25 10:46:47 +0200573 - max-spread-checks
Willy Tarreau6a06a402007-07-15 20:15:28 +0200574 - maxconn
Willy Tarreau81c25d02011-09-07 15:17:21 +0200575 - maxconnrate
William Lallemandd85f9172012-11-09 17:05:39 +0100576 - maxcomprate
William Lallemand072a2bf2012-11-20 17:01:01 +0100577 - maxcompcpuusage
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100578 - maxpipes
Willy Tarreau93e7c002013-10-07 18:51:07 +0200579 - maxsessrate
Willy Tarreau403edff2012-09-06 11:58:37 +0200580 - maxsslconn
Willy Tarreaue43d5322013-10-07 20:01:52 +0200581 - maxsslrate
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200582 - maxzlibmem
Willy Tarreau6a06a402007-07-15 20:15:28 +0200583 - noepoll
584 - nokqueue
585 - nopoll
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100586 - nosplice
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300587 - nogetaddrinfo
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +0000588 - noreuseport
Willy Tarreaufe255b72007-10-14 23:09:26 +0200589 - spread-checks
Baptiste Assmann5626f482015-08-23 10:00:10 +0200590 - server-state-base
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200591 - server-state-file
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200592 - tune.buffers.limit
593 - tune.buffers.reserve
Willy Tarreau27a674e2009-08-17 07:23:33 +0200594 - tune.bufsize
Willy Tarreau43961d52010-10-04 20:39:20 +0200595 - tune.chksize
William Lallemandf3747832012-11-09 12:33:10 +0100596 - tune.comp.maxlevel
Willy Tarreau193b8c62012-11-22 00:17:38 +0100597 - tune.http.cookielen
Willy Tarreauac1932d2011-10-24 19:14:41 +0200598 - tune.http.maxhdr
Willy Tarreau7e312732014-02-12 16:35:14 +0100599 - tune.idletimer
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100600 - tune.lua.forced-yield
Willy Tarreau32f61e22015-03-18 17:54:59 +0100601 - tune.lua.maxmem
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100602 - tune.lua.session-timeout
603 - tune.lua.task-timeout
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +0200604 - tune.lua.service-timeout
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100605 - tune.maxaccept
606 - tune.maxpollevents
Willy Tarreau27a674e2009-08-17 07:23:33 +0200607 - tune.maxrewrite
Willy Tarreauf3045d22015-04-29 16:24:50 +0200608 - tune.pattern.cache-size
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200609 - tune.pipesize
Willy Tarreaue803de22010-01-21 17:43:04 +0100610 - tune.rcvbuf.client
611 - tune.rcvbuf.server
Willy Tarreaub22fc302015-12-14 12:04:35 +0100612 - tune.recv_enough
Willy Tarreaue803de22010-01-21 17:43:04 +0100613 - tune.sndbuf.client
614 - tune.sndbuf.server
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100615 - tune.ssl.cachesize
Willy Tarreaubfd59462013-02-21 07:46:09 +0100616 - tune.ssl.lifetime
Emeric Brun8dc60392014-05-09 13:52:00 +0200617 - tune.ssl.force-private-cache
Willy Tarreaubfd59462013-02-21 07:46:09 +0100618 - tune.ssl.maxrecord
Remi Gacognef46cd6e2014-06-12 14:58:40 +0200619 - tune.ssl.default-dh-param
Christopher Faulet31af49d2015-06-09 17:29:50 +0200620 - tune.ssl.ssl-ctx-cache-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200621 - tune.vars.global-max-size
Christopher Fauletff2613e2016-11-09 11:36:17 +0100622 - tune.vars.proc-max-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200623 - tune.vars.reqres-max-size
624 - tune.vars.sess-max-size
625 - tune.vars.txn-max-size
William Lallemanda509e4c2012-11-07 16:54:34 +0100626 - tune.zlib.memlevel
627 - tune.zlib.windowsize
Willy Tarreaud72758d2010-01-12 10:42:19 +0100628
Willy Tarreau6a06a402007-07-15 20:15:28 +0200629 * Debugging
630 - debug
631 - quiet
Willy Tarreau6a06a402007-07-15 20:15:28 +0200632
633
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006343.1. Process management and security
Willy Tarreau6a06a402007-07-15 20:15:28 +0200635------------------------------------
636
Emeric Brunc8e8d122012-10-02 18:42:10 +0200637ca-base <dir>
638 Assigns a default directory to fetch SSL CA certificates and CRLs from when a
Emeric Brunfd33a262012-10-11 16:28:27 +0200639 relative path is used with "ca-file" or "crl-file" directives. Absolute
640 locations specified in "ca-file" and "crl-file" prevail and ignore "ca-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +0200641
Willy Tarreau6a06a402007-07-15 20:15:28 +0200642chroot <jail dir>
643 Changes current directory to <jail dir> and performs a chroot() there before
644 dropping privileges. This increases the security level in case an unknown
645 vulnerability would be exploited, since it would make it very hard for the
646 attacker to exploit the system. This only works when the process is started
647 with superuser privileges. It is important to ensure that <jail_dir> is both
648 empty and unwritable to anyone.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100649
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100650cpu-map <"all"|"odd"|"even"|process_num> <cpu-set>...
651 On Linux 2.6 and above, it is possible to bind a process to a specific CPU
652 set. This means that the process will never run on other CPUs. The "cpu-map"
653 directive specifies CPU sets for process sets. The first argument is the
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100654 process number to bind. This process must have a number between 1 and 32 or
655 64, depending on the machine's word size, and any process IDs above nbproc
656 are ignored. It is possible to specify all processes at once using "all",
657 only odd numbers using "odd" or even numbers using "even", just like with the
658 "bind-process" directive. The second and forthcoming arguments are CPU sets.
659 Each CPU set is either a unique number between 0 and 31 or 63 or a range with
660 two such numbers delimited by a dash ('-'). Multiple CPU numbers or ranges
661 may be specified, and the processes will be allowed to bind to all of them.
662 Obviously, multiple "cpu-map" directives may be specified. Each "cpu-map"
663 directive will replace the previous ones when they overlap.
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100664
Emeric Brunc8e8d122012-10-02 18:42:10 +0200665crt-base <dir>
666 Assigns a default directory to fetch SSL certificates from when a relative
667 path is used with "crtfile" directives. Absolute locations specified after
668 "crtfile" prevail and ignore "crt-base".
669
Willy Tarreau6a06a402007-07-15 20:15:28 +0200670daemon
671 Makes the process fork into background. This is the recommended mode of
672 operation. It is equivalent to the command line "-D" argument. It can be
673 disabled by the command line "-db" argument.
674
David Carlier8167f302015-06-01 13:50:06 +0200675deviceatlas-json-file <path>
676 Sets the path of the DeviceAtlas JSON data file to be loaded by the API.
677 The path must be a valid JSON data file and accessible by Haproxy process.
678
679deviceatlas-log-level <value>
680 Sets the level of informations returned by the API. This directive is
681 optional and set to 0 by default if not set.
682
683deviceatlas-separator <char>
684 Sets the character separator for the API properties results. This directive
685 is optional and set to | by default if not set.
686
Cyril Bonté0306c4a2015-10-26 22:37:38 +0100687deviceatlas-properties-cookie <name>
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200688 Sets the client cookie's name used for the detection if the DeviceAtlas
689 Client-side component was used during the request. This directive is optional
690 and set to DAPROPS by default if not set.
David Carlier29b3ca32015-09-25 14:09:21 +0100691
Simon Horman98637e52014-06-20 12:30:16 +0900692external-check
693 Allows the use of an external agent to perform health checks.
694 This is disabled by default as a security precaution.
695 See "option external-check".
696
Willy Tarreau6a06a402007-07-15 20:15:28 +0200697gid <number>
698 Changes the process' group ID to <number>. It is recommended that the group
699 ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
700 be started with a user belonging to this group, or with superuser privileges.
Michael Schererab012dd2013-01-12 18:35:19 +0100701 Note that if haproxy is started from a user having supplementary groups, it
702 will only be able to drop these groups if started with superuser privileges.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200703 See also "group" and "uid".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100704
Willy Tarreau6a06a402007-07-15 20:15:28 +0200705group <group name>
706 Similar to "gid" but uses the GID of group name <group name> from /etc/group.
707 See also "gid" and "user".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100708
Dragan Dosen7ad31542015-09-28 17:16:47 +0200709log <address> [len <length>] [format <format>] <facility> [max level [min level]]
Willy Tarreau6a06a402007-07-15 20:15:28 +0200710 Adds a global syslog server. Up to two global servers can be defined. They
711 will receive logs for startups and exits, as well as all logs from proxies
Robert Tsai81ae1952007-12-05 10:47:29 +0100712 configured with "log global".
713
714 <address> can be one of:
715
Willy Tarreau2769aa02007-12-27 18:26:09 +0100716 - An IPv4 address optionally followed by a colon and a UDP port. If
Robert Tsai81ae1952007-12-05 10:47:29 +0100717 no port is specified, 514 is used by default (the standard syslog
718 port).
719
David du Colombier24bb5f52011-03-17 10:40:23 +0100720 - An IPv6 address followed by a colon and optionally a UDP port. If
721 no port is specified, 514 is used by default (the standard syslog
722 port).
723
Robert Tsai81ae1952007-12-05 10:47:29 +0100724 - A filesystem path to a UNIX domain socket, keeping in mind
725 considerations for chroot (be sure the path is accessible inside
726 the chroot) and uid/gid (be sure the path is appropriately
727 writeable).
728
William Lallemandb2f07452015-05-12 14:27:13 +0200729 You may want to reference some environment variables in the address
730 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +0100731
Willy Tarreau18324f52014-06-27 18:10:07 +0200732 <length> is an optional maximum line length. Log lines larger than this value
733 will be truncated before being sent. The reason is that syslog
734 servers act differently on log line length. All servers support the
735 default value of 1024, but some servers simply drop larger lines
736 while others do log them. If a server supports long lines, it may
737 make sense to set this value here in order to avoid truncating long
738 lines. Similarly, if a server drops long lines, it is preferable to
739 truncate them before sending them. Accepted values are 80 to 65535
740 inclusive. The default value of 1024 is generally fine for all
741 standard usages. Some specific cases of long captures or
742 JSON-formated logs may require larger values.
743
Dragan Dosen7ad31542015-09-28 17:16:47 +0200744 <format> is the log format used when generating syslog messages. It may be
745 one of the following :
746
747 rfc3164 The RFC3164 syslog message format. This is the default.
748 (https://tools.ietf.org/html/rfc3164)
749
750 rfc5424 The RFC5424 syslog message format.
751 (https://tools.ietf.org/html/rfc5424)
752
Robert Tsai81ae1952007-12-05 10:47:29 +0100753 <facility> must be one of the 24 standard syslog facilities :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200754
755 kern user mail daemon auth syslog lpr news
756 uucp cron auth2 ftp ntp audit alert cron2
757 local0 local1 local2 local3 local4 local5 local6 local7
758
759 An optional level can be specified to filter outgoing messages. By default,
Willy Tarreauf7edefa2009-05-10 17:20:05 +0200760 all messages are sent. If a maximum level is specified, only messages with a
761 severity at least as important as this level will be sent. An optional minimum
762 level can be specified. If it is set, logs emitted with a more severe level
763 than this one will be capped to this level. This is used to avoid sending
764 "emerg" messages on all terminals on some default syslog configurations.
765 Eight levels are known :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200766
Cyril Bontédc4d9032012-04-08 21:57:39 +0200767 emerg alert crit err warning notice info debug
Willy Tarreau6a06a402007-07-15 20:15:28 +0200768
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100769log-send-hostname [<string>]
770 Sets the hostname field in the syslog header. If optional "string" parameter
771 is set the header is set to the string contents, otherwise uses the hostname
772 of the system. Generally used if one is not relaying logs through an
773 intermediate syslog server or for simply customizing the hostname printed in
774 the logs.
775
Kevinm48936af2010-12-22 16:08:21 +0000776log-tag <string>
777 Sets the tag field in the syslog header to this string. It defaults to the
778 program name as launched from the command line, which usually is "haproxy".
779 Sometimes it can be useful to differentiate between multiple processes
Willy Tarreau094af4e2015-01-07 15:03:42 +0100780 running on the same host. See also the per-proxy "log-tag" directive.
Kevinm48936af2010-12-22 16:08:21 +0000781
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100782lua-load <file>
783 This global directive loads and executes a Lua file. This directive can be
784 used multiple times.
785
Willy Tarreau6a06a402007-07-15 20:15:28 +0200786nbproc <number>
787 Creates <number> processes when going daemon. This requires the "daemon"
788 mode. By default, only one process is created, which is the recommended mode
789 of operation. For systems limited to small sets of file descriptors per
790 process, it may be needed to fork multiple daemons. USING MULTIPLE PROCESSES
791 IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. See also "daemon".
792
793pidfile <pidfile>
794 Writes pids of all daemons into file <pidfile>. This option is equivalent to
795 the "-p" command line argument. The file must be accessible to the user
796 starting the process. See also "daemon".
797
Willy Tarreau1d549722016-02-16 12:41:57 +0100798presetenv <name> <value>
799 Sets environment variable <name> to value <value>. If the variable exists, it
800 is NOT overwritten. The changes immediately take effect so that the next line
801 in the configuration file sees the new value. See also "setenv", "resetenv",
802 and "unsetenv".
803
804resetenv [<name> ...]
805 Removes all environment variables except the ones specified in argument. It
806 allows to use a clean controlled environment before setting new values with
807 setenv or unsetenv. Please note that some internal functions may make use of
808 some environment variables, such as time manipulation functions, but also
809 OpenSSL or even external checks. This must be used with extreme care and only
810 after complete validation. The changes immediately take effect so that the
811 next line in the configuration file sees the new environment. See also
812 "setenv", "presetenv", and "unsetenv".
813
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100814stats bind-process [ all | odd | even | <number 1-64>[-<number 1-64>] ] ...
Willy Tarreau35b7b162012-10-22 23:17:18 +0200815 Limits the stats socket to a certain set of processes numbers. By default the
816 stats socket is bound to all processes, causing a warning to be emitted when
817 nbproc is greater than 1 because there is no way to select the target process
818 when connecting. However, by using this setting, it becomes possible to pin
819 the stats socket to a specific set of processes, typically the first one. The
820 warning will automatically be disabled when this setting is used, whatever
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100821 the number of processes used. The maximum process ID depends on the machine's
Willy Tarreauae302532014-05-07 19:22:24 +0200822 word size (32 or 64). A better option consists in using the "process" setting
823 of the "stats socket" line to force the process on each line.
Willy Tarreau35b7b162012-10-22 23:17:18 +0200824
Baptiste Assmann5626f482015-08-23 10:00:10 +0200825server-state-base <directory>
826 Specifies the directory prefix to be prepended in front of all servers state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +0200827 file names which do not start with a '/'. See also "server-state-file",
828 "load-server-state-from-file" and "server-state-file-name".
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200829
830server-state-file <file>
831 Specifies the path to the file containing state of servers. If the path starts
832 with a slash ('/'), it is considered absolute, otherwise it is considered
833 relative to the directory specified using "server-state-base" (if set) or to
834 the current directory. Before reloading HAProxy, it is possible to save the
835 servers' current state using the stats command "show servers state". The
836 output of this command must be written in the file pointed by <file>. When
837 starting up, before handling traffic, HAProxy will read, load and apply state
838 for each server found in the file and available in its current running
Baptiste Assmann01c6cc32015-08-23 11:45:29 +0200839 configuration. See also "server-state-base" and "show servers state",
840 "load-server-state-from-file" and "server-state-file-name"
Baptiste Assmann5626f482015-08-23 10:00:10 +0200841
Willy Tarreau1d549722016-02-16 12:41:57 +0100842setenv <name> <value>
843 Sets environment variable <name> to value <value>. If the variable exists, it
844 is overwritten. The changes immediately take effect so that the next line in
845 the configuration file sees the new value. See also "presetenv", "resetenv",
846 and "unsetenv".
847
Willy Tarreau610f04b2014-02-13 11:36:41 +0100848ssl-default-bind-ciphers <ciphers>
849 This setting is only available when support for OpenSSL was built in. It sets
850 the default string describing the list of cipher algorithms ("cipher suite")
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300851 that are negotiated during the SSL/TLS handshake for all "bind" lines which
Willy Tarreau610f04b2014-02-13 11:36:41 +0100852 do not explicitly define theirs. The format of the string is defined in
853 "man 1 ciphers" from OpenSSL man pages, and can be for instance a string such
854 as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes). Please check the
855 "bind" keyword for more information.
856
Emeric Brun2c86cbf2014-10-30 15:56:50 +0100857ssl-default-bind-options [<option>]...
858 This setting is only available when support for OpenSSL was built in. It sets
859 default ssl-options to force on all "bind" lines. Please check the "bind"
860 keyword to see available options.
861
862 Example:
863 global
864 ssl-default-bind-options no-sslv3 no-tls-tickets
865
Willy Tarreau610f04b2014-02-13 11:36:41 +0100866ssl-default-server-ciphers <ciphers>
867 This setting is only available when support for OpenSSL was built in. It
868 sets the default string describing the list of cipher algorithms that are
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300869 negotiated during the SSL/TLS handshake with the server, for all "server"
Willy Tarreau610f04b2014-02-13 11:36:41 +0100870 lines which do not explicitly define theirs. The format of the string is
871 defined in "man 1 ciphers". Please check the "server" keyword for more
872 information.
873
Emeric Brun2c86cbf2014-10-30 15:56:50 +0100874ssl-default-server-options [<option>]...
875 This setting is only available when support for OpenSSL was built in. It sets
876 default ssl-options to force on all "server" lines. Please check the "server"
877 keyword to see available options.
878
Remi Gacogne47783ef2015-05-29 15:53:22 +0200879ssl-dh-param-file <file>
880 This setting is only available when support for OpenSSL was built in. It sets
881 the default DH parameters that are used during the SSL/TLS handshake when
882 ephemeral Diffie-Hellman (DHE) key exchange is used, for all "bind" lines
883 which do not explicitely define theirs. It will be overridden by custom DH
884 parameters found in a bind certificate file if any. If custom DH parameters
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200885 are not specified either by using ssl-dh-param-file or by setting them
886 directly in the certificate file, pre-generated DH parameters of the size
887 specified by tune.ssl.default-dh-param will be used. Custom parameters are
888 known to be more secure and therefore their use is recommended.
Remi Gacogne47783ef2015-05-29 15:53:22 +0200889 Custom DH parameters may be generated by using the OpenSSL command
890 "openssl dhparam <size>", where size should be at least 2048, as 1024-bit DH
891 parameters should not be considered secure anymore.
892
Emeric Brun850efd52014-01-29 12:24:34 +0100893ssl-server-verify [none|required]
894 The default behavior for SSL verify on servers side. If specified to 'none',
895 servers certificates are not verified. The default is 'required' except if
896 forced using cmdline option '-dV'.
897
Willy Tarreauabb175f2012-09-24 12:43:26 +0200898stats socket [<address:port>|<path>] [param*]
899 Binds a UNIX socket to <path> or a TCPv4/v6 address to <address:port>.
900 Connections to this socket will return various statistics outputs and even
901 allow some commands to be issued to change some runtime settings. Please
Kevin Decherf949c7202015-10-13 23:26:44 +0200902 consult section 9.2 "Unix Socket commands" of Management Guide for more
903 details.
Willy Tarreau6162db22009-10-10 17:13:00 +0200904
Willy Tarreauabb175f2012-09-24 12:43:26 +0200905 All parameters supported by "bind" lines are supported, for instance to
906 restrict access to some users or their access rights. Please consult
907 section 5.1 for more information.
Willy Tarreaufbee7132007-10-18 13:53:22 +0200908
909stats timeout <timeout, in milliseconds>
910 The default timeout on the stats socket is set to 10 seconds. It is possible
911 to change this value with "stats timeout". The value must be passed in
Willy Tarreaubefdff12007-12-02 22:27:38 +0100912 milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.
Willy Tarreaufbee7132007-10-18 13:53:22 +0200913
914stats maxconn <connections>
915 By default, the stats socket is limited to 10 concurrent connections. It is
916 possible to change this value with "stats maxconn".
917
Willy Tarreau6a06a402007-07-15 20:15:28 +0200918uid <number>
919 Changes the process' user ID to <number>. It is recommended that the user ID
920 is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
921 be started with superuser privileges in order to be able to switch to another
922 one. See also "gid" and "user".
923
924ulimit-n <number>
925 Sets the maximum number of per-process file-descriptors to <number>. By
926 default, it is automatically computed, so it is recommended not to use this
927 option.
928
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100929unix-bind [ prefix <prefix> ] [ mode <mode> ] [ user <user> ] [ uid <uid> ]
930 [ group <group> ] [ gid <gid> ]
931
932 Fixes common settings to UNIX listening sockets declared in "bind" statements.
933 This is mainly used to simplify declaration of those UNIX sockets and reduce
934 the risk of errors, since those settings are most commonly required but are
935 also process-specific. The <prefix> setting can be used to force all socket
936 path to be relative to that directory. This might be needed to access another
937 component's chroot. Note that those paths are resolved before haproxy chroots
938 itself, so they are absolute. The <mode>, <user>, <uid>, <group> and <gid>
939 all have the same meaning as their homonyms used by the "bind" statement. If
940 both are specified, the "bind" statement has priority, meaning that the
941 "unix-bind" settings may be seen as process-wide default settings.
942
Willy Tarreau1d549722016-02-16 12:41:57 +0100943unsetenv [<name> ...]
944 Removes environment variables specified in arguments. This can be useful to
945 hide some sensitive information that are occasionally inherited from the
946 user's environment during some operations. Variables which did not exist are
947 silently ignored so that after the operation, it is certain that none of
948 these variables remain. The changes immediately take effect so that the next
949 line in the configuration file will not see these variables. See also
950 "setenv", "presetenv", and "resetenv".
951
Willy Tarreau6a06a402007-07-15 20:15:28 +0200952user <user name>
953 Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.
954 See also "uid" and "group".
955
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +0200956node <name>
957 Only letters, digits, hyphen and underscore are allowed, like in DNS names.
958
959 This statement is useful in HA configurations where two or more processes or
960 servers share the same IP address. By setting a different node-name on all
961 nodes, it becomes easy to immediately spot what server is handling the
962 traffic.
963
964description <text>
965 Add a text that describes the instance.
966
967 Please note that it is required to escape certain characters (# for example)
968 and this text is inserted into a html page so you should avoid using
969 "<" and ">" characters.
970
Thomas Holmesdb04f192015-05-18 13:21:39 +010097151degrees-data-file <file path>
972 The path of the 51Degrees data file to provide device detection services. The
973 file should be unzipped and accessible by HAProxy with relevavnt permissions.
974
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200975 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +0100976 compiled with USE_51DEGREES.
977
97851degrees-property-name-list [<string>]
979 A list of 51Degrees property names to be load from the dataset. A full list
980 of names is available on the 51Degrees website:
981 https://51degrees.com/resources/property-dictionary
982
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200983 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +0100984 compiled with USE_51DEGREES.
985
Dragan Dosen93b38d92015-06-29 16:43:25 +020098651degrees-property-separator <char>
Thomas Holmesdb04f192015-05-18 13:21:39 +0100987 A char that will be appended to every property value in a response header
988 containing 51Degrees results. If not set that will be set as ','.
989
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200990 Please note that this option is only available when haproxy has been
991 compiled with USE_51DEGREES.
992
99351degrees-cache-size <number>
994 Sets the size of the 51Degrees converter cache to <number> entries. This
995 is an LRU cache which reminds previous device detections and their results.
996 By default, this cache is disabled.
997
998 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +0100999 compiled with USE_51DEGREES.
1000
scientiamobiled0027ed2016-11-04 10:55:08 +01001001wurfl-data-file <file path>
1002 The path of the WURFL data file to provide device detection services. The
1003 file should be accessible by HAProxy with relevant permissions.
1004
1005 Please note that this option is only available when haproxy has been compiled
1006 with USE_WURFL=1.
1007
1008wurfl-information-list [<capability>]*
1009 A space-delimited list of WURFL capabilities, virtual capabilities, property
1010 names we plan to use in injected headers. A full list of capability and
1011 virtual capability names is available on the Scientiamobile website :
1012
1013 https://www.scientiamobile.com/wurflCapability
1014
1015 Valid WURFL properties are:
1016 - wurfl_id Contains the device ID of the matched device.
1017
1018 - wurfl_root_id Contains the device root ID of the matched
1019 device.
1020
1021 - wurfl_isdevroot Tells if the matched device is a root device.
1022 Possible values are "TRUE" or "FALSE".
1023
1024 - wurfl_useragent The original useragent coming with this
1025 particular web request.
1026
1027 - wurfl_api_version Contains a string representing the currently
1028 used Libwurfl API version.
1029
1030 - wurfl_engine_target Contains a string representing the currently
1031 set WURFL Engine Target. Possible values are
1032 "HIGH_ACCURACY", "HIGH_PERFORMANCE", "INVALID".
1033
1034 - wurfl_info A string containing information on the parsed
1035 wurfl.xml and its full path.
1036
1037 - wurfl_last_load_time Contains the UNIX timestamp of the last time
1038 WURFL has been loaded successfully.
1039
1040 - wurfl_normalized_useragent The normalized useragent.
1041
1042 - wurfl_useragent_priority The user agent priority used by WURFL.
1043
1044 Please note that this option is only available when haproxy has been compiled
1045 with USE_WURFL=1.
1046
1047wurfl-information-list-separator <char>
1048 A char that will be used to separate values in a response header containing
1049 WURFL results. If not set that a comma (',') will be used by default.
1050
1051 Please note that this option is only available when haproxy has been compiled
1052 with USE_WURFL=1.
1053
1054wurfl-patch-file [<file path>]
1055 A list of WURFL patch file paths. Note that patches are loaded during startup
1056 thus before the chroot.
1057
1058 Please note that this option is only available when haproxy has been compiled
1059 with USE_WURFL=1.
1060
1061wurfl-engine-mode { accuracy | performance }
1062 Sets the WURFL engine target. You can choose between 'accuracy' or
1063 'performance' targets. In performance mode, desktop web browser detection is
1064 done programmatically without referencing the WURFL data. As a result, most
1065 desktop web browsers are returned as generic_web_browser WURFL ID for
1066 performance. If either performance or accuracy are not defined, performance
1067 mode is enabled by default.
1068
1069 Please note that this option is only available when haproxy has been compiled
1070 with USE_WURFL=1.
1071
1072wurfl-cache-size <U>[,<D>]
1073 Sets the WURFL caching strategy. Here <U> is the Useragent cache size, and
1074 <D> is the internal device cache size. There are three possibilities here :
1075 - "0" : no cache is used.
1076 - <U> : the Single LRU cache is used, the size is expressed in elements.
1077 - <U>,<D> : the Double LRU cache is used, both sizes are in elements. This is
1078 the highest performing option.
1079
1080 Please note that this option is only available when haproxy has been compiled
1081 with USE_WURFL=1.
1082
1083wurfl-useragent-priority { plain | sideloaded_browser }
1084 Tells WURFL if it should prioritize use of the plain user agent ('plain')
1085 over the default sideloaded browser user agent ('sideloaded_browser').
1086
1087 Please note that this option is only available when haproxy has been compiled
1088 with USE_WURFL=1.
1089
Willy Tarreau6a06a402007-07-15 20:15:28 +02001090
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010913.2. Performance tuning
Willy Tarreau6a06a402007-07-15 20:15:28 +02001092-----------------------
1093
Willy Tarreau1746eec2014-04-25 10:46:47 +02001094max-spread-checks <delay in milliseconds>
1095 By default, haproxy tries to spread the start of health checks across the
1096 smallest health check interval of all the servers in a farm. The principle is
1097 to avoid hammering services running on the same server. But when using large
1098 check intervals (10 seconds or more), the last servers in the farm take some
1099 time before starting to be tested, which can be a problem. This parameter is
1100 used to enforce an upper bound on delay between the first and the last check,
1101 even if the servers' check intervals are larger. When servers run with
1102 shorter intervals, their intervals will be respected though.
1103
Willy Tarreau6a06a402007-07-15 20:15:28 +02001104maxconn <number>
1105 Sets the maximum per-process number of concurrent connections to <number>. It
1106 is equivalent to the command-line argument "-n". Proxies will stop accepting
1107 connections when this limit is reached. The "ulimit-n" parameter is
Willy Tarreau8274e102014-06-19 15:31:25 +02001108 automatically adjusted according to this value. See also "ulimit-n". Note:
1109 the "select" poller cannot reliably use more than 1024 file descriptors on
1110 some platforms. If your platform only supports select and reports "select
1111 FAILED" on startup, you need to reduce maxconn until it works (slightly
Willy Tarreaud0256482015-01-15 21:45:22 +01001112 below 500 in general). If this value is not set, it will default to the value
1113 set in DEFAULT_MAXCONN at build time (reported in haproxy -vv) if no memory
1114 limit is enforced, or will be computed based on the memory limit, the buffer
1115 size, memory allocated to compression, SSL cache size, and use or not of SSL
1116 and the associated maxsslconn (which can also be automatic).
Willy Tarreau6a06a402007-07-15 20:15:28 +02001117
Willy Tarreau81c25d02011-09-07 15:17:21 +02001118maxconnrate <number>
1119 Sets the maximum per-process number of connections per second to <number>.
1120 Proxies will stop accepting connections when this limit is reached. It can be
1121 used to limit the global capacity regardless of each frontend capacity. It is
1122 important to note that this can only be used as a service protection measure,
1123 as there will not necessarily be a fair share between frontends when the
1124 limit is reached, so it's a good idea to also limit each frontend to some
1125 value close to its expected share. Also, lowering tune.maxaccept can improve
1126 fairness.
1127
William Lallemandd85f9172012-11-09 17:05:39 +01001128maxcomprate <number>
1129 Sets the maximum per-process input compression rate to <number> kilobytes
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001130 per second. For each session, if the maximum is reached, the compression
William Lallemandd85f9172012-11-09 17:05:39 +01001131 level will be decreased during the session. If the maximum is reached at the
1132 beginning of a session, the session will not compress at all. If the maximum
1133 is not reached, the compression level will be increased up to
1134 tune.comp.maxlevel. A value of zero means there is no limit, this is the
1135 default value.
1136
William Lallemand072a2bf2012-11-20 17:01:01 +01001137maxcompcpuusage <number>
1138 Sets the maximum CPU usage HAProxy can reach before stopping the compression
1139 for new requests or decreasing the compression level of current requests.
1140 It works like 'maxcomprate' but measures CPU usage instead of incoming data
1141 bandwidth. The value is expressed in percent of the CPU used by haproxy. In
1142 case of multiple processes (nbproc > 1), each process manages its individual
1143 usage. A value of 100 disable the limit. The default value is 100. Setting
1144 a lower value will prevent the compression work from slowing the whole
1145 process down and from introducing high latencies.
1146
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001147maxpipes <number>
1148 Sets the maximum per-process number of pipes to <number>. Currently, pipes
1149 are only used by kernel-based tcp splicing. Since a pipe contains two file
1150 descriptors, the "ulimit-n" value will be increased accordingly. The default
1151 value is maxconn/4, which seems to be more than enough for most heavy usages.
1152 The splice code dynamically allocates and releases pipes, and can fall back
1153 to standard copy, so setting this value too low may only impact performance.
1154
Willy Tarreau93e7c002013-10-07 18:51:07 +02001155maxsessrate <number>
1156 Sets the maximum per-process number of sessions per second to <number>.
1157 Proxies will stop accepting connections when this limit is reached. It can be
1158 used to limit the global capacity regardless of each frontend capacity. It is
1159 important to note that this can only be used as a service protection measure,
1160 as there will not necessarily be a fair share between frontends when the
1161 limit is reached, so it's a good idea to also limit each frontend to some
1162 value close to its expected share. Also, lowering tune.maxaccept can improve
1163 fairness.
1164
Willy Tarreau403edff2012-09-06 11:58:37 +02001165maxsslconn <number>
1166 Sets the maximum per-process number of concurrent SSL connections to
1167 <number>. By default there is no SSL-specific limit, which means that the
1168 global maxconn setting will apply to all connections. Setting this limit
1169 avoids having openssl use too much memory and crash when malloc returns NULL
1170 (since it unfortunately does not reliably check for such conditions). Note
1171 that the limit applies both to incoming and outgoing connections, so one
1172 connection which is deciphered then ciphered accounts for 2 SSL connections.
Willy Tarreaud0256482015-01-15 21:45:22 +01001173 If this value is not set, but a memory limit is enforced, this value will be
1174 automatically computed based on the memory limit, maxconn, the buffer size,
1175 memory allocated to compression, SSL cache size, and use of SSL in either
1176 frontends, backends or both. If neither maxconn nor maxsslconn are specified
1177 when there is a memory limit, haproxy will automatically adjust these values
1178 so that 100% of the connections can be made over SSL with no risk, and will
1179 consider the sides where it is enabled (frontend, backend, both).
Willy Tarreau403edff2012-09-06 11:58:37 +02001180
Willy Tarreaue43d5322013-10-07 20:01:52 +02001181maxsslrate <number>
1182 Sets the maximum per-process number of SSL sessions per second to <number>.
1183 SSL listeners will stop accepting connections when this limit is reached. It
1184 can be used to limit the global SSL CPU usage regardless of each frontend
1185 capacity. It is important to note that this can only be used as a service
1186 protection measure, as there will not necessarily be a fair share between
1187 frontends when the limit is reached, so it's a good idea to also limit each
1188 frontend to some value close to its expected share. It is also important to
1189 note that the sessions are accounted before they enter the SSL stack and not
1190 after, which also protects the stack against bad handshakes. Also, lowering
1191 tune.maxaccept can improve fairness.
1192
William Lallemand9d5f5482012-11-07 16:12:57 +01001193maxzlibmem <number>
1194 Sets the maximum amount of RAM in megabytes per process usable by the zlib.
1195 When the maximum amount is reached, future sessions will not compress as long
1196 as RAM is unavailable. When sets to 0, there is no limit.
William Lallemande3a7d992012-11-20 11:25:20 +01001197 The default value is 0. The value is available in bytes on the UNIX socket
1198 with "show info" on the line "MaxZlibMemUsage", the memory used by zlib is
1199 "ZlibMemUsage" in bytes.
1200
Willy Tarreau6a06a402007-07-15 20:15:28 +02001201noepoll
1202 Disables the use of the "epoll" event polling system on Linux. It is
1203 equivalent to the command-line argument "-de". The next polling system
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001204 used will generally be "poll". See also "nopoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001205
1206nokqueue
1207 Disables the use of the "kqueue" event polling system on BSD. It is
1208 equivalent to the command-line argument "-dk". The next polling system
1209 used will generally be "poll". See also "nopoll".
1210
1211nopoll
1212 Disables the use of the "poll" event polling system. It is equivalent to the
1213 command-line argument "-dp". The next polling system used will be "select".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001214 It should never be needed to disable "poll" since it's available on all
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001215 platforms supported by HAProxy. See also "nokqueue" and "noepoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001216
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001217nosplice
1218 Disables the use of kernel tcp splicing between sockets on Linux. It is
1219 equivalent to the command line argument "-dS". Data will then be copied
1220 using conventional and more portable recv/send calls. Kernel tcp splicing is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001221 limited to some very recent instances of kernel 2.6. Most versions between
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001222 2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not
1223 be used. This option makes it easier to globally disable kernel splicing in
1224 case of doubt. See also "option splice-auto", "option splice-request" and
1225 "option splice-response".
1226
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001227nogetaddrinfo
1228 Disables the use of getaddrinfo(3) for name resolving. It is equivalent to
1229 the command line argument "-dG". Deprecated gethostbyname(3) will be used.
1230
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +00001231noreuseport
1232 Disables the use of SO_REUSEPORT - see socket(7). It is equivalent to the
1233 command line argument "-dR".
1234
Willy Tarreaufe255b72007-10-14 23:09:26 +02001235spread-checks <0..50, in percent>
Simon Hormand60d6912013-11-25 10:46:36 +09001236 Sometimes it is desirable to avoid sending agent and health checks to
1237 servers at exact intervals, for instance when many logical servers are
1238 located on the same physical server. With the help of this parameter, it
1239 becomes possible to add some randomness in the check interval between 0
1240 and +/- 50%. A value between 2 and 5 seems to show good results. The
1241 default value remains at 0.
Willy Tarreaufe255b72007-10-14 23:09:26 +02001242
Willy Tarreau33cb0652014-12-23 22:52:37 +01001243tune.buffers.limit <number>
1244 Sets a hard limit on the number of buffers which may be allocated per process.
1245 The default value is zero which means unlimited. The minimum non-zero value
1246 will always be greater than "tune.buffers.reserve" and should ideally always
1247 be about twice as large. Forcing this value can be particularly useful to
1248 limit the amount of memory a process may take, while retaining a sane
1249 behaviour. When this limit is reached, sessions which need a buffer wait for
1250 another one to be released by another session. Since buffers are dynamically
1251 allocated and released, the waiting time is very short and not perceptible
1252 provided that limits remain reasonable. In fact sometimes reducing the limit
1253 may even increase performance by increasing the CPU cache's efficiency. Tests
1254 have shown good results on average HTTP traffic with a limit to 1/10 of the
1255 expected global maxconn setting, which also significantly reduces memory
1256 usage. The memory savings come from the fact that a number of connections
1257 will not allocate 2*tune.bufsize. It is best not to touch this value unless
1258 advised to do so by an haproxy core developer.
1259
Willy Tarreau1058ae72014-12-23 22:40:40 +01001260tune.buffers.reserve <number>
1261 Sets the number of buffers which are pre-allocated and reserved for use only
1262 during memory shortage conditions resulting in failed memory allocations. The
1263 minimum value is 2 and is also the default. There is no reason a user would
1264 want to change this value, it's mostly aimed at haproxy core developers.
1265
Willy Tarreau27a674e2009-08-17 07:23:33 +02001266tune.bufsize <number>
1267 Sets the buffer size to this size (in bytes). Lower values allow more
1268 sessions to coexist in the same amount of RAM, and higher values allow some
1269 applications with very large cookies to work. The default value is 16384 and
1270 can be changed at build time. It is strongly recommended not to change this
1271 from the default value, as very low values will break some services such as
1272 statistics, and values larger than default size will increase memory usage,
1273 possibly causing the system to run out of memory. At least the global maxconn
1274 parameter should be decreased by the same factor as this one is increased.
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04001275 If HTTP request is larger than (tune.bufsize - tune.maxrewrite), haproxy will
1276 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
1277 than this size, haproxy will return HTTP 502 (Bad Gateway).
Willy Tarreau27a674e2009-08-17 07:23:33 +02001278
Willy Tarreau43961d52010-10-04 20:39:20 +02001279tune.chksize <number>
1280 Sets the check buffer size to this size (in bytes). Higher values may help
1281 find string or regex patterns in very large pages, though doing so may imply
1282 more memory and CPU usage. The default value is 16384 and can be changed at
1283 build time. It is not recommended to change this value, but to use better
1284 checks whenever possible.
1285
William Lallemandf3747832012-11-09 12:33:10 +01001286tune.comp.maxlevel <number>
1287 Sets the maximum compression level. The compression level affects CPU
1288 usage during compression. This value affects CPU usage during compression.
1289 Each session using compression initializes the compression algorithm with
1290 this value. The default value is 1.
1291
Willy Tarreau193b8c62012-11-22 00:17:38 +01001292tune.http.cookielen <number>
1293 Sets the maximum length of captured cookies. This is the maximum value that
1294 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
1295 will automatically be truncated to this one. It is important not to set too
1296 high a value because all cookie captures still allocate this size whatever
1297 their configured value (they share a same pool). This value is per request
1298 per response, so the memory allocated is twice this value per connection.
1299 When not specified, the limit is set to 63 characters. It is recommended not
1300 to change this value.
1301
Willy Tarreauac1932d2011-10-24 19:14:41 +02001302tune.http.maxhdr <number>
1303 Sets the maximum number of headers in a request. When a request comes with a
1304 number of headers greater than this value (including the first line), it is
1305 rejected with a "400 Bad Request" status code. Similarly, too large responses
1306 are blocked with "502 Bad Gateway". The default value is 101, which is enough
1307 for all usages, considering that the widely deployed Apache server uses the
1308 same limit. It can be useful to push this limit further to temporarily allow
1309 a buggy application to work by the time it gets fixed. Keep in mind that each
1310 new header consumes 32bits of memory for each session, so don't push this
1311 limit too high.
1312
Willy Tarreau7e312732014-02-12 16:35:14 +01001313tune.idletimer <timeout>
1314 Sets the duration after which haproxy will consider that an empty buffer is
1315 probably associated with an idle stream. This is used to optimally adjust
1316 some packet sizes while forwarding large and small data alternatively. The
1317 decision to use splice() or to send large buffers in SSL is modulated by this
1318 parameter. The value is in milliseconds between 0 and 65535. A value of zero
1319 means that haproxy will not try to detect idle streams. The default is 1000,
1320 which seems to correctly detect end user pauses (eg: read a page before
1321 clicking). There should be not reason for changing this value. Please check
1322 tune.ssl.maxrecord below.
1323
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001324tune.lua.forced-yield <number>
1325 This directive forces the Lua engine to execute a yield each <number> of
1326 instructions executed. This permits interruptng a long script and allows the
1327 HAProxy scheduler to process other tasks like accepting connections or
1328 forwarding traffic. The default value is 10000 instructions. If HAProxy often
1329 executes some Lua code but more reactivity is required, this value can be
1330 lowered. If the Lua code is quite long and its result is absolutely required
1331 to process the data, the <number> can be increased.
1332
Willy Tarreau32f61e22015-03-18 17:54:59 +01001333tune.lua.maxmem
1334 Sets the maximum amount of RAM in megabytes per process usable by Lua. By
1335 default it is zero which means unlimited. It is important to set a limit to
1336 ensure that a bug in a script will not result in the system running out of
1337 memory.
1338
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001339tune.lua.session-timeout <timeout>
1340 This is the execution timeout for the Lua sessions. This is useful for
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001341 preventing infinite loops or spending too much time in Lua. This timeout
1342 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
1343 not taked in account. The default timeout is 4s.
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001344
1345tune.lua.task-timeout <timeout>
1346 Purpose is the same as "tune.lua.session-timeout", but this timeout is
1347 dedicated to the tasks. By default, this timeout isn't set because a task may
1348 remain alive during of the lifetime of HAProxy. For example, a task used to
1349 check servers.
1350
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001351tune.lua.service-timeout <timeout>
1352 This is the execution timeout for the Lua services. This is useful for
1353 preventing infinite loops or spending too much time in Lua. This timeout
1354 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
1355 not taked in account. The default timeout is 4s.
1356
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001357tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +01001358 Sets the maximum number of consecutive connections a process may accept in a
1359 row before switching to other work. In single process mode, higher numbers
1360 give better performance at high connection rates. However in multi-process
1361 modes, keeping a bit of fairness between processes generally is better to
1362 increase performance. This value applies individually to each listener, so
1363 that the number of processes a listener is bound to is taken into account.
1364 This value defaults to 64. In multi-process mode, it is divided by twice
1365 the number of processes the listener is bound to. Setting this value to -1
1366 completely disables the limitation. It should normally not be needed to tweak
1367 this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001368
1369tune.maxpollevents <number>
1370 Sets the maximum amount of events that can be processed at once in a call to
1371 the polling system. The default value is adapted to the operating system. It
1372 has been noticed that reducing it below 200 tends to slightly decrease
1373 latency at the expense of network bandwidth, and increasing it above 200
1374 tends to trade latency for slightly increased bandwidth.
1375
Willy Tarreau27a674e2009-08-17 07:23:33 +02001376tune.maxrewrite <number>
1377 Sets the reserved buffer space to this size in bytes. The reserved space is
1378 used for header rewriting or appending. The first reads on sockets will never
1379 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
1380 bufsize, though that does not make much sense since there are rarely large
1381 numbers of headers to add. Setting it too high prevents processing of large
1382 requests or responses. Setting it too low prevents addition of new headers
1383 to already large requests or to POST requests. It is generally wise to set it
1384 to about 1024. It is automatically readjusted to half of bufsize if it is
1385 larger than that. This means you don't have to worry about it when changing
1386 bufsize.
1387
Willy Tarreauf3045d22015-04-29 16:24:50 +02001388tune.pattern.cache-size <number>
1389 Sets the size of the pattern lookup cache to <number> entries. This is an LRU
1390 cache which reminds previous lookups and their results. It is used by ACLs
1391 and maps on slow pattern lookups, namely the ones using the "sub", "reg",
1392 "dir", "dom", "end", "bin" match methods as well as the case-insensitive
1393 strings. It applies to pattern expressions which means that it will be able
1394 to memorize the result of a lookup among all the patterns specified on a
1395 configuration line (including all those loaded from files). It automatically
1396 invalidates entries which are updated using HTTP actions or on the CLI. The
1397 default cache size is set to 10000 entries, which limits its footprint to
1398 about 5 MB on 32-bit systems and 8 MB on 64-bit systems. There is a very low
1399 risk of collision in this cache, which is in the order of the size of the
1400 cache divided by 2^64. Typically, at 10000 requests per second with the
1401 default cache size of 10000 entries, there's 1% chance that a brute force
1402 attack could cause a single collision after 60 years, or 0.1% after 6 years.
1403 This is considered much lower than the risk of a memory corruption caused by
1404 aging components. If this is not acceptable, the cache can be disabled by
1405 setting this parameter to 0.
1406
Willy Tarreaubd9a0a72011-10-23 21:14:29 +02001407tune.pipesize <number>
1408 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
1409 are the default size for the system. But sometimes when using TCP splicing,
1410 it can improve performance to increase pipe sizes, especially if it is
1411 suspected that pipes are not filled and that many calls to splice() are
1412 performed. This has an impact on the kernel's memory footprint, so this must
1413 not be changed if impacts are not understood.
1414
Willy Tarreaue803de22010-01-21 17:43:04 +01001415tune.rcvbuf.client <number>
1416tune.rcvbuf.server <number>
1417 Forces the kernel socket receive buffer size on the client or the server side
1418 to the specified value in bytes. This value applies to all TCP/HTTP frontends
1419 and backends. It should normally never be set, and the default size (0) lets
1420 the kernel autotune this value depending on the amount of available memory.
1421 However it can sometimes help to set it to very low values (eg: 4096) in
1422 order to save kernel memory by preventing it from buffering too large amounts
1423 of received data. Lower values will significantly increase CPU usage though.
1424
Willy Tarreaub22fc302015-12-14 12:04:35 +01001425tune.recv_enough <number>
1426 Haproxy uses some hints to detect that a short read indicates the end of the
1427 socket buffers. One of them is that a read returns more than <recv_enough>
1428 bytes, which defaults to 10136 (7 segments of 1448 each). This default value
1429 may be changed by this setting to better deal with workloads involving lots
1430 of short messages such as telnet or SSH sessions.
1431
Willy Tarreaue803de22010-01-21 17:43:04 +01001432tune.sndbuf.client <number>
1433tune.sndbuf.server <number>
1434 Forces the kernel socket send buffer size on the client or the server side to
1435 the specified value in bytes. This value applies to all TCP/HTTP frontends
1436 and backends. It should normally never be set, and the default size (0) lets
1437 the kernel autotune this value depending on the amount of available memory.
1438 However it can sometimes help to set it to very low values (eg: 4096) in
1439 order to save kernel memory by preventing it from buffering too large amounts
1440 of received data. Lower values will significantly increase CPU usage though.
1441 Another use case is to prevent write timeouts with extremely slow clients due
1442 to the kernel waiting for a large part of the buffer to be read before
1443 notifying haproxy again.
1444
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001445tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +01001446 Sets the size of the global SSL session cache, in a number of blocks. A block
1447 is large enough to contain an encoded session without peer certificate.
1448 An encoded session with peer certificate is stored in multiple blocks
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001449 depending on the size of the peer certificate. A block uses approximately
Emeric Brunaf9619d2012-11-28 18:47:52 +01001450 200 bytes of memory. The default value may be forced at build time, otherwise
1451 defaults to 20000. When the cache is full, the most idle entries are purged
1452 and reassigned. Higher values reduce the occurrence of such a purge, hence
1453 the number of CPU-intensive SSL handshakes by ensuring that all users keep
1454 their session as long as possible. All entries are pre-allocated upon startup
Emeric Brun22890a12012-12-28 14:41:32 +01001455 and are shared between all processes if "nbproc" is greater than 1. Setting
1456 this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001457
Emeric Brun8dc60392014-05-09 13:52:00 +02001458tune.ssl.force-private-cache
1459 This boolean disables SSL session cache sharing between all processes. It
1460 should normally not be used since it will force many renegotiations due to
1461 clients hitting a random process. But it may be required on some operating
1462 systems where none of the SSL cache synchronization method may be used. In
1463 this case, adding a first layer of hash-based load balancing before the SSL
1464 layer might limit the impact of the lack of session sharing.
1465
Emeric Brun4f65bff2012-11-16 15:11:00 +01001466tune.ssl.lifetime <timeout>
1467 Sets how long a cached SSL session may remain valid. This time is expressed
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001468 in seconds and defaults to 300 (5 min). It is important to understand that it
Emeric Brun4f65bff2012-11-16 15:11:00 +01001469 does not guarantee that sessions will last that long, because if the cache is
1470 full, the longest idle sessions will be purged despite their configured
1471 lifetime. The real usefulness of this setting is to prevent sessions from
1472 being used for too long.
1473
Willy Tarreaubfd59462013-02-21 07:46:09 +01001474tune.ssl.maxrecord <number>
1475 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
1476 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
1477 data only once it has received a full record. With large records, it means
1478 that clients might have to download up to 16kB of data before starting to
1479 process them. Limiting the value can improve page load times on browsers
1480 located over high latency or low bandwidth networks. It is suggested to find
1481 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
1482 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
1483 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
1484 2859 gave good results during tests. Use "strace -e trace=write" to find the
Willy Tarreau7e312732014-02-12 16:35:14 +01001485 best value. Haproxy will automatically switch to this setting after an idle
1486 stream has been detected (see tune.idletimer above).
Willy Tarreaubfd59462013-02-21 07:46:09 +01001487
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001488tune.ssl.default-dh-param <number>
1489 Sets the maximum size of the Diffie-Hellman parameters used for generating
1490 the ephemeral/temporary Diffie-Hellman key in case of DHE key exchange. The
1491 final size will try to match the size of the server's RSA (or DSA) key (e.g,
1492 a 2048 bits temporary DH key for a 2048 bits RSA key), but will not exceed
1493 this maximum value. Default value if 1024. Only 1024 or higher values are
1494 allowed. Higher values will increase the CPU load, and values greater than
1495 1024 bits are not supported by Java 7 and earlier clients. This value is not
Remi Gacogne47783ef2015-05-29 15:53:22 +02001496 used if static Diffie-Hellman parameters are supplied either directly
1497 in the certificate file or by using the ssl-dh-param-file parameter.
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001498
Christopher Faulet31af49d2015-06-09 17:29:50 +02001499tune.ssl.ssl-ctx-cache-size <number>
1500 Sets the size of the cache used to store generated certificates to <number>
1501 entries. This is a LRU cache. Because generating a SSL certificate
1502 dynamically is expensive, they are cached. The default cache size is set to
1503 1000 entries.
1504
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001505tune.vars.global-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01001506tune.vars.proc-max-size <size>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001507tune.vars.reqres-max-size <size>
1508tune.vars.sess-max-size <size>
1509tune.vars.txn-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01001510 These five tunes help to manage the maximum amount of memory used by the
1511 variables system. "global" limits the overall amount of memory available for
1512 all scopes. "proc" limits the memory for the process scope, "sess" limits the
1513 memory for the session scope, "txn" for the transaction scope, and "reqres"
1514 limits the memory for each request or response processing.
1515 Memory accounting is hierarchical, meaning more coarse grained limits include
1516 the finer grained ones: "proc" includes "sess", "sess" includes "txn", and
1517 "txn" includes "reqres".
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001518
Daniel Schneller0b547052016-03-21 20:46:57 +01001519 For example, when "tune.vars.sess-max-size" is limited to 100,
1520 "tune.vars.txn-max-size" and "tune.vars.reqres-max-size" cannot exceed
1521 100 either. If we create a variable "txn.var" that contains 100 bytes,
1522 all available space is consumed.
1523 Notice that exceeding the limits at runtime will not result in an error
1524 message, but values might be cut off or corrupted. So make sure to accurately
1525 plan for the amount of space needed to store all your variables.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001526
William Lallemanda509e4c2012-11-07 16:54:34 +01001527tune.zlib.memlevel <number>
1528 Sets the memLevel parameter in zlib initialization for each session. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001529 defines how much memory should be allocated for the internal compression
William Lallemanda509e4c2012-11-07 16:54:34 +01001530 state. A value of 1 uses minimum memory but is slow and reduces compression
1531 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
1532 between 1 and 9. The default value is 8.
1533
1534tune.zlib.windowsize <number>
1535 Sets the window size (the size of the history buffer) as a parameter of the
1536 zlib initialization for each session. Larger values of this parameter result
1537 in better compression at the expense of memory usage. Can be a value between
1538 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001539
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015403.3. Debugging
1541--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02001542
1543debug
1544 Enables debug mode which dumps to stdout all exchanges, and disables forking
1545 into background. It is the equivalent of the command-line argument "-d". It
1546 should never be used in a production configuration since it may prevent full
1547 system startup.
1548
1549quiet
1550 Do not display any message during startup. It is equivalent to the command-
1551 line argument "-q".
1552
Emeric Brunf099e792010-09-27 12:05:28 +02001553
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010015543.4. Userlists
1555--------------
1556It is possible to control access to frontend/backend/listen sections or to
1557http stats by allowing only authenticated and authorized users. To do this,
1558it is required to create at least one userlist and to define users.
1559
1560userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +01001561 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001562 used to store authentication & authorization data for independent customers.
1563
1564group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +01001565 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001566 attach users to this group by using a comma separated list of names
1567 proceeded by "users" keyword.
1568
Cyril Bontéf0c60612010-02-06 14:44:47 +01001569user <username> [password|insecure-password <password>]
1570 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001571 Adds user <username> to the current userlist. Both secure (encrypted) and
1572 insecure (unencrypted) passwords can be used. Encrypted passwords are
Cyril Bonté78caf842010-03-10 22:41:43 +01001573 evaluated using the crypt(3) function so depending of the system's
1574 capabilities, different algorithms are supported. For example modern Glibc
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001575 based Linux system supports MD5, SHA-256, SHA-512 and of course classic,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001576 DES-based method of encrypting passwords.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001577
1578
1579 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +01001580 userlist L1
1581 group G1 users tiger,scott
1582 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001583
Cyril Bontéf0c60612010-02-06 14:44:47 +01001584 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
1585 user scott insecure-password elgato
1586 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001587
Cyril Bontéf0c60612010-02-06 14:44:47 +01001588 userlist L2
1589 group G1
1590 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001591
Cyril Bontéf0c60612010-02-06 14:44:47 +01001592 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
1593 user scott insecure-password elgato groups G1,G2
1594 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001595
1596 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001597
Emeric Brunf099e792010-09-27 12:05:28 +02001598
15993.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02001600----------
Emeric Brun94900952015-06-11 18:25:54 +02001601It is possible to propagate entries of any data-types in stick-tables between
1602several haproxy instances over TCP connections in a multi-master fashion. Each
1603instance pushes its local updates and insertions to remote peers. The pushed
1604values overwrite remote ones without aggregation. Interrupted exchanges are
1605automatically detected and recovered from the last known point.
1606In addition, during a soft restart, the old process connects to the new one
1607using such a TCP connection to push all its entries before the new process
1608tries to connect to other peers. That ensures very fast replication during a
1609reload, it typically takes a fraction of a second even for large tables.
1610Note that Server IDs are used to identify servers remotely, so it is important
1611that configurations look similar or at least that the same IDs are forced on
1612each server on all participants.
Emeric Brunf099e792010-09-27 12:05:28 +02001613
1614peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001615 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02001616 which is referenced by one or more stick-tables.
1617
Willy Tarreau77e4bd12015-05-01 20:02:17 +02001618disabled
1619 Disables a peers section. It disables both listening and any synchronization
1620 related to this section. This is provided to disable synchronization of stick
1621 tables without having to comment out all "peers" references.
1622
1623enable
1624 This re-enables a disabled peers section which was previously disabled.
1625
Emeric Brunf099e792010-09-27 12:05:28 +02001626peer <peername> <ip>:<port>
1627 Defines a peer inside a peers section.
1628 If <peername> is set to the local peer name (by default hostname, or forced
1629 using "-L" command line option), haproxy will listen for incoming remote peer
1630 connection on <ip>:<port>. Otherwise, <ip>:<port> defines where to connect to
1631 to join the remote peer, and <peername> is used at the protocol level to
1632 identify and validate the remote peer on the server side.
1633
1634 During a soft restart, local peer <ip>:<port> is used by the old instance to
1635 connect the new one and initiate a complete replication (teaching process).
1636
1637 It is strongly recommended to have the exact same peers declaration on all
1638 peers and to only rely on the "-L" command line argument to change the local
1639 peer name. This makes it easier to maintain coherent configuration files
1640 across all peers.
1641
William Lallemandb2f07452015-05-12 14:27:13 +02001642 You may want to reference some environment variables in the address
1643 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01001644
Cyril Bontédc4d9032012-04-08 21:57:39 +02001645 Example:
Emeric Brunf099e792010-09-27 12:05:28 +02001646 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001647 peer haproxy1 192.168.0.1:1024
1648 peer haproxy2 192.168.0.2:1024
1649 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02001650
1651 backend mybackend
1652 mode tcp
1653 balance roundrobin
1654 stick-table type ip size 20k peers mypeers
1655 stick on src
1656
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001657 server srv1 192.168.0.30:80
1658 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02001659
1660
Simon Horman51a1cf62015-02-03 13:00:44 +090016613.6. Mailers
1662------------
1663It is possible to send email alerts when the state of servers changes.
1664If configured email alerts are sent to each mailer that is configured
1665in a mailers section. Email is sent to mailers using SMTP.
1666
Pieter Baauw386a1272015-08-16 15:26:24 +02001667mailers <mailersect>
Simon Horman51a1cf62015-02-03 13:00:44 +09001668 Creates a new mailer list with the name <mailersect>. It is an
1669 independent section which is referenced by one or more proxies.
1670
1671mailer <mailername> <ip>:<port>
1672 Defines a mailer inside a mailers section.
1673
1674 Example:
1675 mailers mymailers
1676 mailer smtp1 192.168.0.1:587
1677 mailer smtp2 192.168.0.2:587
1678
1679 backend mybackend
1680 mode tcp
1681 balance roundrobin
1682
1683 email-alert mailers mymailers
1684 email-alert from test1@horms.org
1685 email-alert to test2@horms.org
1686
1687 server srv1 192.168.0.30:80
1688 server srv2 192.168.0.31:80
1689
Pieter Baauw235fcfc2016-02-13 15:33:40 +01001690timeout mail <time>
1691 Defines the time available for a mail/connection to be made and send to
1692 the mail-server. If not defined the default value is 10 seconds. To allow
1693 for at least two SYN-ACK packets to be send during initial TCP handshake it
1694 is advised to keep this value above 4 seconds.
1695
1696 Example:
1697 mailers mymailers
1698 timeout mail 20s
1699 mailer smtp1 192.168.0.1:587
Simon Horman51a1cf62015-02-03 13:00:44 +09001700
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017014. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02001702----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001703
Willy Tarreau6a06a402007-07-15 20:15:28 +02001704Proxy configuration can be located in a set of sections :
William Lallemand6e62fb62015-04-28 16:55:23 +02001705 - defaults [<name>]
Willy Tarreau6a06a402007-07-15 20:15:28 +02001706 - frontend <name>
1707 - backend <name>
1708 - listen <name>
1709
1710A "defaults" section sets default parameters for all other sections following
1711its declaration. Those default parameters are reset by the next "defaults"
1712section. See below for the list of parameters which can be set in a "defaults"
Willy Tarreau0ba27502007-12-24 16:55:16 +01001713section. The name is optional but its use is encouraged for better readability.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001714
1715A "frontend" section describes a set of listening sockets accepting client
1716connections.
1717
1718A "backend" section describes a set of servers to which the proxy will connect
1719to forward incoming connections.
1720
1721A "listen" section defines a complete proxy with its frontend and backend
1722parts combined in one section. It is generally useful for TCP-only traffic.
1723
Willy Tarreau0ba27502007-12-24 16:55:16 +01001724All proxy names must be formed from upper and lower case letters, digits,
1725'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
1726case-sensitive, which means that "www" and "WWW" are two different proxies.
1727
1728Historically, all proxy names could overlap, it just caused troubles in the
1729logs. Since the introduction of content switching, it is mandatory that two
1730proxies with overlapping capabilities (frontend/backend) have different names.
1731However, it is still permitted that a frontend and a backend share the same
1732name, as this configuration seems to be commonly encountered.
1733
1734Right now, two major proxy modes are supported : "tcp", also known as layer 4,
1735and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001736bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01001737protocol, and can interact with it by allowing, blocking, switching, adding,
1738modifying, or removing arbitrary contents in requests or responses, based on
1739arbitrary criteria.
1740
Willy Tarreau70dffda2014-01-30 03:07:23 +01001741In HTTP mode, the processing applied to requests and responses flowing over
1742a connection depends in the combination of the frontend's HTTP options and
1743the backend's. HAProxy supports 5 connection modes :
1744
1745 - KAL : keep alive ("option http-keep-alive") which is the default mode : all
1746 requests and responses are processed, and connections remain open but idle
1747 between responses and new requests.
1748
1749 - TUN: tunnel ("option http-tunnel") : this was the default mode for versions
1750 1.0 to 1.5-dev21 : only the first request and response are processed, and
1751 everything else is forwarded with no analysis at all. This mode should not
1752 be used as it creates lots of trouble with logging and HTTP processing.
1753
1754 - PCL: passive close ("option httpclose") : exactly the same as tunnel mode,
1755 but with "Connection: close" appended in both directions to try to make
1756 both ends close after the first request/response exchange.
1757
1758 - SCL: server close ("option http-server-close") : the server-facing
1759 connection is closed after the end of the response is received, but the
1760 client-facing connection remains open.
1761
1762 - FCL: forced close ("option forceclose") : the connection is actively closed
1763 after the end of the response.
1764
1765The effective mode that will be applied to a connection passing through a
1766frontend and a backend can be determined by both proxy modes according to the
1767following matrix, but in short, the modes are symmetric, keep-alive is the
1768weakest option and force close is the strongest.
1769
1770 Backend mode
1771
1772 | KAL | TUN | PCL | SCL | FCL
1773 ----+-----+-----+-----+-----+----
1774 KAL | KAL | TUN | PCL | SCL | FCL
1775 ----+-----+-----+-----+-----+----
1776 TUN | TUN | TUN | PCL | SCL | FCL
1777 Frontend ----+-----+-----+-----+-----+----
1778 mode PCL | PCL | PCL | PCL | FCL | FCL
1779 ----+-----+-----+-----+-----+----
1780 SCL | SCL | SCL | FCL | SCL | FCL
1781 ----+-----+-----+-----+-----+----
1782 FCL | FCL | FCL | FCL | FCL | FCL
1783
Willy Tarreau0ba27502007-12-24 16:55:16 +01001784
Willy Tarreau70dffda2014-01-30 03:07:23 +01001785
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017864.1. Proxy keywords matrix
1787--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001788
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001789The following list of keywords is supported. Most of them may only be used in a
1790limited set of section types. Some of them are marked as "deprecated" because
1791they are inherited from an old syntax which may be confusing or functionally
1792limited, and there are new recommended keywords to replace them. Keywords
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001793marked with "(*)" can be optionally inverted using the "no" prefix, eg. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001794option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02001795and must be disabled for a specific instance. Such options may also be prefixed
1796with "default" in order to restore default settings regardless of what has been
1797specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001798
Willy Tarreau6a06a402007-07-15 20:15:28 +02001799
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001800 keyword defaults frontend listen backend
1801------------------------------------+----------+----------+---------+---------
1802acl - X X X
Willy Tarreau294d0f02015-08-10 19:40:12 +02001803appsession - - - -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001804backlog X X X -
1805balance X - X X
1806bind - X X -
1807bind-process X X X X
1808block - X X X
1809capture cookie - X X -
1810capture request header - X X -
1811capture response header - X X -
1812clitimeout (deprecated) X X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02001813compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001814contimeout (deprecated) X - X X
1815cookie X - X X
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02001816declare capture - X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001817default-server X - X X
1818default_backend X X X -
1819description - X X X
1820disabled X X X X
1821dispatch - - X X
Simon Horman51a1cf62015-02-03 13:00:44 +09001822email-alert from X X X X
Simon Horman64e34162015-02-06 11:11:57 +09001823email-alert level X X X X
Simon Horman51a1cf62015-02-03 13:00:44 +09001824email-alert mailers X X X X
1825email-alert myhostname X X X X
1826email-alert to X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001827enabled X X X X
1828errorfile X X X X
1829errorloc X X X X
1830errorloc302 X X X X
1831-- keyword -------------------------- defaults - frontend - listen -- backend -
1832errorloc303 X X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001833force-persist - X X X
Christopher Fauletc3fe5332016-04-07 15:30:10 +02001834filter - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001835fullconn X - X X
1836grace X X X X
1837hash-type X - X X
1838http-check disable-on-404 X - X X
Willy Tarreaubd741542010-03-16 18:46:54 +01001839http-check expect - - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02001840http-check send-state X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001841http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02001842http-response - X X X
Willy Tarreau30631952015-08-06 15:05:24 +02001843http-reuse X - X X
Baptiste Assmann2c42ef52013-10-09 21:57:02 +02001844http-send-name-header - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001845id - X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001846ignore-persist - X X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001847load-server-state-from-file X - X X
William Lallemand0f99e342011-10-12 17:50:54 +02001848log (*) X X X X
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01001849log-format X X X -
Dragan Dosen7ad31542015-09-28 17:16:47 +02001850log-format-sd X X X -
Willy Tarreau094af4e2015-01-07 15:03:42 +01001851log-tag X X X X
Willy Tarreauc35362a2014-04-25 13:58:37 +02001852max-keep-alive-queue X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001853maxconn X X X -
1854mode X X X X
1855monitor fail - X X -
1856monitor-net X X X -
1857monitor-uri X X X -
1858option abortonclose (*) X - X X
1859option accept-invalid-http-request (*) X X X -
1860option accept-invalid-http-response (*) X - X X
1861option allbackups (*) X - X X
1862option checkcache (*) X - X X
1863option clitcpka (*) X X X -
1864option contstats (*) X X X -
1865option dontlog-normal (*) X X X -
1866option dontlognull (*) X X X -
1867option forceclose (*) X X X X
1868-- keyword -------------------------- defaults - frontend - listen -- backend -
1869option forwardfor X X X X
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02001870option http-buffer-request (*) X X X X
Willy Tarreau82649f92015-05-01 22:40:51 +02001871option http-ignore-probes (*) X X X -
Willy Tarreau16bfb022010-01-16 19:48:41 +01001872option http-keep-alive (*) X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02001873option http-no-delay (*) X X X X
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02001874option http-pretend-keepalive (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001875option http-server-close (*) X X X X
Willy Tarreau02bce8b2014-01-30 00:15:28 +01001876option http-tunnel (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001877option http-use-proxy-header (*) X X X -
1878option httpchk X - X X
1879option httpclose (*) X X X X
1880option httplog X X X X
1881option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001882option independent-streams (*) X X X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02001883option ldap-check X - X X
Simon Horman98637e52014-06-20 12:30:16 +09001884option external-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001885option log-health-checks (*) X - X X
1886option log-separate-errors (*) X X X -
1887option logasap (*) X X X -
1888option mysql-check X - X X
1889option nolinger (*) X X X X
1890option originalto X X X X
1891option persist (*) X - X X
Baptiste Assmann809e22a2015-10-12 20:22:55 +02001892option pgsql-check X - X X
1893option prefer-last-server (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001894option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02001895option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001896option smtpchk X - X X
1897option socket-stats (*) X X X -
1898option splice-auto (*) X X X X
1899option splice-request (*) X X X X
1900option splice-response (*) X X X X
Christopher Fauletba7bc162016-11-07 21:07:38 +01001901option spop-check - - - X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001902option srvtcpka (*) X - X X
1903option ssl-hello-chk X - X X
1904-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreaued179852013-12-16 01:07:00 +01001905option tcp-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001906option tcp-smart-accept (*) X X X -
1907option tcp-smart-connect (*) X - X X
1908option tcpka X X X X
1909option tcplog X X X X
1910option transparent (*) X - X X
Simon Horman98637e52014-06-20 12:30:16 +09001911external-check command X - X X
1912external-check path X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001913persist rdp-cookie X - X X
1914rate-limit sessions X X X -
1915redirect - X X X
1916redisp (deprecated) X - X X
1917redispatch (deprecated) X - X X
1918reqadd - X X X
1919reqallow - X X X
1920reqdel - X X X
1921reqdeny - X X X
1922reqiallow - X X X
1923reqidel - X X X
1924reqideny - X X X
1925reqipass - X X X
1926reqirep - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001927reqitarpit - X X X
1928reqpass - X X X
1929reqrep - X X X
1930-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001931reqtarpit - X X X
1932retries X - X X
1933rspadd - X X X
1934rspdel - X X X
1935rspdeny - X X X
1936rspidel - X X X
1937rspideny - X X X
1938rspirep - X X X
1939rsprep - X X X
1940server - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001941server-state-file-name X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001942source X - X X
1943srvtimeout (deprecated) X - X X
Baptiste Assmann5a549212015-10-12 20:30:24 +02001944stats admin - X X X
1945stats auth X X X X
1946stats enable X X X X
1947stats hide-version X X X X
1948stats http-request - X X X
1949stats realm X X X X
1950stats refresh X X X X
1951stats scope X X X X
1952stats show-desc X X X X
1953stats show-legends X X X X
1954stats show-node X X X X
1955stats uri X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001956-- keyword -------------------------- defaults - frontend - listen -- backend -
1957stick match - - X X
1958stick on - - X X
1959stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02001960stick store-response - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001961stick-table - - X X
Willy Tarreau938c7fe2014-04-25 14:21:39 +02001962tcp-check connect - - X X
1963tcp-check expect - - X X
1964tcp-check send - - X X
1965tcp-check send-binary - - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02001966tcp-request connection - X X -
1967tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02001968tcp-request inspect-delay - X X X
Willy Tarreau4f614292016-10-21 17:49:36 +02001969tcp-request session - X X -
Emeric Brun0a3b67f2010-09-24 15:34:53 +02001970tcp-response content - - X X
1971tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001972timeout check X - X X
1973timeout client X X X -
Willy Tarreau05cdd962014-05-10 14:30:07 +02001974timeout client-fin X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001975timeout clitimeout (deprecated) X X X -
1976timeout connect X - X X
1977timeout contimeout (deprecated) X - X X
1978timeout http-keep-alive X X X X
1979timeout http-request X X X X
1980timeout queue X - X X
1981timeout server X - X X
Willy Tarreau05cdd962014-05-10 14:30:07 +02001982timeout server-fin X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001983timeout srvtimeout (deprecated) X - X X
1984timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02001985timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001986transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01001987unique-id-format X X X -
1988unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001989use_backend - X X -
Willy Tarreau4a5cade2012-04-05 21:09:48 +02001990use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001991------------------------------------+----------+----------+---------+---------
1992 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02001993
Willy Tarreau0ba27502007-12-24 16:55:16 +01001994
Willy Tarreauc57f0e22009-05-10 13:12:33 +020019954.2. Alphabetically sorted keywords reference
1996---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001997
1998This section provides a description of each keyword and its usage.
1999
2000
2001acl <aclname> <criterion> [flags] [operator] <value> ...
2002 Declare or complete an access list.
2003 May be used in sections : defaults | frontend | listen | backend
2004 no | yes | yes | yes
2005 Example:
2006 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2007 acl invalid_src src_port 0:1023
2008 acl local_dst hdr(host) -i localhost
2009
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002010 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002011
2012
Cyril Bontéb21570a2009-11-29 20:04:48 +01002013appsession <cookie> len <length> timeout <holdtime>
2014 [request-learn] [prefix] [mode <path-parameters|query-string>]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002015 Define session stickiness on an existing application cookie.
2016 May be used in sections : defaults | frontend | listen | backend
2017 no | no | yes | yes
2018 Arguments :
2019 <cookie> this is the name of the cookie used by the application and which
2020 HAProxy will have to learn for each new session.
2021
Cyril Bontéb21570a2009-11-29 20:04:48 +01002022 <length> this is the max number of characters that will be memorized and
Willy Tarreau0ba27502007-12-24 16:55:16 +01002023 checked in each cookie value.
2024
2025 <holdtime> this is the time after which the cookie will be removed from
2026 memory if unused. If no unit is specified, this time is in
2027 milliseconds.
2028
Cyril Bontébf47aeb2009-10-15 00:15:40 +02002029 request-learn
2030 If this option is specified, then haproxy will be able to learn
2031 the cookie found in the request in case the server does not
2032 specify any in response. This is typically what happens with
2033 PHPSESSID cookies, or when haproxy's session expires before
2034 the application's session and the correct server is selected.
2035 It is recommended to specify this option to improve reliability.
2036
Cyril Bontéb21570a2009-11-29 20:04:48 +01002037 prefix When this option is specified, haproxy will match on the cookie
2038 prefix (or URL parameter prefix). The appsession value is the
2039 data following this prefix.
2040
2041 Example :
2042 appsession ASPSESSIONID len 64 timeout 3h prefix
2043
2044 This will match the cookie ASPSESSIONIDXXXX=XXXXX,
2045 the appsession value will be XXXX=XXXXX.
2046
2047 mode This option allows to change the URL parser mode.
2048 2 modes are currently supported :
2049 - path-parameters :
2050 The parser looks for the appsession in the path parameters
2051 part (each parameter is separated by a semi-colon), which is
2052 convenient for JSESSIONID for example.
2053 This is the default mode if the option is not set.
2054 - query-string :
2055 In this mode, the parser will look for the appsession in the
2056 query string.
2057
Willy Tarreau294d0f02015-08-10 19:40:12 +02002058 As of version 1.6, appsessions was removed. It is more flexible and more
2059 convenient to use stick-tables instead, and stick-tables support multi-master
2060 replication and data conservation across reloads, which appsessions did not.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002061
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01002062 See also : "cookie", "capture cookie", "balance", "stick", "stick-table",
2063 "ignore-persist", "nbproc" and "bind-process".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002064
2065
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01002066backlog <conns>
2067 Give hints to the system about the approximate listen backlog desired size
2068 May be used in sections : defaults | frontend | listen | backend
2069 yes | yes | yes | no
2070 Arguments :
2071 <conns> is the number of pending connections. Depending on the operating
2072 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02002073 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01002074
2075 In order to protect against SYN flood attacks, one solution is to increase
2076 the system's SYN backlog size. Depending on the system, sometimes it is just
2077 tunable via a system parameter, sometimes it is not adjustable at all, and
2078 sometimes the system relies on hints given by the application at the time of
2079 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
2080 to the listen() syscall. On systems which can make use of this value, it can
2081 sometimes be useful to be able to specify a different value, hence this
2082 backlog parameter.
2083
2084 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
2085 used as a hint and the system accepts up to the smallest greater power of
2086 two, and never more than some limits (usually 32768).
2087
2088 See also : "maxconn" and the target operating system's tuning guide.
2089
2090
Willy Tarreau0ba27502007-12-24 16:55:16 +01002091balance <algorithm> [ <arguments> ]
Willy Tarreau226071e2014-04-10 11:55:45 +02002092balance url_param <param> [check_post]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002093 Define the load balancing algorithm to be used in a backend.
2094 May be used in sections : defaults | frontend | listen | backend
2095 yes | no | yes | yes
2096 Arguments :
2097 <algorithm> is the algorithm used to select a server when doing load
2098 balancing. This only applies when no persistence information
2099 is available, or when a connection is redispatched to another
2100 server. <algorithm> may be one of the following :
2101
2102 roundrobin Each server is used in turns, according to their weights.
2103 This is the smoothest and fairest algorithm when the server's
2104 processing time remains equally distributed. This algorithm
2105 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02002106 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08002107 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02002108 large farms, when a server becomes up after having been down
2109 for a very short time, it may sometimes take a few hundreds
2110 requests for it to be re-integrated into the farm and start
2111 receiving traffic. This is normal, though very rare. It is
2112 indicated here in case you would have the chance to observe
2113 it, so that you don't worry.
2114
2115 static-rr Each server is used in turns, according to their weights.
2116 This algorithm is as similar to roundrobin except that it is
2117 static, which means that changing a server's weight on the
2118 fly will have no effect. On the other hand, it has no design
2119 limitation on the number of servers, and when a server goes
2120 up, it is always immediately reintroduced into the farm, once
2121 the full map is recomputed. It also uses slightly less CPU to
2122 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01002123
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01002124 leastconn The server with the lowest number of connections receives the
2125 connection. Round-robin is performed within groups of servers
2126 of the same load to ensure that all servers will be used. Use
2127 of this algorithm is recommended where very long sessions are
2128 expected, such as LDAP, SQL, TSE, etc... but is not very well
2129 suited for protocols using short sessions such as HTTP. This
2130 algorithm is dynamic, which means that server weights may be
2131 adjusted on the fly for slow starts for instance.
2132
Willy Tarreauf09c6602012-02-13 17:12:08 +01002133 first The first server with available connection slots receives the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002134 connection. The servers are chosen from the lowest numeric
Willy Tarreauf09c6602012-02-13 17:12:08 +01002135 identifier to the highest (see server parameter "id"), which
2136 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02002137 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01002138 not make sense to use this algorithm without setting maxconn.
2139 The purpose of this algorithm is to always use the smallest
2140 number of servers so that extra servers can be powered off
2141 during non-intensive hours. This algorithm ignores the server
2142 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02002143 or IMAP than HTTP, though it can be useful there too. In
2144 order to use this algorithm efficiently, it is recommended
2145 that a cloud controller regularly checks server usage to turn
2146 them off when unused, and regularly checks backend queue to
2147 turn new servers on when the queue inflates. Alternatively,
2148 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01002149
Willy Tarreau0ba27502007-12-24 16:55:16 +01002150 source The source IP address is hashed and divided by the total
2151 weight of the running servers to designate which server will
2152 receive the request. This ensures that the same client IP
2153 address will always reach the same server as long as no
2154 server goes down or up. If the hash result changes due to the
2155 number of running servers changing, many clients will be
2156 directed to a different server. This algorithm is generally
2157 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002158 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01002159 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002160 static by default, which means that changing a server's
2161 weight on the fly will have no effect, but this can be
2162 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002163
Oskar Stolc8dc41842012-05-19 10:19:54 +01002164 uri This algorithm hashes either the left part of the URI (before
2165 the question mark) or the whole URI (if the "whole" parameter
2166 is present) and divides the hash value by the total weight of
2167 the running servers. The result designates which server will
2168 receive the request. This ensures that the same URI will
2169 always be directed to the same server as long as no server
2170 goes up or down. This is used with proxy caches and
2171 anti-virus proxies in order to maximize the cache hit rate.
2172 Note that this algorithm may only be used in an HTTP backend.
2173 This algorithm is static by default, which means that
2174 changing a server's weight on the fly will have no effect,
2175 but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002176
Oskar Stolc8dc41842012-05-19 10:19:54 +01002177 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002178 "depth", both followed by a positive integer number. These
2179 options may be helpful when it is needed to balance servers
2180 based on the beginning of the URI only. The "len" parameter
2181 indicates that the algorithm should only consider that many
2182 characters at the beginning of the URI to compute the hash.
2183 Note that having "len" set to 1 rarely makes sense since most
2184 URIs start with a leading "/".
2185
2186 The "depth" parameter indicates the maximum directory depth
2187 to be used to compute the hash. One level is counted for each
2188 slash in the request. If both parameters are specified, the
2189 evaluation stops when either is reached.
2190
Willy Tarreau0ba27502007-12-24 16:55:16 +01002191 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002192 the query string of each HTTP GET request.
2193
2194 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02002195 request entity will be searched for the parameter argument,
2196 when it is not found in a query string after a question mark
Willy Tarreau226071e2014-04-10 11:55:45 +02002197 ('?') in the URL. The message body will only start to be
2198 analyzed once either the advertised amount of data has been
2199 received or the request buffer is full. In the unlikely event
2200 that chunked encoding is used, only the first chunk is
Cyril Bontédc4d9032012-04-08 21:57:39 +02002201 scanned. Parameter values separated by a chunk boundary, may
Willy Tarreau226071e2014-04-10 11:55:45 +02002202 be randomly balanced if at all. This keyword used to support
2203 an optional <max_wait> parameter which is now ignored.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002204
2205 If the parameter is found followed by an equal sign ('=') and
2206 a value, then the value is hashed and divided by the total
2207 weight of the running servers. The result designates which
2208 server will receive the request.
2209
2210 This is used to track user identifiers in requests and ensure
2211 that a same user ID will always be sent to the same server as
2212 long as no server goes up or down. If no value is found or if
2213 the parameter is not found, then a round robin algorithm is
2214 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002215 backend. This algorithm is static by default, which means
2216 that changing a server's weight on the fly will have no
2217 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002218
Cyril Bontédc4d9032012-04-08 21:57:39 +02002219 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
2220 request. Just as with the equivalent ACL 'hdr()' function,
2221 the header name in parenthesis is not case sensitive. If the
2222 header is absent or if it does not contain any value, the
2223 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01002224
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002225 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01002226 reducing the hash algorithm to the main domain part with some
2227 specific headers such as 'Host'. For instance, in the Host
2228 value "haproxy.1wt.eu", only "1wt" will be considered.
2229
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002230 This algorithm is static by default, which means that
2231 changing a server's weight on the fly will have no effect,
2232 but this can be changed using "hash-type".
2233
Emeric Brun736aa232009-06-30 17:56:00 +02002234 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02002235 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02002236 The RDP cookie <name> (or "mstshash" if omitted) will be
2237 looked up and hashed for each incoming TCP request. Just as
2238 with the equivalent ACL 'req_rdp_cookie()' function, the name
2239 is not case-sensitive. This mechanism is useful as a degraded
2240 persistence mode, as it makes it possible to always send the
2241 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002242 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02002243 used instead.
2244
2245 Note that for this to work, the frontend must ensure that an
2246 RDP cookie is already present in the request buffer. For this
2247 you must use 'tcp-request content accept' rule combined with
2248 a 'req_rdp_cookie_cnt' ACL.
2249
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002250 This algorithm is static by default, which means that
2251 changing a server's weight on the fly will have no effect,
2252 but this can be changed using "hash-type".
2253
Cyril Bontédc4d9032012-04-08 21:57:39 +02002254 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09002255
Willy Tarreau0ba27502007-12-24 16:55:16 +01002256 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002257 algorithms. Right now, only "url_param" and "uri" support an
2258 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002259
Willy Tarreau3cd9af22009-03-15 14:06:41 +01002260 The load balancing algorithm of a backend is set to roundrobin when no other
2261 algorithm, mode nor option have been set. The algorithm may only be set once
2262 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002263
2264 Examples :
2265 balance roundrobin
2266 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002267 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01002268 balance hdr(User-Agent)
2269 balance hdr(host)
2270 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002271
2272 Note: the following caveats and limitations on using the "check_post"
2273 extension with "url_param" must be considered :
2274
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002275 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002276 to determine if the parameters will be found in the body or entity which
2277 may contain binary data. Therefore another method may be required to
2278 restrict consideration of POST requests that have no URL parameters in
2279 the body. (see acl reqideny http_end)
2280
2281 - using a <max_wait> value larger than the request buffer size does not
2282 make sense and is useless. The buffer size is set at build time, and
2283 defaults to 16 kB.
2284
2285 - Content-Encoding is not supported, the parameter search will probably
2286 fail; and load balancing will fall back to Round Robin.
2287
2288 - Expect: 100-continue is not supported, load balancing will fall back to
2289 Round Robin.
2290
2291 - Transfer-Encoding (RFC2616 3.6.1) is only supported in the first chunk.
2292 If the entire parameter value is not present in the first chunk, the
2293 selection of server is undefined (actually, defined by how little
2294 actually appeared in the first chunk).
2295
2296 - This feature does not support generation of a 100, 411 or 501 response.
2297
2298 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002299 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002300 white space or control characters are found, indicating the end of what
2301 might be a URL parameter list. This is probably not a concern with SGML
2302 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002303
Willy Tarreau294d0f02015-08-10 19:40:12 +02002304 See also : "dispatch", "cookie", "transparent", "hash-type" and "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002305
2306
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002307bind [<address>]:<port_range> [, ...] [param*]
2308bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002309 Define one or several listening addresses and/or ports in a frontend.
2310 May be used in sections : defaults | frontend | listen | backend
2311 no | yes | yes | no
2312 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002313 <address> is optional and can be a host name, an IPv4 address, an IPv6
2314 address, or '*'. It designates the address the frontend will
2315 listen on. If unset, all IPv4 addresses of the system will be
2316 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01002317 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01002318 Optionally, an address family prefix may be used before the
2319 address to force the family regardless of the address format,
2320 which can be useful to specify a path to a unix socket with
2321 no slash ('/'). Currently supported prefixes are :
2322 - 'ipv4@' -> address is always IPv4
2323 - 'ipv6@' -> address is always IPv6
2324 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau70f72e02014-07-08 00:37:50 +02002325 - 'abns@' -> address is in abstract namespace (Linux only).
2326 Note: since abstract sockets are not "rebindable", they
2327 do not cope well with multi-process mode during
2328 soft-restart, so it is better to avoid them if
2329 nbproc is greater than 1. The effect is that if the
2330 new process fails to start, only one of the old ones
2331 will be able to rebind to the socket.
Willy Tarreau40aa0702013-03-10 23:51:38 +01002332 - 'fd@<n>' -> use file descriptor <n> inherited from the
2333 parent. The fd must be bound and may or may not already
2334 be listening.
William Lallemandb2f07452015-05-12 14:27:13 +02002335 You may want to reference some environment variables in the
2336 address parameter, see section 2.3 about environment
2337 variables.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002338
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002339 <port_range> is either a unique TCP port, or a port range for which the
2340 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002341 above. The port is mandatory for TCP listeners. Note that in
2342 the case of an IPv6 address, the port is always the number
2343 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002344 - a numerical port (ex: '80')
2345 - a dash-delimited ports range explicitly stating the lower
2346 and upper bounds (ex: '2000-2100') which are included in
2347 the range.
2348
2349 Particular care must be taken against port ranges, because
2350 every <address:port> couple consumes one socket (= a file
2351 descriptor), so it's easy to consume lots of descriptors
2352 with a simple range, and to run out of sockets. Also, each
2353 <address:port> couple must be used only once among all
2354 instances running on a same system. Please note that binding
2355 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002356 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002357 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002358
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002359 <path> is a UNIX socket path beginning with a slash ('/'). This is
2360 alternative to the TCP listening port. Haproxy will then
2361 receive UNIX connections on the socket located at this place.
2362 The path must begin with a slash and by default is absolute.
2363 It can be relative to the prefix defined by "unix-bind" in
2364 the global section. Note that the total length of the prefix
2365 followed by the socket path cannot exceed some system limits
2366 for UNIX sockets, which commonly are set to 107 characters.
2367
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002368 <param*> is a list of parameters common to all sockets declared on the
2369 same line. These numerous parameters depend on OS and build
2370 options and have a complete section dedicated to them. Please
2371 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002372
Willy Tarreau0ba27502007-12-24 16:55:16 +01002373 It is possible to specify a list of address:port combinations delimited by
2374 commas. The frontend will then listen on all of these addresses. There is no
2375 fixed limit to the number of addresses and ports which can be listened on in
2376 a frontend, as well as there is no limit to the number of "bind" statements
2377 in a frontend.
2378
2379 Example :
2380 listen http_proxy
2381 bind :80,:443
2382 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002383 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01002384
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002385 listen http_https_proxy
2386 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02002387 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002388
Willy Tarreau24709282013-03-10 21:32:12 +01002389 listen http_https_proxy_explicit
2390 bind ipv6@:80
2391 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
2392 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
2393
Willy Tarreaudad36a32013-03-11 01:20:04 +01002394 listen external_bind_app1
William Lallemandb2f07452015-05-12 14:27:13 +02002395 bind "fd@${FD_APP1}"
Willy Tarreaudad36a32013-03-11 01:20:04 +01002396
Willy Tarreau55dcaf62015-09-27 15:03:15 +02002397 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
2398 sun_path length is used for the address length. Some other programs
2399 such as socat use the string length only by default. Pass the option
2400 ",unix-tightsocklen=0" to any abstract socket definition in socat to
2401 make it compatible with HAProxy's.
2402
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002403 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002404 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002405
2406
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002407bind-process [ all | odd | even | <number 1-64>[-<number 1-64>] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002408 Limit visibility of an instance to a certain set of processes numbers.
2409 May be used in sections : defaults | frontend | listen | backend
2410 yes | yes | yes | yes
2411 Arguments :
2412 all All process will see this instance. This is the default. It
2413 may be used to override a default value.
2414
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002415 odd This instance will be enabled on processes 1,3,5,...63. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002416 option may be combined with other numbers.
2417
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002418 even This instance will be enabled on processes 2,4,6,...64. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002419 option may be combined with other numbers. Do not use it
2420 with less than 2 processes otherwise some instances might be
2421 missing from all processes.
2422
Willy Tarreau110ecc12012-11-15 17:50:01 +01002423 number The instance will be enabled on this process number or range,
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002424 whose values must all be between 1 and 32 or 64 depending on
Willy Tarreau102df612014-05-07 23:56:38 +02002425 the machine's word size. If a proxy is bound to process
2426 numbers greater than the configured global.nbproc, it will
2427 either be forced to process #1 if a single process was
2428 specified, or to all processes otherwise.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002429
2430 This keyword limits binding of certain instances to certain processes. This
2431 is useful in order not to have too many processes listening to the same
2432 ports. For instance, on a dual-core machine, it might make sense to set
2433 'nbproc 2' in the global section, then distributes the listeners among 'odd'
2434 and 'even' instances.
2435
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002436 At the moment, it is not possible to reference more than 32 or 64 processes
2437 using this keyword, but this should be more than enough for most setups.
2438 Please note that 'all' really means all processes regardless of the machine's
2439 word size, and is not limited to the first 32 or 64.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002440
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002441 Each "bind" line may further be limited to a subset of the proxy's processes,
2442 please consult the "process" bind keyword in section 5.1.
2443
Willy Tarreaub369a042014-09-16 13:21:03 +02002444 When a frontend has no explicit "bind-process" line, it tries to bind to all
2445 the processes referenced by its "bind" lines. That means that frontends can
2446 easily adapt to their listeners' processes.
2447
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002448 If some backends are referenced by frontends bound to other processes, the
2449 backend automatically inherits the frontend's processes.
2450
2451 Example :
2452 listen app_ip1
2453 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002454 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002455
2456 listen app_ip2
2457 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002458 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002459
2460 listen management
2461 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002462 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002463
Willy Tarreau110ecc12012-11-15 17:50:01 +01002464 listen management
2465 bind 10.0.0.4:80
2466 bind-process 1-4
2467
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002468 See also : "nbproc" in global section, and "process" in section 5.1.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002469
2470
Willy Tarreau0ba27502007-12-24 16:55:16 +01002471block { if | unless } <condition>
2472 Block a layer 7 request if/unless a condition is matched
2473 May be used in sections : defaults | frontend | listen | backend
2474 no | yes | yes | yes
2475
2476 The HTTP request will be blocked very early in the layer 7 processing
2477 if/unless <condition> is matched. A 403 error will be returned if the request
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002478 is blocked. The condition has to reference ACLs (see section 7). This is
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02002479 typically used to deny access to certain sensitive resources if some
Willy Tarreau0ba27502007-12-24 16:55:16 +01002480 conditions are met or not met. There is no fixed limit to the number of
2481 "block" statements per instance.
2482
2483 Example:
2484 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2485 acl invalid_src src_port 0:1023
2486 acl local_dst hdr(host) -i localhost
2487 block if invalid_src || local_dst
2488
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002489 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002490
2491
2492capture cookie <name> len <length>
2493 Capture and log a cookie in the request and in the response.
2494 May be used in sections : defaults | frontend | listen | backend
2495 no | yes | yes | no
2496 Arguments :
2497 <name> is the beginning of the name of the cookie to capture. In order
2498 to match the exact name, simply suffix the name with an equal
2499 sign ('='). The full name will appear in the logs, which is
2500 useful with application servers which adjust both the cookie name
2501 and value (eg: ASPSESSIONXXXXX).
2502
2503 <length> is the maximum number of characters to report in the logs, which
2504 include the cookie name, the equal sign and the value, all in the
2505 standard "name=value" form. The string will be truncated on the
2506 right if it exceeds <length>.
2507
2508 Only the first cookie is captured. Both the "cookie" request headers and the
2509 "set-cookie" response headers are monitored. This is particularly useful to
2510 check for application bugs causing session crossing or stealing between
2511 users, because generally the user's cookies can only change on a login page.
2512
2513 When the cookie was not presented by the client, the associated log column
2514 will report "-". When a request does not cause a cookie to be assigned by the
2515 server, a "-" is reported in the response column.
2516
2517 The capture is performed in the frontend only because it is necessary that
2518 the log format does not change for a given frontend depending on the
2519 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01002520 "capture cookie" statement in a frontend. The maximum capture length is set
2521 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
2522 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002523
2524 Example:
2525 capture cookie ASPSESSION len 32
2526
2527 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002528 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002529
2530
2531capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01002532 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002533 May be used in sections : defaults | frontend | listen | backend
2534 no | yes | yes | no
2535 Arguments :
2536 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002537 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01002538 appear in the requests, with the first letter of each word in
2539 upper case. The header name will not appear in the logs, only the
2540 value is reported, but the position in the logs is respected.
2541
2542 <length> is the maximum number of characters to extract from the value and
2543 report in the logs. The string will be truncated on the right if
2544 it exceeds <length>.
2545
Willy Tarreau4460d032012-11-21 23:37:37 +01002546 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01002547 value will be added to the logs between braces ('{}'). If multiple headers
2548 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002549 in the same order they were declared in the configuration. Non-existent
2550 headers will be logged just as an empty string. Common uses for request
2551 header captures include the "Host" field in virtual hosting environments, the
2552 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002553 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002554 environments to find where the request came from.
2555
2556 Note that when capturing headers such as "User-agent", some spaces may be
2557 logged, making the log analysis more difficult. Thus be careful about what
2558 you log if you know your log parser is not smart enough to rely on the
2559 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002560
Willy Tarreau0900abb2012-11-22 00:21:46 +01002561 There is no limit to the number of captured request headers nor to their
2562 length, though it is wise to keep them low to limit memory usage per session.
2563 In order to keep log format consistent for a same frontend, header captures
2564 can only be declared in a frontend. It is not possible to specify a capture
2565 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002566
2567 Example:
2568 capture request header Host len 15
2569 capture request header X-Forwarded-For len 15
Cyril Bontéd1b0f7c2015-10-26 22:37:39 +01002570 capture request header Referer len 15
Willy Tarreau0ba27502007-12-24 16:55:16 +01002571
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002572 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01002573 about logging.
2574
2575
2576capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01002577 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002578 May be used in sections : defaults | frontend | listen | backend
2579 no | yes | yes | no
2580 Arguments :
2581 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002582 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01002583 appear in the response, with the first letter of each word in
2584 upper case. The header name will not appear in the logs, only the
2585 value is reported, but the position in the logs is respected.
2586
2587 <length> is the maximum number of characters to extract from the value and
2588 report in the logs. The string will be truncated on the right if
2589 it exceeds <length>.
2590
Willy Tarreau4460d032012-11-21 23:37:37 +01002591 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01002592 result will be added to the logs between braces ('{}') after the captured
2593 request headers. If multiple headers are captured, they will be delimited by
2594 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002595 the configuration. Non-existent headers will be logged just as an empty
2596 string. Common uses for response header captures include the "Content-length"
2597 header which indicates how many bytes are expected to be returned, the
2598 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002599
Willy Tarreau0900abb2012-11-22 00:21:46 +01002600 There is no limit to the number of captured response headers nor to their
2601 length, though it is wise to keep them low to limit memory usage per session.
2602 In order to keep log format consistent for a same frontend, header captures
2603 can only be declared in a frontend. It is not possible to specify a capture
2604 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002605
2606 Example:
2607 capture response header Content-length len 9
2608 capture response header Location len 15
2609
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002610 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01002611 about logging.
2612
2613
Cyril Bontéf0c60612010-02-06 14:44:47 +01002614clitimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002615 Set the maximum inactivity time on the client side.
2616 May be used in sections : defaults | frontend | listen | backend
2617 yes | yes | yes | no
2618 Arguments :
2619 <timeout> is the timeout value is specified in milliseconds by default, but
2620 can be in any other unit if the number is suffixed by the unit,
2621 as explained at the top of this document.
2622
2623 The inactivity timeout applies when the client is expected to acknowledge or
2624 send data. In HTTP mode, this timeout is particularly important to consider
2625 during the first phase, when the client sends the request, and during the
2626 response while it is reading data sent by the server. The value is specified
2627 in milliseconds by default, but can be in any other unit if the number is
2628 suffixed by the unit, as specified at the top of this document. In TCP mode
2629 (and to a lesser extent, in HTTP mode), it is highly recommended that the
2630 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002631 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01002632 losses by specifying timeouts that are slightly above multiples of 3 seconds
2633 (eg: 4 or 5 seconds).
2634
2635 This parameter is specific to frontends, but can be specified once for all in
2636 "defaults" sections. This is in fact one of the easiest solutions not to
2637 forget about it. An unspecified timeout results in an infinite timeout, which
2638 is not recommended. Such a usage is accepted and works but reports a warning
2639 during startup because it may results in accumulation of expired sessions in
2640 the system if the system's timeouts are not configured either.
2641
2642 This parameter is provided for compatibility but is currently deprecated.
2643 Please use "timeout client" instead.
2644
Willy Tarreau036fae02008-01-06 13:24:40 +01002645 See also : "timeout client", "timeout http-request", "timeout server", and
2646 "srvtimeout".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002647
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002648compression algo <algorithm> ...
2649compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02002650compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02002651 Enable HTTP compression.
2652 May be used in sections : defaults | frontend | listen | backend
2653 yes | yes | yes | yes
2654 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002655 algo is followed by the list of supported compression algorithms.
2656 type is followed by the list of MIME types that will be compressed.
2657 offload makes haproxy work as a compression offloader only (see notes).
2658
2659 The currently supported algorithms are :
Willy Tarreauc91840a2015-03-28 17:00:39 +01002660 identity this is mostly for debugging, and it was useful for developing
2661 the compression feature. Identity does not apply any change on
2662 data.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002663
Willy Tarreauc91840a2015-03-28 17:00:39 +01002664 gzip applies gzip compression. This setting is only available when
Baptiste Assmannf085d632015-12-21 17:57:32 +01002665 support for zlib or libslz was built in.
Willy Tarreauc91840a2015-03-28 17:00:39 +01002666
2667 deflate same as "gzip", but with deflate algorithm and zlib format.
2668 Note that this algorithm has ambiguous support on many
2669 browsers and no support at all from recent ones. It is
2670 strongly recommended not to use it for anything else than
2671 experimentation. This setting is only available when support
Baptiste Assmannf085d632015-12-21 17:57:32 +01002672 for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002673
Willy Tarreauc91840a2015-03-28 17:00:39 +01002674 raw-deflate same as "deflate" without the zlib wrapper, and used as an
2675 alternative when the browser wants "deflate". All major
2676 browsers understand it and despite violating the standards,
2677 it is known to work better than "deflate", at least on MSIE
2678 and some versions of Safari. Do not use it in conjunction
2679 with "deflate", use either one or the other since both react
2680 to the same Accept-Encoding token. This setting is only
Baptiste Assmannf085d632015-12-21 17:57:32 +01002681 available when support for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002682
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04002683 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002684 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04002685 If backend servers support HTTP compression, these directives
2686 will be no-op: haproxy will see the compressed response and will not
2687 compress again. If backend servers do not support HTTP compression and
2688 there is Accept-Encoding header in request, haproxy will compress the
2689 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02002690
2691 The "offload" setting makes haproxy remove the Accept-Encoding header to
2692 prevent backend servers from compressing responses. It is strongly
2693 recommended not to do this because this means that all the compression work
2694 will be done on the single point where haproxy is located. However in some
2695 deployment scenarios, haproxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04002696 with broken HTTP compression implementation which can't be turned off.
2697 In that case haproxy can be used to prevent that gateway from emitting
2698 invalid payloads. In this case, simply removing the header in the
2699 configuration does not work because it applies before the header is parsed,
2700 so that prevents haproxy from compressing. The "offload" setting should
Willy Tarreauffea9fd2014-07-12 16:37:02 +02002701 then be used for such scenarios. Note: for now, the "offload" setting is
2702 ignored when set in a defaults section.
William Lallemand82fe75c2012-10-23 10:25:10 +02002703
William Lallemand05097442012-11-20 12:14:28 +01002704 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002705 * the request does not advertise a supported compression algorithm in the
2706 "Accept-Encoding" header
2707 * the response message is not HTTP/1.1
William Lallemandd3002612012-11-26 14:34:47 +01002708 * HTTP status code is not 200
William Lallemand8bb4e342013-12-10 17:28:48 +01002709 * response header "Transfer-Encoding" contains "chunked" (Temporary
2710 Workaround)
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002711 * response contain neither a "Content-Length" header nor a
2712 "Transfer-Encoding" whose last value is "chunked"
2713 * response contains a "Content-Type" header whose first value starts with
2714 "multipart"
2715 * the response contains the "no-transform" value in the "Cache-control"
2716 header
2717 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
2718 and later
2719 * The response contains a "Content-Encoding" header, indicating that the
2720 response is already compressed (see compression offload)
William Lallemand05097442012-11-20 12:14:28 +01002721
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002722 Note: The compression does not rewrite Etag headers, and does not emit the
2723 Warning header.
William Lallemand05097442012-11-20 12:14:28 +01002724
William Lallemand82fe75c2012-10-23 10:25:10 +02002725 Examples :
2726 compression algo gzip
2727 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01002728
Christopher Fauletc3fe5332016-04-07 15:30:10 +02002729
Cyril Bontéf0c60612010-02-06 14:44:47 +01002730contimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002731 Set the maximum time to wait for a connection attempt to a server to succeed.
2732 May be used in sections : defaults | frontend | listen | backend
2733 yes | no | yes | yes
2734 Arguments :
2735 <timeout> is the timeout value is specified in milliseconds by default, but
2736 can be in any other unit if the number is suffixed by the unit,
2737 as explained at the top of this document.
2738
2739 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002740 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01002741 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01002742 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
2743 connect timeout also presets the queue timeout to the same value if this one
2744 has not been specified. Historically, the contimeout was also used to set the
2745 tarpit timeout in a listen section, which is not possible in a pure frontend.
2746
2747 This parameter is specific to backends, but can be specified once for all in
2748 "defaults" sections. This is in fact one of the easiest solutions not to
2749 forget about it. An unspecified timeout results in an infinite timeout, which
2750 is not recommended. Such a usage is accepted and works but reports a warning
2751 during startup because it may results in accumulation of failed sessions in
2752 the system if the system's timeouts are not configured either.
2753
2754 This parameter is provided for backwards compatibility but is currently
2755 deprecated. Please use "timeout connect", "timeout queue" or "timeout tarpit"
2756 instead.
2757
2758 See also : "timeout connect", "timeout queue", "timeout tarpit",
2759 "timeout server", "contimeout".
2760
2761
Willy Tarreau55165fe2009-05-10 12:02:55 +02002762cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02002763 [ postonly ] [ preserve ] [ httponly ] [ secure ]
2764 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002765 Enable cookie-based persistence in a backend.
2766 May be used in sections : defaults | frontend | listen | backend
2767 yes | no | yes | yes
2768 Arguments :
2769 <name> is the name of the cookie which will be monitored, modified or
2770 inserted in order to bring persistence. This cookie is sent to
2771 the client via a "Set-Cookie" header in the response, and is
2772 brought back by the client in a "Cookie" header in all requests.
2773 Special care should be taken to choose a name which does not
2774 conflict with any likely application cookie. Also, if the same
2775 backends are subject to be used by the same clients (eg:
2776 HTTP/HTTPS), care should be taken to use different cookie names
2777 between all backends if persistence between them is not desired.
2778
2779 rewrite This keyword indicates that the cookie will be provided by the
2780 server and that haproxy will have to modify its value to set the
2781 server's identifier in it. This mode is handy when the management
2782 of complex combinations of "Set-cookie" and "Cache-control"
2783 headers is left to the application. The application can then
2784 decide whether or not it is appropriate to emit a persistence
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01002785 cookie. Since all responses should be monitored, this mode
2786 doesn't work in HTTP tunnel mode. Unless the application
2787 behaviour is very complex and/or broken, it is advised not to
2788 start with this mode for new deployments. This keyword is
2789 incompatible with "insert" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002790
2791 insert This keyword indicates that the persistence cookie will have to
Willy Tarreaua79094d2010-08-31 22:54:15 +02002792 be inserted by haproxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002793
Willy Tarreaua79094d2010-08-31 22:54:15 +02002794 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002795 server. When used without the "preserve" option, if the server
2796 emits a cookie with the same name, it will be remove before
2797 processing. For this reason, this mode can be used to upgrade
2798 existing configurations running in the "rewrite" mode. The cookie
2799 will only be a session cookie and will not be stored on the
2800 client's disk. By default, unless the "indirect" option is added,
2801 the server will see the cookies emitted by the client. Due to
2802 caching effects, it is generally wise to add the "nocache" or
2803 "postonly" keywords (see below). The "insert" keyword is not
2804 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002805
2806 prefix This keyword indicates that instead of relying on a dedicated
2807 cookie for the persistence, an existing one will be completed.
2808 This may be needed in some specific environments where the client
2809 does not support more than one single cookie and the application
2810 already needs it. In this case, whenever the server sets a cookie
2811 named <name>, it will be prefixed with the server's identifier
2812 and a delimiter. The prefix will be removed from all client
2813 requests so that the server still finds the cookie it emitted.
2814 Since all requests and responses are subject to being modified,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01002815 this mode doesn't work with tunnel mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02002816 not compatible with "rewrite" and "insert". Note: it is highly
2817 recommended not to use "indirect" with "prefix", otherwise server
2818 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002819
Willy Tarreaua79094d2010-08-31 22:54:15 +02002820 indirect When this option is specified, no cookie will be emitted to a
2821 client which already has a valid one for the server which has
2822 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002823 it will be removed, unless the "preserve" option is also set. In
2824 "insert" mode, this will additionally remove cookies from the
2825 requests transmitted to the server, making the persistence
2826 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02002827 Note: it is highly recommended not to use "indirect" with
2828 "prefix", otherwise server cookie updates would not be sent to
2829 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002830
2831 nocache This option is recommended in conjunction with the insert mode
2832 when there is a cache between the client and HAProxy, as it
2833 ensures that a cacheable response will be tagged non-cacheable if
2834 a cookie needs to be inserted. This is important because if all
2835 persistence cookies are added on a cacheable home page for
2836 instance, then all customers will then fetch the page from an
2837 outer cache and will all share the same persistence cookie,
2838 leading to one server receiving much more traffic than others.
2839 See also the "insert" and "postonly" options.
2840
2841 postonly This option ensures that cookie insertion will only be performed
2842 on responses to POST requests. It is an alternative to the
2843 "nocache" option, because POST responses are not cacheable, so
2844 this ensures that the persistence cookie will never get cached.
2845 Since most sites do not need any sort of persistence before the
2846 first POST which generally is a login request, this is a very
2847 efficient method to optimize caching without risking to find a
2848 persistence cookie in the cache.
2849 See also the "insert" and "nocache" options.
2850
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002851 preserve This option may only be used with "insert" and/or "indirect". It
2852 allows the server to emit the persistence cookie itself. In this
2853 case, if a cookie is found in the response, haproxy will leave it
2854 untouched. This is useful in order to end persistence after a
2855 logout request for instance. For this, the server just has to
2856 emit a cookie with an invalid value (eg: empty) or with a date in
2857 the past. By combining this mechanism with the "disable-on-404"
2858 check option, it is possible to perform a completely graceful
2859 shutdown because users will definitely leave the server after
2860 they logout.
2861
Willy Tarreau4992dd22012-05-31 21:02:17 +02002862 httponly This option tells haproxy to add an "HttpOnly" cookie attribute
2863 when a cookie is inserted. This attribute is used so that a
2864 user agent doesn't share the cookie with non-HTTP components.
2865 Please check RFC6265 for more information on this attribute.
2866
2867 secure This option tells haproxy to add a "Secure" cookie attribute when
2868 a cookie is inserted. This attribute is used so that a user agent
2869 never emits this cookie over non-secure channels, which means
2870 that a cookie learned with this flag will be presented only over
2871 SSL/TLS connections. Please check RFC6265 for more information on
2872 this attribute.
2873
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002874 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002875 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01002876 name. If the domain begins with a dot, the browser is allowed to
2877 use it for any host ending with that name. It is also possible to
2878 specify several domain names by invoking this option multiple
2879 times. Some browsers might have small limits on the number of
2880 domains, so be careful when doing that. For the record, sending
2881 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002882
Willy Tarreau996a92c2010-10-13 19:30:47 +02002883 maxidle This option allows inserted cookies to be ignored after some idle
2884 time. It only works with insert-mode cookies. When a cookie is
2885 sent to the client, the date this cookie was emitted is sent too.
2886 Upon further presentations of this cookie, if the date is older
2887 than the delay indicated by the parameter (in seconds), it will
2888 be ignored. Otherwise, it will be refreshed if needed when the
2889 response is sent to the client. This is particularly useful to
2890 prevent users who never close their browsers from remaining for
2891 too long on the same server (eg: after a farm size change). When
2892 this option is set and a cookie has no date, it is always
2893 accepted, but gets refreshed in the response. This maintains the
2894 ability for admins to access their sites. Cookies that have a
2895 date in the future further than 24 hours are ignored. Doing so
2896 lets admins fix timezone issues without risking kicking users off
2897 the site.
2898
2899 maxlife This option allows inserted cookies to be ignored after some life
2900 time, whether they're in use or not. It only works with insert
2901 mode cookies. When a cookie is first sent to the client, the date
2902 this cookie was emitted is sent too. Upon further presentations
2903 of this cookie, if the date is older than the delay indicated by
2904 the parameter (in seconds), it will be ignored. If the cookie in
2905 the request has no date, it is accepted and a date will be set.
2906 Cookies that have a date in the future further than 24 hours are
2907 ignored. Doing so lets admins fix timezone issues without risking
2908 kicking users off the site. Contrary to maxidle, this value is
2909 not refreshed, only the first visit date counts. Both maxidle and
2910 maxlife may be used at the time. This is particularly useful to
2911 prevent users who never close their browsers from remaining for
2912 too long on the same server (eg: after a farm size change). This
2913 is stronger than the maxidle method in that it forces a
2914 redispatch after some absolute delay.
2915
Willy Tarreau0ba27502007-12-24 16:55:16 +01002916 There can be only one persistence cookie per HTTP backend, and it can be
2917 declared in a defaults section. The value of the cookie will be the value
2918 indicated after the "cookie" keyword in a "server" statement. If no cookie
2919 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002920
Willy Tarreau0ba27502007-12-24 16:55:16 +01002921 Examples :
2922 cookie JSESSIONID prefix
2923 cookie SRV insert indirect nocache
2924 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02002925 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01002926
Willy Tarreau294d0f02015-08-10 19:40:12 +02002927 See also : "balance source", "capture cookie", "server" and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002928
Willy Tarreau983e01e2010-01-11 18:42:06 +01002929
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02002930declare capture [ request | response ] len <length>
2931 Declares a capture slot.
2932 May be used in sections : defaults | frontend | listen | backend
2933 no | yes | yes | no
2934 Arguments:
2935 <length> is the length allowed for the capture.
2936
2937 This declaration is only available in the frontend or listen section, but the
2938 reserved slot can be used in the backends. The "request" keyword allocates a
2939 capture slot for use in the request, and "response" allocates a capture slot
2940 for use in the response.
2941
2942 See also: "capture-req", "capture-res" (sample converters),
Baptiste Assmann5ac425c2015-10-21 23:13:46 +02002943 "capture.req.hdr", "capture.res.hdr" (sample fetches),
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02002944 "http-request capture" and "http-response capture".
2945
2946
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002947default-server [param*]
2948 Change default options for a server in a backend
2949 May be used in sections : defaults | frontend | listen | backend
2950 yes | no | yes | yes
2951 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01002952 <param*> is a list of parameters for this server. The "default-server"
2953 keyword accepts an important number of options and has a complete
2954 section dedicated to it. Please refer to section 5 for more
2955 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002956
Willy Tarreau983e01e2010-01-11 18:42:06 +01002957 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002958 default-server inter 1000 weight 13
2959
2960 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01002961
Willy Tarreau983e01e2010-01-11 18:42:06 +01002962
Willy Tarreau0ba27502007-12-24 16:55:16 +01002963default_backend <backend>
2964 Specify the backend to use when no "use_backend" rule has been matched.
2965 May be used in sections : defaults | frontend | listen | backend
2966 yes | yes | yes | no
2967 Arguments :
2968 <backend> is the name of the backend to use.
2969
2970 When doing content-switching between frontend and backends using the
2971 "use_backend" keyword, it is often useful to indicate which backend will be
2972 used when no rule has matched. It generally is the dynamic backend which
2973 will catch all undetermined requests.
2974
Willy Tarreau0ba27502007-12-24 16:55:16 +01002975 Example :
2976
2977 use_backend dynamic if url_dyn
2978 use_backend static if url_css url_img extension_img
2979 default_backend dynamic
2980
Willy Tarreau98d04852015-05-26 12:18:29 +02002981 See also : "use_backend"
Willy Tarreau2769aa02007-12-27 18:26:09 +01002982
Willy Tarreau0ba27502007-12-24 16:55:16 +01002983
Baptiste Assmann27f51342013-10-09 06:51:49 +02002984description <string>
2985 Describe a listen, frontend or backend.
2986 May be used in sections : defaults | frontend | listen | backend
2987 no | yes | yes | yes
2988 Arguments : string
2989
2990 Allows to add a sentence to describe the related object in the HAProxy HTML
2991 stats page. The description will be printed on the right of the object name
2992 it describes.
2993 No need to backslash spaces in the <string> arguments.
2994
2995
Willy Tarreau0ba27502007-12-24 16:55:16 +01002996disabled
2997 Disable a proxy, frontend or backend.
2998 May be used in sections : defaults | frontend | listen | backend
2999 yes | yes | yes | yes
3000 Arguments : none
3001
3002 The "disabled" keyword is used to disable an instance, mainly in order to
3003 liberate a listening port or to temporarily disable a service. The instance
3004 will still be created and its configuration will be checked, but it will be
3005 created in the "stopped" state and will appear as such in the statistics. It
3006 will not receive any traffic nor will it send any health-checks or logs. It
3007 is possible to disable many instances at once by adding the "disabled"
3008 keyword in a "defaults" section.
3009
3010 See also : "enabled"
3011
3012
Willy Tarreau5ce94572010-06-07 14:35:41 +02003013dispatch <address>:<port>
3014 Set a default server address
3015 May be used in sections : defaults | frontend | listen | backend
3016 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003017 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02003018
3019 <address> is the IPv4 address of the default server. Alternatively, a
3020 resolvable hostname is supported, but this name will be resolved
3021 during start-up.
3022
3023 <ports> is a mandatory port specification. All connections will be sent
3024 to this port, and it is not permitted to use port offsets as is
3025 possible with normal servers.
3026
Willy Tarreau787aed52011-04-15 06:45:37 +02003027 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02003028 server can take the connection. In the past it was used to forward non
3029 persistent connections to an auxiliary load balancer. Due to its simple
3030 syntax, it has also been used for simple TCP relays. It is recommended not to
3031 use it for more clarity, and to use the "server" directive instead.
3032
3033 See also : "server"
3034
3035
Willy Tarreau0ba27502007-12-24 16:55:16 +01003036enabled
3037 Enable a proxy, frontend or backend.
3038 May be used in sections : defaults | frontend | listen | backend
3039 yes | yes | yes | yes
3040 Arguments : none
3041
3042 The "enabled" keyword is used to explicitly enable an instance, when the
3043 defaults has been set to "disabled". This is very rarely used.
3044
3045 See also : "disabled"
3046
3047
3048errorfile <code> <file>
3049 Return a file contents instead of errors generated by HAProxy
3050 May be used in sections : defaults | frontend | listen | backend
3051 yes | yes | yes | yes
3052 Arguments :
3053 <code> is the HTTP status code. Currently, HAProxy is capable of
CJ Ess108b1dd2015-04-07 12:03:37 -04003054 generating codes 200, 400, 403, 405, 408, 429, 500, 502, 503, and
3055 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003056
3057 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003058 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01003059 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01003060 error pages, and to use absolute paths, since files are read
3061 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003062
3063 It is important to understand that this keyword is not meant to rewrite
3064 errors returned by the server, but errors detected and returned by HAProxy.
3065 This is why the list of supported errors is limited to a small set.
3066
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003067 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3068
Willy Tarreau0ba27502007-12-24 16:55:16 +01003069 The files are returned verbatim on the TCP socket. This allows any trick such
3070 as redirections to another URL or site, as well as tricks to clean cookies,
3071 force enable or disable caching, etc... The package provides default error
3072 files returning the same contents as default errors.
3073
Willy Tarreau59140a22009-02-22 12:02:30 +01003074 The files should not exceed the configured buffer size (BUFSIZE), which
3075 generally is 8 or 16 kB, otherwise they will be truncated. It is also wise
3076 not to put any reference to local contents (eg: images) in order to avoid
3077 loops between the client and HAProxy when all servers are down, causing an
3078 error to be returned instead of an image. For better HTTP compliance, it is
3079 recommended that all header lines end with CR-LF and not LF alone.
3080
Willy Tarreau0ba27502007-12-24 16:55:16 +01003081 The files are read at the same time as the configuration and kept in memory.
3082 For this reason, the errors continue to be returned even when the process is
3083 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01003084 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01003085 403 status code and interrogating a blocked URL.
3086
3087 See also : "errorloc", "errorloc302", "errorloc303"
3088
Willy Tarreau59140a22009-02-22 12:02:30 +01003089 Example :
3090 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
Willy Tarreau989222a2016-01-15 10:26:26 +01003091 errorfile 408 /dev/null # work around Chrome pre-connect bug
Willy Tarreau59140a22009-02-22 12:02:30 +01003092 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
3093 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
3094
Willy Tarreau2769aa02007-12-27 18:26:09 +01003095
3096errorloc <code> <url>
3097errorloc302 <code> <url>
3098 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3099 May be used in sections : defaults | frontend | listen | backend
3100 yes | yes | yes | yes
3101 Arguments :
3102 <code> is the HTTP status code. Currently, HAProxy is capable of
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003103 generating codes 200, 400, 403, 408, 500, 502, 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003104
3105 <url> it is the exact contents of the "Location" header. It may contain
3106 either a relative URI to an error page hosted on the same site,
3107 or an absolute URI designating an error page on another site.
3108 Special care should be given to relative URIs to avoid redirect
3109 loops if the URI itself may generate the same error (eg: 500).
3110
3111 It is important to understand that this keyword is not meant to rewrite
3112 errors returned by the server, but errors detected and returned by HAProxy.
3113 This is why the list of supported errors is limited to a small set.
3114
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003115 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3116
Willy Tarreau2769aa02007-12-27 18:26:09 +01003117 Note that both keyword return the HTTP 302 status code, which tells the
3118 client to fetch the designated URL using the same HTTP method. This can be
3119 quite problematic in case of non-GET methods such as POST, because the URL
3120 sent to the client might not be allowed for something other than GET. To
Willy Tarreau989222a2016-01-15 10:26:26 +01003121 work around this problem, please use "errorloc303" which send the HTTP 303
Willy Tarreau2769aa02007-12-27 18:26:09 +01003122 status code, indicating to the client that the URL must be fetched with a GET
3123 request.
3124
3125 See also : "errorfile", "errorloc303"
3126
3127
3128errorloc303 <code> <url>
3129 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3130 May be used in sections : defaults | frontend | listen | backend
3131 yes | yes | yes | yes
3132 Arguments :
3133 <code> is the HTTP status code. Currently, HAProxy is capable of
3134 generating codes 400, 403, 408, 500, 502, 503, and 504.
3135
3136 <url> it is the exact contents of the "Location" header. It may contain
3137 either a relative URI to an error page hosted on the same site,
3138 or an absolute URI designating an error page on another site.
3139 Special care should be given to relative URIs to avoid redirect
3140 loops if the URI itself may generate the same error (eg: 500).
3141
3142 It is important to understand that this keyword is not meant to rewrite
3143 errors returned by the server, but errors detected and returned by HAProxy.
3144 This is why the list of supported errors is limited to a small set.
3145
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003146 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3147
Willy Tarreau2769aa02007-12-27 18:26:09 +01003148 Note that both keyword return the HTTP 303 status code, which tells the
3149 client to fetch the designated URL using the same HTTP GET method. This
3150 solves the usual problems associated with "errorloc" and the 302 code. It is
3151 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003152 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003153
3154 See also : "errorfile", "errorloc", "errorloc302"
3155
3156
Simon Horman51a1cf62015-02-03 13:00:44 +09003157email-alert from <emailaddr>
3158 Declare the from email address to be used in both the envelope and header
3159 of email alerts. This is the address that email alerts are sent from.
3160 May be used in sections: defaults | frontend | listen | backend
3161 yes | yes | yes | yes
3162
3163 Arguments :
3164
3165 <emailaddr> is the from email address to use when sending email alerts
3166
3167 Also requires "email-alert mailers" and "email-alert to" to be set
3168 and if so sending email alerts is enabled for the proxy.
3169
Simon Horman64e34162015-02-06 11:11:57 +09003170 See also : "email-alert level", "email-alert mailers",
Cyril Bonté307ee1e2015-09-28 23:16:06 +02003171 "email-alert myhostname", "email-alert to", section 3.6 about
3172 mailers.
Simon Horman64e34162015-02-06 11:11:57 +09003173
3174
3175email-alert level <level>
3176 Declare the maximum log level of messages for which email alerts will be
3177 sent. This acts as a filter on the sending of email alerts.
3178 May be used in sections: defaults | frontend | listen | backend
3179 yes | yes | yes | yes
3180
3181 Arguments :
3182
3183 <level> One of the 8 syslog levels:
3184 emerg alert crit err warning notice info debug
3185 The above syslog levels are ordered from lowest to highest.
3186
3187 By default level is alert
3188
3189 Also requires "email-alert from", "email-alert mailers" and
3190 "email-alert to" to be set and if so sending email alerts is enabled
3191 for the proxy.
3192
Simon Horman1421e212015-04-30 13:10:35 +09003193 Alerts are sent when :
3194
3195 * An un-paused server is marked as down and <level> is alert or lower
3196 * A paused server is marked as down and <level> is notice or lower
3197 * A server is marked as up or enters the drain state and <level>
3198 is notice or lower
3199 * "option log-health-checks" is enabled, <level> is info or lower,
3200 and a health check status update occurs
3201
Simon Horman64e34162015-02-06 11:11:57 +09003202 See also : "email-alert from", "email-alert mailers",
3203 "email-alert myhostname", "email-alert to",
Simon Horman51a1cf62015-02-03 13:00:44 +09003204 section 3.6 about mailers.
3205
3206
3207email-alert mailers <mailersect>
3208 Declare the mailers to be used when sending email alerts
3209 May be used in sections: defaults | frontend | listen | backend
3210 yes | yes | yes | yes
3211
3212 Arguments :
3213
3214 <mailersect> is the name of the mailers section to send email alerts.
3215
3216 Also requires "email-alert from" and "email-alert to" to be set
3217 and if so sending email alerts is enabled for the proxy.
3218
Simon Horman64e34162015-02-06 11:11:57 +09003219 See also : "email-alert from", "email-alert level", "email-alert myhostname",
3220 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003221
3222
3223email-alert myhostname <hostname>
3224 Declare the to hostname address to be used when communicating with
3225 mailers.
3226 May be used in sections: defaults | frontend | listen | backend
3227 yes | yes | yes | yes
3228
3229 Arguments :
3230
Baptiste Assmann738bad92015-12-21 15:27:53 +01003231 <hostname> is the hostname to use when communicating with mailers
Simon Horman51a1cf62015-02-03 13:00:44 +09003232
3233 By default the systems hostname is used.
3234
3235 Also requires "email-alert from", "email-alert mailers" and
3236 "email-alert to" to be set and if so sending email alerts is enabled
3237 for the proxy.
3238
Simon Horman64e34162015-02-06 11:11:57 +09003239 See also : "email-alert from", "email-alert level", "email-alert mailers",
3240 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003241
3242
3243email-alert to <emailaddr>
3244 Declare both the recipent address in the envelope and to address in the
3245 header of email alerts. This is the address that email alerts are sent to.
3246 May be used in sections: defaults | frontend | listen | backend
3247 yes | yes | yes | yes
3248
3249 Arguments :
3250
3251 <emailaddr> is the to email address to use when sending email alerts
3252
3253 Also requires "email-alert mailers" and "email-alert to" to be set
3254 and if so sending email alerts is enabled for the proxy.
3255
Simon Horman64e34162015-02-06 11:11:57 +09003256 See also : "email-alert from", "email-alert level", "email-alert mailers",
Simon Horman51a1cf62015-02-03 13:00:44 +09003257 "email-alert myhostname", section 3.6 about mailers.
3258
3259
Willy Tarreau4de91492010-01-22 19:10:05 +01003260force-persist { if | unless } <condition>
3261 Declare a condition to force persistence on down servers
3262 May be used in sections: defaults | frontend | listen | backend
3263 no | yes | yes | yes
3264
3265 By default, requests are not dispatched to down servers. It is possible to
3266 force this using "option persist", but it is unconditional and redispatches
3267 to a valid server if "option redispatch" is set. That leaves with very little
3268 possibilities to force some requests to reach a server which is artificially
3269 marked down for maintenance operations.
3270
3271 The "force-persist" statement allows one to declare various ACL-based
3272 conditions which, when met, will cause a request to ignore the down status of
3273 a server and still try to connect to it. That makes it possible to start a
3274 server, still replying an error to the health checks, and run a specially
3275 configured browser to test the service. Among the handy methods, one could
3276 use a specific source IP address, or a specific cookie. The cookie also has
3277 the advantage that it can easily be added/removed on the browser from a test
3278 page. Once the service is validated, it is then possible to open the service
3279 to the world by returning a valid response to health checks.
3280
3281 The forced persistence is enabled when an "if" condition is met, or unless an
3282 "unless" condition is met. The final redispatch is always disabled when this
3283 is used.
3284
Cyril Bonté0d4bf012010-04-25 23:21:46 +02003285 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02003286 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01003287
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003288
3289filter <name> [param*]
3290 Add the filter <name> in the filter list attached to the proxy.
3291 May be used in sections : defaults | frontend | listen | backend
3292 no | yes | yes | yes
3293 Arguments :
3294 <name> is the name of the filter. Officially supported filters are
3295 referenced in section 9.
3296
3297 <param*> is a list of parameters accpeted by the filter <name>. The
3298 parsing of these parameters are the responsibility of the
3299 filter. Please refer to the documention of the corresponding
3300 filter (section 9) from all details on the supported parameters.
3301
3302 Multiple occurrences of the filter line can be used for the same proxy. The
3303 same filter can be referenced many times if needed.
3304
3305 Example:
3306 listen
3307 bind *:80
3308
3309 filter trace name BEFORE-HTTP-COMP
3310 filter compression
3311 filter trace name AFTER-HTTP-COMP
3312
3313 compression algo gzip
3314 compression offload
3315
3316 server srv1 192.168.0.1:80
3317
3318 See also : section 9.
3319
Willy Tarreau4de91492010-01-22 19:10:05 +01003320
Willy Tarreau2769aa02007-12-27 18:26:09 +01003321fullconn <conns>
3322 Specify at what backend load the servers will reach their maxconn
3323 May be used in sections : defaults | frontend | listen | backend
3324 yes | no | yes | yes
3325 Arguments :
3326 <conns> is the number of connections on the backend which will make the
3327 servers use the maximal number of connections.
3328
Willy Tarreau198a7442008-01-17 12:05:32 +01003329 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01003330 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01003331 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01003332 load. The server will then always accept at least <minconn> connections,
3333 never more than <maxconn>, and the limit will be on the ramp between both
3334 values when the backend has less than <conns> concurrent connections. This
3335 makes it possible to limit the load on the servers during normal loads, but
3336 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003337 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003338
Willy Tarreaufbb78422011-06-05 15:38:35 +02003339 Since it's hard to get this value right, haproxy automatically sets it to
3340 10% of the sum of the maxconns of all frontends that may branch to this
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01003341 backend (based on "use_backend" and "default_backend" rules). That way it's
3342 safe to leave it unset. However, "use_backend" involving dynamic names are
3343 not counted since there is no way to know if they could match or not.
Willy Tarreaufbb78422011-06-05 15:38:35 +02003344
Willy Tarreau2769aa02007-12-27 18:26:09 +01003345 Example :
3346 # The servers will accept between 100 and 1000 concurrent connections each
3347 # and the maximum of 1000 will be reached when the backend reaches 10000
3348 # connections.
3349 backend dynamic
3350 fullconn 10000
3351 server srv1 dyn1:80 minconn 100 maxconn 1000
3352 server srv2 dyn2:80 minconn 100 maxconn 1000
3353
3354 See also : "maxconn", "server"
3355
3356
3357grace <time>
3358 Maintain a proxy operational for some time after a soft stop
3359 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01003360 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01003361 Arguments :
3362 <time> is the time (by default in milliseconds) for which the instance
3363 will remain operational with the frontend sockets still listening
3364 when a soft-stop is received via the SIGUSR1 signal.
3365
3366 This may be used to ensure that the services disappear in a certain order.
3367 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003368 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01003369 needed by the equipment to detect the failure.
3370
3371 Note that currently, there is very little benefit in using this parameter,
3372 and it may in fact complicate the soft-reconfiguration process more than
3373 simplify it.
3374
Willy Tarreau0ba27502007-12-24 16:55:16 +01003375
Andrew Rodland17be45e2016-10-25 17:04:12 -04003376hash-balance-factor <factor>
3377 Specify the balancing factor for bounded-load consistent hashing
3378 May be used in sections : defaults | frontend | listen | backend
3379 yes | no | no | yes
3380 Arguments :
3381 <factor> is the control for the maximum number of concurrent requests to
3382 send to a server, expressed as a percentage of the average number
3383 of concurrent requests across all of the active servers.
3384
3385 Specifying a "hash-balance-factor" for a server with "hash-type consistent"
3386 enables an algorithm that prevents any one server from getting too many
3387 requests at once, even if some hash buckets receive many more requests than
3388 others. Setting <factor> to 0 (the default) disables the feature. Otherwise,
3389 <factor> is a percentage greater than 100. For example, if <factor> is 150,
3390 then no server will be allowed to have a load more than 1.5 times the average.
3391 If server weights are used, they will be respected.
3392
3393 If the first-choice server is disqualified, the algorithm will choose another
3394 server based on the request hash, until a server with additional capacity is
3395 found. A higher <factor> allows more imbalance between the servers, while a
3396 lower <factor> means that more servers will be checked on average, affecting
3397 performance. Reasonable values are from 125 to 200.
3398
3399 See also : "balance" and "hash-type".
3400
3401
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003402hash-type <method> <function> <modifier>
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003403 Specify a method to use for mapping hashes to servers
3404 May be used in sections : defaults | frontend | listen | backend
3405 yes | no | yes | yes
3406 Arguments :
Bhaskar98634f02013-10-29 23:30:51 -04003407 <method> is the method used to select a server from the hash computed by
3408 the <function> :
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003409
Bhaskar98634f02013-10-29 23:30:51 -04003410 map-based the hash table is a static array containing all alive servers.
3411 The hashes will be very smooth, will consider weights, but
3412 will be static in that weight changes while a server is up
3413 will be ignored. This means that there will be no slow start.
3414 Also, since a server is selected by its position in the array,
3415 most mappings are changed when the server count changes. This
3416 means that when a server goes up or down, or when a server is
3417 added to a farm, most connections will be redistributed to
3418 different servers. This can be inconvenient with caches for
3419 instance.
Willy Tarreau798a39c2010-11-24 15:04:29 +01003420
Bhaskar98634f02013-10-29 23:30:51 -04003421 consistent the hash table is a tree filled with many occurrences of each
3422 server. The hash key is looked up in the tree and the closest
3423 server is chosen. This hash is dynamic, it supports changing
3424 weights while the servers are up, so it is compatible with the
3425 slow start feature. It has the advantage that when a server
3426 goes up or down, only its associations are moved. When a
3427 server is added to the farm, only a few part of the mappings
3428 are redistributed, making it an ideal method for caches.
3429 However, due to its principle, the distribution will never be
3430 very smooth and it may sometimes be necessary to adjust a
3431 server's weight or its ID to get a more balanced distribution.
3432 In order to get the same distribution on multiple load
3433 balancers, it is important that all servers have the exact
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003434 same IDs. Note: consistent hash uses sdbm and avalanche if no
3435 hash function is specified.
Bhaskar98634f02013-10-29 23:30:51 -04003436
3437 <function> is the hash function to be used :
3438
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003439 sdbm this function was created initially for sdbm (a public-domain
Bhaskar98634f02013-10-29 23:30:51 -04003440 reimplementation of ndbm) database library. It was found to do
3441 well in scrambling bits, causing better distribution of the keys
3442 and fewer splits. It also happens to be a good general hashing
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003443 function with good distribution, unless the total server weight
3444 is a multiple of 64, in which case applying the avalanche
3445 modifier may help.
Bhaskar98634f02013-10-29 23:30:51 -04003446
3447 djb2 this function was first proposed by Dan Bernstein many years ago
3448 on comp.lang.c. Studies have shown that for certain workload this
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003449 function provides a better distribution than sdbm. It generally
3450 works well with text-based inputs though it can perform extremely
3451 poorly with numeric-only input or when the total server weight is
3452 a multiple of 33, unless the avalanche modifier is also used.
3453
Willy Tarreaua0f42712013-11-14 14:30:35 +01003454 wt6 this function was designed for haproxy while testing other
3455 functions in the past. It is not as smooth as the other ones, but
3456 is much less sensible to the input data set or to the number of
3457 servers. It can make sense as an alternative to sdbm+avalanche or
3458 djb2+avalanche for consistent hashing or when hashing on numeric
3459 data such as a source IP address or a visitor identifier in a URL
3460 parameter.
3461
Willy Tarreau324f07f2015-01-20 19:44:50 +01003462 crc32 this is the most common CRC32 implementation as used in Ethernet,
3463 gzip, PNG, etc. It is slower than the other ones but may provide
3464 a better distribution or less predictable results especially when
3465 used on strings.
3466
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003467 <modifier> indicates an optional method applied after hashing the key :
3468
3469 avalanche This directive indicates that the result from the hash
3470 function above should not be used in its raw form but that
3471 a 4-byte full avalanche hash must be applied first. The
3472 purpose of this step is to mix the resulting bits from the
3473 previous hash in order to avoid any undesired effect when
3474 the input contains some limited values or when the number of
3475 servers is a multiple of one of the hash's components (64
3476 for SDBM, 33 for DJB2). Enabling avalanche tends to make the
3477 result less predictable, but it's also not as smooth as when
3478 using the original function. Some testing might be needed
3479 with some workloads. This hash is one of the many proposed
3480 by Bob Jenkins.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003481
Bhaskar98634f02013-10-29 23:30:51 -04003482 The default hash type is "map-based" and is recommended for most usages. The
3483 default function is "sdbm", the selection of a function should be based on
3484 the range of the values being hashed.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003485
Andrew Rodland17be45e2016-10-25 17:04:12 -04003486 See also : "balance", "hash-balance-factor", "server"
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003487
3488
Willy Tarreau0ba27502007-12-24 16:55:16 +01003489http-check disable-on-404
3490 Enable a maintenance mode upon HTTP/404 response to health-checks
3491 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01003492 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01003493 Arguments : none
3494
3495 When this option is set, a server which returns an HTTP code 404 will be
3496 excluded from further load-balancing, but will still receive persistent
3497 connections. This provides a very convenient method for Web administrators
3498 to perform a graceful shutdown of their servers. It is also important to note
3499 that a server which is detected as failed while it was in this mode will not
3500 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
3501 will immediately be reinserted into the farm. The status on the stats page
3502 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01003503 option only works in conjunction with the "httpchk" option. If this option
3504 is used with "http-check expect", then it has precedence over it so that 404
3505 responses will still be considered as soft-stop.
3506
3507 See also : "option httpchk", "http-check expect"
3508
3509
3510http-check expect [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003511 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01003512 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02003513 yes | no | yes | yes
Willy Tarreaubd741542010-03-16 18:46:54 +01003514 Arguments :
3515 <match> is a keyword indicating how to look for a specific pattern in the
3516 response. The keyword may be one of "status", "rstatus",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003517 "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01003518 exclamation mark ("!") to negate the match. Spaces are allowed
3519 between the exclamation mark and the keyword. See below for more
3520 details on the supported keywords.
3521
3522 <pattern> is the pattern to look for. It may be a string or a regular
3523 expression. If the pattern contains spaces, they must be escaped
3524 with the usual backslash ('\').
3525
3526 By default, "option httpchk" considers that response statuses 2xx and 3xx
3527 are valid, and that others are invalid. When "http-check expect" is used,
3528 it defines what is considered valid or invalid. Only one "http-check"
3529 statement is supported in a backend. If a server fails to respond or times
3530 out, the check obviously fails. The available matches are :
3531
3532 status <string> : test the exact string match for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003533 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003534 response's status code is exactly this string. If the
3535 "status" keyword is prefixed with "!", then the response
3536 will be considered invalid if the status code matches.
3537
3538 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003539 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003540 response's status code matches the expression. If the
3541 "rstatus" keyword is prefixed with "!", then the response
3542 will be considered invalid if the status code matches.
3543 This is mostly used to check for multiple codes.
3544
3545 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003546 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003547 response's body contains this exact string. If the
3548 "string" keyword is prefixed with "!", then the response
3549 will be considered invalid if the body contains this
3550 string. This can be used to look for a mandatory word at
3551 the end of a dynamic page, or to detect a failure when a
3552 specific error appears on the check page (eg: a stack
3553 trace).
3554
3555 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003556 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003557 response's body matches this expression. If the "rstring"
3558 keyword is prefixed with "!", then the response will be
3559 considered invalid if the body matches the expression.
3560 This can be used to look for a mandatory word at the end
3561 of a dynamic page, or to detect a failure when a specific
3562 error appears on the check page (eg: a stack trace).
3563
3564 It is important to note that the responses will be limited to a certain size
3565 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
3566 Thus, too large responses may not contain the mandatory pattern when using
3567 "string" or "rstring". If a large response is absolutely required, it is
3568 possible to change the default max size by setting the global variable.
3569 However, it is worth keeping in mind that parsing very large responses can
3570 waste some CPU cycles, especially when regular expressions are used, and that
3571 it is always better to focus the checks on smaller resources.
3572
Cyril Bonté32602d22015-01-30 00:07:07 +01003573 Also "http-check expect" doesn't support HTTP keep-alive. Keep in mind that it
3574 will automatically append a "Connection: close" header, meaning that this
3575 header should not be present in the request provided by "option httpchk".
3576
Willy Tarreaubd741542010-03-16 18:46:54 +01003577 Last, if "http-check expect" is combined with "http-check disable-on-404",
3578 then this last one has precedence when the server responds with 404.
3579
3580 Examples :
3581 # only accept status 200 as valid
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003582 http-check expect status 200
Willy Tarreaubd741542010-03-16 18:46:54 +01003583
3584 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003585 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01003586
3587 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003588 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01003589
3590 # check that we have a correct hexadecimal tag before /html
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003591 http-check expect rstring <!--tag:[0-9a-f]*</html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01003592
Willy Tarreaubd741542010-03-16 18:46:54 +01003593 See also : "option httpchk", "http-check disable-on-404"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003594
3595
Willy Tarreauef781042010-01-27 11:53:01 +01003596http-check send-state
3597 Enable emission of a state header with HTTP health checks
3598 May be used in sections : defaults | frontend | listen | backend
3599 yes | no | yes | yes
3600 Arguments : none
3601
3602 When this option is set, haproxy will systematically send a special header
3603 "X-Haproxy-Server-State" with a list of parameters indicating to each server
3604 how they are seen by haproxy. This can be used for instance when a server is
3605 manipulated without access to haproxy and the operator needs to know whether
3606 haproxy still sees it up or not, or if the server is the last one in a farm.
3607
3608 The header is composed of fields delimited by semi-colons, the first of which
3609 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
3610 checks on the total number before transition, just as appears in the stats
3611 interface. Next headers are in the form "<variable>=<value>", indicating in
3612 no specific order some values available in the stats interface :
Joseph Lynch514061c2015-01-15 17:52:59 -08003613 - a variable "address", containing the address of the backend server.
3614 This corresponds to the <address> field in the server declaration. For
3615 unix domain sockets, it will read "unix".
3616
3617 - a variable "port", containing the port of the backend server. This
3618 corresponds to the <port> field in the server declaration. For unix
3619 domain sockets, it will read "unix".
3620
Willy Tarreauef781042010-01-27 11:53:01 +01003621 - a variable "name", containing the name of the backend followed by a slash
3622 ("/") then the name of the server. This can be used when a server is
3623 checked in multiple backends.
3624
3625 - a variable "node" containing the name of the haproxy node, as set in the
3626 global "node" variable, otherwise the system's hostname if unspecified.
3627
3628 - a variable "weight" indicating the weight of the server, a slash ("/")
3629 and the total weight of the farm (just counting usable servers). This
3630 helps to know if other servers are available to handle the load when this
3631 one fails.
3632
3633 - a variable "scur" indicating the current number of concurrent connections
3634 on the server, followed by a slash ("/") then the total number of
3635 connections on all servers of the same backend.
3636
3637 - a variable "qcur" indicating the current number of requests in the
3638 server's queue.
3639
3640 Example of a header received by the application server :
3641 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
3642 scur=13/22; qcur=0
3643
3644 See also : "option httpchk", "http-check disable-on-404"
3645
Willy Tarreaube1d34d2016-06-26 19:37:59 +02003646http-request { allow | tarpit | auth [realm <realm>] | redirect <rule> |
3647 deny [deny_status <status>] |
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003648 add-header <name> <fmt> | set-header <name> <fmt> |
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003649 capture <sample> [ len <length> | id <id> ] |
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003650 del-header <name> | set-nice <nice> | set-log-level <level> |
Sasha Pachev218f0642014-06-16 12:05:59 -06003651 replace-header <name> <match-regex> <replace-fmt> |
3652 replace-value <name> <match-regex> <replace-fmt> |
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01003653 set-method <fmt> | set-path <fmt> | set-query <fmt> |
3654 set-uri <fmt> | set-tos <tos> | set-mark <mark> |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003655 add-acl(<file name>) <key fmt> |
3656 del-acl(<file name>) <key fmt> |
3657 del-map(<file name>) <key fmt> |
Baptiste Assmannbb7e86a2014-09-03 18:29:47 +02003658 set-map(<file name>) <key fmt> <value fmt> |
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003659 set-var(<var name>) <expr> |
Christopher Faulet85d79c92016-11-09 16:54:56 +01003660 unset-var(<var name>) |
Thierry FOURNIER90da1912015-03-05 11:17:06 +01003661 { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] |
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02003662 sc-inc-gpc0(<sc-id>) |
Thierry FOURNIER236657b2015-08-19 08:25:14 +02003663 sc-set-gpt0(<sc-id>) <int> |
Willy Tarreau2d392c22015-08-24 01:43:45 +02003664 silent-drop |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003665 }
Cyril Bontéf0c60612010-02-06 14:44:47 +01003666 [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003667 Access control for Layer 7 requests
3668
3669 May be used in sections: defaults | frontend | listen | backend
3670 no | yes | yes | yes
3671
Willy Tarreau20b0de52012-12-24 15:45:22 +01003672 The http-request statement defines a set of rules which apply to layer 7
3673 processing. The rules are evaluated in their declaration order when they are
3674 met in a frontend, listen or backend section. Any rule may optionally be
3675 followed by an ACL-based condition, in which case it will only be evaluated
3676 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003677
Willy Tarreau20b0de52012-12-24 15:45:22 +01003678 The first keyword is the rule's action. Currently supported actions include :
3679 - "allow" : this stops the evaluation of the rules and lets the request
3680 pass the check. No further "http-request" rules are evaluated.
3681
3682 - "deny" : this stops the evaluation of the rules and immediately rejects
Willy Tarreaube1d34d2016-06-26 19:37:59 +02003683 the request and emits an HTTP 403 error, or optionally the status code
3684 specified as an argument to "deny_status". The list of permitted status
3685 codes is limited to those that can be overridden by the "errorfile"
3686 directive. No further "http-request" rules are evaluated.
Willy Tarreau20b0de52012-12-24 15:45:22 +01003687
Willy Tarreauccbcc372012-12-27 12:37:57 +01003688 - "tarpit" : this stops the evaluation of the rules and immediately blocks
3689 the request without responding for a delay specified by "timeout tarpit"
3690 or "timeout connect" if the former is not set. After that delay, if the
3691 client is still connected, an HTTP error 500 is returned so that the
3692 client does not suspect it has been tarpitted. Logs will report the flags
3693 "PT". The goal of the tarpit rule is to slow down robots during an attack
3694 when they're limited on the number of concurrent requests. It can be very
3695 efficient against very dumb robots, and will significantly reduce the
3696 load on firewalls compared to a "deny" rule. But when facing "correctly"
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003697 developed robots, it can make things worse by forcing haproxy and the
Willy Tarreau2d392c22015-08-24 01:43:45 +02003698 front firewall to support insane number of concurrent connections. See
3699 also the "silent-drop" action below.
Willy Tarreauccbcc372012-12-27 12:37:57 +01003700
Willy Tarreau20b0de52012-12-24 15:45:22 +01003701 - "auth" : this stops the evaluation of the rules and immediately responds
3702 with an HTTP 401 or 407 error code to invite the user to present a valid
3703 user name and password. No further "http-request" rules are evaluated. An
3704 optional "realm" parameter is supported, it sets the authentication realm
3705 that is returned with the response (typically the application's name).
3706
Willy Tarreau81499eb2012-12-27 12:19:02 +01003707 - "redirect" : this performs an HTTP redirection based on a redirect rule.
3708 This is exactly the same as the "redirect" statement except that it
3709 inserts a redirect rule which can be processed in the middle of other
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01003710 "http-request" rules and that these rules use the "log-format" strings.
3711 See the "redirect" keyword for the rule's syntax.
Willy Tarreau81499eb2012-12-27 12:19:02 +01003712
Willy Tarreau20b0de52012-12-24 15:45:22 +01003713 - "add-header" appends an HTTP header field whose name is specified in
3714 <name> and whose value is defined by <fmt> which follows the log-format
3715 rules (see Custom Log Format in section 8.2.4). This is particularly
3716 useful to pass connection-specific information to the server (eg: the
3717 client's SSL certificate), or to combine several headers into one. This
3718 rule is not final, so it is possible to add other similar rules. Note
3719 that header addition is performed immediately, so one rule might reuse
3720 the resulting header from a previous rule.
3721
3722 - "set-header" does the same as "add-header" except that the header name
3723 is first removed if it existed. This is useful when passing security
3724 information to the server, where the header must not be manipulated by
Willy Tarreau85603282015-01-21 20:39:27 +01003725 external users. Note that the new value is computed before the removal so
3726 it is possible to concatenate a value to an existing header.
Willy Tarreau20b0de52012-12-24 15:45:22 +01003727
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003728 - "del-header" removes all HTTP header fields whose name is specified in
3729 <name>.
3730
Sasha Pachev218f0642014-06-16 12:05:59 -06003731 - "replace-header" matches the regular expression in all occurrences of
3732 header field <name> according to <match-regex>, and replaces them with
3733 the <replace-fmt> argument. Format characters are allowed in replace-fmt
3734 and work like in <fmt> arguments in "add-header". The match is only
3735 case-sensitive. It is important to understand that this action only
3736 considers whole header lines, regardless of the number of values they
3737 may contain. This usage is suited to headers naturally containing commas
3738 in their value, such as If-Modified-Since and so on.
3739
3740 Example:
3741
3742 http-request replace-header Cookie foo=([^;]*);(.*) foo=\1;ip=%bi;\2
3743
3744 applied to:
3745
3746 Cookie: foo=foobar; expires=Tue, 14-Jun-2016 01:40:45 GMT;
3747
3748 outputs:
3749
3750 Cookie: foo=foobar;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT;
3751
3752 assuming the backend IP is 192.168.1.20
3753
3754 - "replace-value" works like "replace-header" except that it matches the
3755 regex against every comma-delimited value of the header field <name>
3756 instead of the entire header. This is suited for all headers which are
3757 allowed to carry more than one value. An example could be the Accept
3758 header.
3759
3760 Example:
3761
3762 http-request replace-value X-Forwarded-For ^192\.168\.(.*)$ 172.16.\1
3763
3764 applied to:
3765
3766 X-Forwarded-For: 192.168.10.1, 192.168.13.24, 10.0.0.37
3767
3768 outputs:
3769
3770 X-Forwarded-For: 172.16.10.1, 172.16.13.24, 10.0.0.37
3771
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01003772 - "set-method" rewrites the request method with the result of the
3773 evaluation of format string <fmt>. There should be very few valid reasons
3774 for having to do so as this is more likely to break something than to fix
3775 it.
3776
3777 - "set-path" rewrites the request path with the result of the evaluation of
3778 format string <fmt>. The query string, if any, is left intact. If a
3779 scheme and authority is found before the path, they are left intact as
3780 well. If the request doesn't have a path ("*"), this one is replaced with
3781 the format. This can be used to prepend a directory component in front of
3782 a path for example. See also "set-query" and "set-uri".
3783
3784 Example :
3785 # prepend the host name before the path
3786 http-request set-path /%[hdr(host)]%[path]
3787
3788 - "set-query" rewrites the request's query string which appears after the
3789 first question mark ("?") with the result of the evaluation of format
3790 string <fmt>. The part prior to the question mark is left intact. If the
3791 request doesn't contain a question mark and the new value is not empty,
3792 then one is added at the end of the URI, followed by the new value. If
3793 a question mark was present, it will never be removed even if the value
3794 is empty. This can be used to add or remove parameters from the query
3795 string. See also "set-query" and "set-uri".
3796
3797 Example :
3798 # replace "%3D" with "=" in the query string
3799 http-request set-query %[query,regsub(%3D,=,g)]
3800
3801 - "set-uri" rewrites the request URI with the result of the evaluation of
3802 format string <fmt>. The scheme, authority, path and query string are all
3803 replaced at once. This can be used to rewrite hosts in front of proxies,
3804 or to perform complex modifications to the URI such as moving parts
3805 between the path and the query string. See also "set-path" and
3806 "set-query".
3807
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003808 - "set-nice" sets the "nice" factor of the current request being processed.
3809 It only has effect against the other requests being processed at the same
3810 time. The default value is 0, unless altered by the "nice" setting on the
3811 "bind" line. The accepted range is -1024..1024. The higher the value, the
3812 nicest the request will be. Lower values will make the request more
3813 important than other ones. This can be useful to improve the speed of
3814 some requests, or lower the priority of non-important requests. Using
3815 this setting without prior experimentation can cause some major slowdown.
3816
Willy Tarreau9a355ec2013-06-11 17:45:46 +02003817 - "set-log-level" is used to change the log level of the current request
3818 when a certain condition is met. Valid levels are the 8 syslog levels
3819 (see the "log" keyword) plus the special level "silent" which disables
3820 logging for this request. This rule is not final so the last matching
3821 rule wins. This rule can be useful to disable health checks coming from
3822 another equipment.
3823
Willy Tarreau42cf39e2013-06-11 18:51:32 +02003824 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
3825 the client to the value passed in <tos> on platforms which support this.
3826 This value represents the whole 8 bits of the IP TOS field, and can be
3827 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
3828 that only the 6 higher bits are used in DSCP or TOS, and the two lower
3829 bits are always 0. This can be used to adjust some routing behaviour on
3830 border routers based on some information from the request. See RFC 2474,
3831 2597, 3260 and 4594 for more information.
3832
Willy Tarreau51347ed2013-06-11 19:34:13 +02003833 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
3834 client to the value passed in <mark> on platforms which support it. This
3835 value is an unsigned 32 bit value which can be matched by netfilter and
3836 by the routing table. It can be expressed both in decimal or hexadecimal
3837 format (prefixed by "0x"). This can be useful to force certain packets to
3838 take a different route (for example a cheaper network path for bulk
3839 downloads). This works on Linux kernels 2.6.32 and above and requires
3840 admin privileges.
3841
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003842 - "add-acl" is used to add a new entry into an ACL. The ACL must be loaded
3843 from a file (even a dummy empty file). The file name of the ACL to be
3844 updated is passed between parentheses. It takes one argument: <key fmt>,
3845 which follows log-format rules, to collect content of the new entry. It
3846 performs a lookup in the ACL before insertion, to avoid duplicated (or
3847 more) values. This lookup is done by a linear search and can be expensive
3848 with large lists! It is the equivalent of the "add acl" command from the
3849 stats socket, but can be triggered by an HTTP request.
3850
3851 - "del-acl" is used to delete an entry from an ACL. The ACL must be loaded
3852 from a file (even a dummy empty file). The file name of the ACL to be
3853 updated is passed between parentheses. It takes one argument: <key fmt>,
3854 which follows log-format rules, to collect content of the entry to delete.
3855 It is the equivalent of the "del acl" command from the stats socket, but
3856 can be triggered by an HTTP request.
3857
3858 - "del-map" is used to delete an entry from a MAP. The MAP must be loaded
3859 from a file (even a dummy empty file). The file name of the MAP to be
3860 updated is passed between parentheses. It takes one argument: <key fmt>,
3861 which follows log-format rules, to collect content of the entry to delete.
3862 It takes one argument: "file name" It is the equivalent of the "del map"
3863 command from the stats socket, but can be triggered by an HTTP request.
3864
3865 - "set-map" is used to add a new entry into a MAP. The MAP must be loaded
3866 from a file (even a dummy empty file). The file name of the MAP to be
3867 updated is passed between parentheses. It takes 2 arguments: <key fmt>,
3868 which follows log-format rules, used to collect MAP key, and <value fmt>,
3869 which follows log-format rules, used to collect content for the new entry.
3870 It performs a lookup in the MAP before insertion, to avoid duplicated (or
3871 more) values. This lookup is done by a linear search and can be expensive
3872 with large lists! It is the equivalent of the "set map" command from the
3873 stats socket, but can be triggered by an HTTP request.
3874
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003875 - capture <sample> [ len <length> | id <id> ] :
Willy Tarreaua9083d02015-05-08 15:27:59 +02003876 captures sample expression <sample> from the request buffer, and converts
3877 it to a string of at most <len> characters. The resulting string is
3878 stored into the next request "capture" slot, so it will possibly appear
3879 next to some captured HTTP headers. It will then automatically appear in
3880 the logs, and it will be possible to extract it using sample fetch rules
3881 to feed it into headers or anything. The length should be limited given
3882 that this size will be allocated for each capture during the whole
3883 session life. Please check section 7.3 (Fetching samples) and "capture
3884 request header" for more information.
3885
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003886 If the keyword "id" is used instead of "len", the action tries to store
3887 the captured string in a previously declared capture slot. This is useful
3888 to run captures in backends. The slot id can be declared by a previous
3889 directive "http-request capture" or with the "declare capture" keyword.
Baptiste Assmanne9544932015-11-03 23:31:35 +01003890 If the slot <id> doesn't exist, then HAProxy fails parsing the
3891 configuration to prevent unexpected behavior at run time.
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003892
Willy Tarreau09448f72014-06-25 18:12:15 +02003893 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
3894 enables tracking of sticky counters from current request. These rules
3895 do not stop evaluation and do not change default action. Three sets of
3896 counters may be simultaneously tracked by the same connection. The first
3897 "track-sc0" rule executed enables tracking of the counters of the
3898 specified table as the first set. The first "track-sc1" rule executed
3899 enables tracking of the counters of the specified table as the second
3900 set. The first "track-sc2" rule executed enables tracking of the
3901 counters of the specified table as the third set. It is a recommended
3902 practice to use the first set of counters for the per-frontend counters
3903 and the second set for the per-backend ones. But this is just a
3904 guideline, all may be used everywhere.
3905
3906 These actions take one or two arguments :
3907 <key> is mandatory, and is a sample expression rule as described
3908 in section 7.3. It describes what elements of the incoming
3909 request or connection will be analysed, extracted, combined,
3910 and used to select which table entry to update the counters.
3911
3912 <table> is an optional table to be used instead of the default one,
3913 which is the stick-table declared in the current proxy. All
3914 the counters for the matches and updates for the key will
3915 then be performed in that table until the session ends.
3916
3917 Once a "track-sc*" rule is executed, the key is looked up in the table
3918 and if it is not found, an entry is allocated for it. Then a pointer to
3919 that entry is kept during all the session's life, and this entry's
3920 counters are updated as often as possible, every time the session's
3921 counters are updated, and also systematically when the session ends.
3922 Counters are only updated for events that happen after the tracking has
3923 been started. As an exception, connection counters and request counters
3924 are systematically updated so that they reflect useful information.
3925
3926 If the entry tracks concurrent connection counters, one connection is
3927 counted for as long as the entry is tracked, and the entry will not
3928 expire during that time. Tracking counters also provides a performance
3929 advantage over just checking the keys, because only one table lookup is
3930 performed for all ACL checks that make use of it.
3931
Thierry FOURNIER236657b2015-08-19 08:25:14 +02003932 - sc-set-gpt0(<sc-id>) <int> :
3933 This action sets the GPT0 tag according to the sticky counter designated
3934 by <sc-id> and the value of <int>. The expected result is a boolean. If
3935 an error occurs, this action silently fails and the actions evaluation
3936 continues.
3937
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02003938 - sc-inc-gpc0(<sc-id>):
3939 This action increments the GPC0 counter according with the sticky counter
3940 designated by <sc-id>. If an error occurs, this action silently fails and
3941 the actions evaluation continues.
3942
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003943 - set-var(<var-name>) <expr> :
3944 Is used to set the contents of a variable. The variable is declared
3945 inline.
3946
Daniel Schneller0b547052016-03-21 20:46:57 +01003947 <var-name> The name of the variable starts with an indication about
3948 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +01003949 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +01003950 "sess" : the variable is shared with the whole session
3951 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003952 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +01003953 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003954 processing
Daniel Schneller0b547052016-03-21 20:46:57 +01003955 "res" : the variable is shared only during response
3956 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003957 This prefix is followed by a name. The separator is a '.'.
Daniel Schneller0b547052016-03-21 20:46:57 +01003958 The name may only contain characters 'a-z', 'A-Z', '0-9'
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003959 and '_'.
3960
3961 <expr> Is a standard HAProxy expression formed by a sample-fetch
3962 followed by some converters.
3963
3964 Example:
3965
3966 http-request set-var(req.my_var) req.fhdr(user-agent),lower
3967
Christopher Faulet85d79c92016-11-09 16:54:56 +01003968 - unset-var(<var-name>) :
3969 Is used to unset a variable. See above for details about <var-name>.
3970
3971 Example:
3972
3973 http-request unset-var(req.my_var)
3974
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02003975 - set-src <expr> :
3976 Is used to set the source IP address to the value of specified
3977 expression. Useful when a proxy in front of HAProxy rewrites source IP,
3978 but provides the correct IP in a HTTP header; or you want to mask
3979 source IP for privacy.
3980
3981 <expr> Is a standard HAProxy expression formed by a sample-fetch
3982 followed by some converters.
3983
3984 Example:
3985
3986 http-request set-src hdr(x-forwarded-for)
3987 http-request set-src src,ipmask(24)
3988
Willy Tarreau00005ce2016-10-21 15:07:45 +02003989 When possible, set-src preserves the original source port as long as the
3990 address family allows it, otherwise the source port is set to 0.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02003991
William Lallemand44be6402016-05-25 01:51:35 +02003992 - set-src-port <expr> :
3993 Is used to set the source port address to the value of specified
3994 expression.
3995
3996 <expr> Is a standard HAProxy expression formed by a sample-fetch
3997 followed by some converters.
3998
3999 Example:
4000
4001 http-request set-src-port hdr(x-port)
4002 http-request set-src-port int(4000)
4003
Willy Tarreau00005ce2016-10-21 15:07:45 +02004004 When possible, set-src-port preserves the original source address as long
4005 as the address family supports a port, otherwise it forces the source
4006 address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +02004007
William Lallemand13e9b0c2016-05-25 02:34:07 +02004008 - set-dst <expr> :
4009 Is used to set the destination IP address to the value of specified
4010 expression. Useful when a proxy in front of HAProxy rewrites destination
4011 IP, but provides the correct IP in a HTTP header; or you want to mask
4012 the IP for privacy. If you want to connect to the new address/port, use
4013 '0.0.0.0:0' as a server address in the backend.
4014
4015 <expr> Is a standard HAProxy expression formed by a sample-fetch
4016 followed by some converters.
4017
4018 Example:
4019
4020 http-request set-dst hdr(x-dst)
4021 http-request set-dst dst,ipmask(24)
4022
Willy Tarreau00005ce2016-10-21 15:07:45 +02004023 When possible, set-dst preserves the original destination port as long as
4024 the address family allows it, otherwise the destination port is set to 0.
4025
William Lallemand13e9b0c2016-05-25 02:34:07 +02004026 - set-dst-port <expr> :
4027 Is used to set the destination port address to the value of specified
4028 expression. If you want to connect to the new address/port, use
4029 '0.0.0.0:0' as a server address in the backend.
4030
4031 <expr> Is a standard HAProxy expression formed by a sample-fetch
4032 followed by some converters.
4033
4034 Example:
4035
4036 http-request set-dst-port hdr(x-port)
4037 http-request set-dst-port int(4000)
4038
Willy Tarreau00005ce2016-10-21 15:07:45 +02004039 When possible, set-dst-port preserves the original destination address as
4040 long as the address family supports a port, otherwise it forces the
4041 destination address to IPv4 "0.0.0.0" before rewriting the port.
4042
Willy Tarreau2d392c22015-08-24 01:43:45 +02004043 - "silent-drop" : this stops the evaluation of the rules and makes the
4044 client-facing connection suddenly disappear using a system-dependant way
4045 that tries to prevent the client from being notified. The effect it then
4046 that the client still sees an established connection while there's none
4047 on HAProxy. The purpose is to achieve a comparable effect to "tarpit"
4048 except that it doesn't use any local resource at all on the machine
4049 running HAProxy. It can resist much higher loads than "tarpit", and slow
4050 down stronger attackers. It is important to undestand the impact of using
4051 this mechanism. All stateful equipments placed between the client and
4052 HAProxy (firewalls, proxies, load balancers) will also keep the
4053 established connection for a long time and may suffer from this action.
4054 On modern Linux systems running with enough privileges, the TCP_REPAIR
4055 socket option is used to block the emission of a TCP reset. On other
4056 systems, the socket's TTL is reduced to 1 so that the TCP reset doesn't
4057 pass the first router, though it's still delivered to local networks. Do
4058 not use it unless you fully understand how it works.
4059
Willy Tarreau20b0de52012-12-24 15:45:22 +01004060 There is no limit to the number of http-request statements per instance.
4061
4062 It is important to know that http-request rules are processed very early in
4063 the HTTP processing, just after "block" rules and before "reqdel" or "reqrep"
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08004064 or "reqadd" rules. That way, headers added by "add-header"/"set-header" are
4065 visible by almost all further ACL rules.
4066
4067 Using "reqadd"/"reqdel"/"reqrep" to manipulate request headers is discouraged
4068 in newer versions (>= 1.5). But if you need to use regular expression to
4069 delete headers, you can still use "reqdel". Also please use
4070 "http-request deny/allow/tarpit" instead of "reqdeny"/"reqpass"/"reqtarpit".
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004071
4072 Example:
Cyril Bonté78caf842010-03-10 22:41:43 +01004073 acl nagios src 192.168.129.3
4074 acl local_net src 192.168.0.0/16
4075 acl auth_ok http_auth(L1)
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004076
Cyril Bonté78caf842010-03-10 22:41:43 +01004077 http-request allow if nagios
4078 http-request allow if local_net auth_ok
4079 http-request auth realm Gimme if local_net auth_ok
4080 http-request deny
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004081
Cyril Bonté78caf842010-03-10 22:41:43 +01004082 Example:
4083 acl auth_ok http_auth_group(L1) G1
Cyril Bonté78caf842010-03-10 22:41:43 +01004084 http-request auth unless auth_ok
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004085
Willy Tarreau20b0de52012-12-24 15:45:22 +01004086 Example:
4087 http-request set-header X-Haproxy-Current-Date %T
4088 http-request set-header X-SSL %[ssl_fc]
Willy Tarreaufca42612015-08-27 17:15:05 +02004089 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id,hex]
Willy Tarreau20b0de52012-12-24 15:45:22 +01004090 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
4091 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
4092 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
4093 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
4094 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
4095 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
4096
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004097 Example:
4098 acl key req.hdr(X-Add-Acl-Key) -m found
4099 acl add path /addacl
4100 acl del path /delacl
4101
4102 acl myhost hdr(Host) -f myhost.lst
4103
4104 http-request add-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key add
4105 http-request del-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key del
4106
4107 Example:
4108 acl value req.hdr(X-Value) -m found
4109 acl setmap path /setmap
4110 acl delmap path /delmap
4111
4112 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
4113
4114 http-request set-map(map.lst) %[src] %[req.hdr(X-Value)] if setmap value
4115 http-request del-map(map.lst) %[src] if delmap
4116
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02004117 See also : "stats http-request", section 3.4 about userlists and section 7
4118 about ACL usage.
Willy Tarreauef781042010-01-27 11:53:01 +01004119
Willy Tarreauf4c43c12013-06-11 17:01:13 +02004120http-response { allow | deny | add-header <name> <fmt> | set-nice <nice> |
Willy Tarreau51d861a2015-05-22 17:30:48 +02004121 capture <sample> id <id> | redirect <rule> |
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02004122 set-header <name> <fmt> | del-header <name> |
Sasha Pachev218f0642014-06-16 12:05:59 -06004123 replace-header <name> <regex-match> <replace-fmt> |
4124 replace-value <name> <regex-match> <replace-fmt> |
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004125 set-status <status> |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004126 set-log-level <level> | set-mark <mark> | set-tos <tos> |
4127 add-acl(<file name>) <key fmt> |
4128 del-acl(<file name>) <key fmt> |
4129 del-map(<file name>) <key fmt> |
Thierry FOURNIER90da1912015-03-05 11:17:06 +01004130 set-map(<file name>) <key fmt> <value fmt> |
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004131 set-var(<var-name>) <expr> |
Christopher Faulet85d79c92016-11-09 16:54:56 +01004132 unset-var(<var-name>) |
Ruoshan Huange4edc6b2016-07-14 15:07:45 +08004133 { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] |
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02004134 sc-inc-gpc0(<sc-id>) |
Thierry FOURNIER236657b2015-08-19 08:25:14 +02004135 sc-set-gpt0(<sc-id>) <int> |
Willy Tarreau2d392c22015-08-24 01:43:45 +02004136 silent-drop |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004137 }
Lukas Tribus2dd1d1a2013-06-19 23:34:41 +02004138 [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004139 Access control for Layer 7 responses
4140
4141 May be used in sections: defaults | frontend | listen | backend
4142 no | yes | yes | yes
4143
4144 The http-response statement defines a set of rules which apply to layer 7
4145 processing. The rules are evaluated in their declaration order when they are
4146 met in a frontend, listen or backend section. Any rule may optionally be
4147 followed by an ACL-based condition, in which case it will only be evaluated
4148 if the condition is true. Since these rules apply on responses, the backend
4149 rules are applied first, followed by the frontend's rules.
4150
4151 The first keyword is the rule's action. Currently supported actions include :
4152 - "allow" : this stops the evaluation of the rules and lets the response
4153 pass the check. No further "http-response" rules are evaluated for the
4154 current section.
4155
4156 - "deny" : this stops the evaluation of the rules and immediately rejects
4157 the response and emits an HTTP 502 error. No further "http-response"
4158 rules are evaluated.
4159
4160 - "add-header" appends an HTTP header field whose name is specified in
4161 <name> and whose value is defined by <fmt> which follows the log-format
4162 rules (see Custom Log Format in section 8.2.4). This may be used to send
4163 a cookie to a client for example, or to pass some internal information.
4164 This rule is not final, so it is possible to add other similar rules.
4165 Note that header addition is performed immediately, so one rule might
4166 reuse the resulting header from a previous rule.
4167
4168 - "set-header" does the same as "add-header" except that the header name
4169 is first removed if it existed. This is useful when passing security
4170 information to the server, where the header must not be manipulated by
4171 external users.
4172
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02004173 - "del-header" removes all HTTP header fields whose name is specified in
4174 <name>.
4175
Sasha Pachev218f0642014-06-16 12:05:59 -06004176 - "replace-header" matches the regular expression in all occurrences of
4177 header field <name> according to <match-regex>, and replaces them with
4178 the <replace-fmt> argument. Format characters are allowed in replace-fmt
4179 and work like in <fmt> arguments in "add-header". The match is only
4180 case-sensitive. It is important to understand that this action only
4181 considers whole header lines, regardless of the number of values they
4182 may contain. This usage is suited to headers naturally containing commas
4183 in their value, such as Set-Cookie, Expires and so on.
4184
4185 Example:
4186
4187 http-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
4188
4189 applied to:
4190
4191 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
4192
4193 outputs:
4194
4195 Set-Cookie: C=1;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT
4196
4197 assuming the backend IP is 192.168.1.20.
4198
4199 - "replace-value" works like "replace-header" except that it matches the
4200 regex against every comma-delimited value of the header field <name>
4201 instead of the entire header. This is suited for all headers which are
4202 allowed to carry more than one value. An example could be the Accept
4203 header.
4204
4205 Example:
4206
4207 http-response replace-value Cache-control ^public$ private
4208
4209 applied to:
4210
4211 Cache-Control: max-age=3600, public
4212
4213 outputs:
4214
4215 Cache-Control: max-age=3600, private
4216
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004217 - "set-status" replaces the response status code with <status> which must
4218 be an integer between 100 and 999. Note that the reason is automatically
4219 adapted to the new code.
4220
4221 Example:
4222
4223 # return "431 Request Header Fields Too Large"
4224 http-response set-status 431
4225
Willy Tarreauf4c43c12013-06-11 17:01:13 +02004226 - "set-nice" sets the "nice" factor of the current request being processed.
4227 It only has effect against the other requests being processed at the same
4228 time. The default value is 0, unless altered by the "nice" setting on the
4229 "bind" line. The accepted range is -1024..1024. The higher the value, the
4230 nicest the request will be. Lower values will make the request more
4231 important than other ones. This can be useful to improve the speed of
4232 some requests, or lower the priority of non-important requests. Using
4233 this setting without prior experimentation can cause some major slowdown.
4234
Willy Tarreau9a355ec2013-06-11 17:45:46 +02004235 - "set-log-level" is used to change the log level of the current request
4236 when a certain condition is met. Valid levels are the 8 syslog levels
4237 (see the "log" keyword) plus the special level "silent" which disables
4238 logging for this request. This rule is not final so the last matching
4239 rule wins. This rule can be useful to disable health checks coming from
4240 another equipment.
4241
Willy Tarreau42cf39e2013-06-11 18:51:32 +02004242 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
4243 the client to the value passed in <tos> on platforms which support this.
4244 This value represents the whole 8 bits of the IP TOS field, and can be
4245 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
4246 that only the 6 higher bits are used in DSCP or TOS, and the two lower
4247 bits are always 0. This can be used to adjust some routing behaviour on
4248 border routers based on some information from the request. See RFC 2474,
4249 2597, 3260 and 4594 for more information.
4250
Willy Tarreau51347ed2013-06-11 19:34:13 +02004251 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
4252 client to the value passed in <mark> on platforms which support it. This
4253 value is an unsigned 32 bit value which can be matched by netfilter and
4254 by the routing table. It can be expressed both in decimal or hexadecimal
4255 format (prefixed by "0x"). This can be useful to force certain packets to
4256 take a different route (for example a cheaper network path for bulk
4257 downloads). This works on Linux kernels 2.6.32 and above and requires
4258 admin privileges.
4259
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004260 - "add-acl" is used to add a new entry into an ACL. The ACL must be loaded
4261 from a file (even a dummy empty file). The file name of the ACL to be
4262 updated is passed between parentheses. It takes one argument: <key fmt>,
4263 which follows log-format rules, to collect content of the new entry. It
4264 performs a lookup in the ACL before insertion, to avoid duplicated (or
4265 more) values. This lookup is done by a linear search and can be expensive
4266 with large lists! It is the equivalent of the "add acl" command from the
4267 stats socket, but can be triggered by an HTTP response.
4268
4269 - "del-acl" is used to delete an entry from an ACL. The ACL must be loaded
4270 from a file (even a dummy empty file). The file name of the ACL to be
4271 updated is passed between parentheses. It takes one argument: <key fmt>,
4272 which follows log-format rules, to collect content of the entry to delete.
4273 It is the equivalent of the "del acl" command from the stats socket, but
4274 can be triggered by an HTTP response.
4275
4276 - "del-map" is used to delete an entry from a MAP. The MAP must be loaded
4277 from a file (even a dummy empty file). The file name of the MAP to be
4278 updated is passed between parentheses. It takes one argument: <key fmt>,
4279 which follows log-format rules, to collect content of the entry to delete.
4280 It takes one argument: "file name" It is the equivalent of the "del map"
4281 command from the stats socket, but can be triggered by an HTTP response.
4282
4283 - "set-map" is used to add a new entry into a MAP. The MAP must be loaded
4284 from a file (even a dummy empty file). The file name of the MAP to be
4285 updated is passed between parentheses. It takes 2 arguments: <key fmt>,
4286 which follows log-format rules, used to collect MAP key, and <value fmt>,
4287 which follows log-format rules, used to collect content for the new entry.
4288 It performs a lookup in the MAP before insertion, to avoid duplicated (or
4289 more) values. This lookup is done by a linear search and can be expensive
4290 with large lists! It is the equivalent of the "set map" command from the
4291 stats socket, but can be triggered by an HTTP response.
4292
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02004293 - capture <sample> id <id> :
4294 captures sample expression <sample> from the response buffer, and converts
4295 it to a string. The resulting string is stored into the next request
4296 "capture" slot, so it will possibly appear next to some captured HTTP
4297 headers. It will then automatically appear in the logs, and it will be
4298 possible to extract it using sample fetch rules to feed it into headers or
4299 anything. Please check section 7.3 (Fetching samples) and "capture
4300 response header" for more information.
4301
4302 The keyword "id" is the id of the capture slot which is used for storing
4303 the string. The capture slot must be defined in an associated frontend.
4304 This is useful to run captures in backends. The slot id can be declared by
4305 a previous directive "http-response capture" or with the "declare capture"
4306 keyword.
Baptiste Assmanne9544932015-11-03 23:31:35 +01004307 If the slot <id> doesn't exist, then HAProxy fails parsing the
4308 configuration to prevent unexpected behavior at run time.
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02004309
Willy Tarreau51d861a2015-05-22 17:30:48 +02004310 - "redirect" : this performs an HTTP redirection based on a redirect rule.
4311 This supports a format string similarly to "http-request redirect" rules,
4312 with the exception that only the "location" type of redirect is possible
4313 on the response. See the "redirect" keyword for the rule's syntax. When
4314 a redirect rule is applied during a response, connections to the server
4315 are closed so that no data can be forwarded from the server to the client.
4316
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004317 - set-var(<var-name>) expr:
4318 Is used to set the contents of a variable. The variable is declared
4319 inline.
4320
Daniel Schneller0b547052016-03-21 20:46:57 +01004321 <var-name> The name of the variable starts with an indication about
4322 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +01004323 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +01004324 "sess" : the variable is shared with the whole session
4325 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004326 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +01004327 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004328 processing
Daniel Schneller0b547052016-03-21 20:46:57 +01004329 "res" : the variable is shared only during response
4330 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004331 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +01004332 The name may only contain characters 'a-z', 'A-Z', '0-9',
4333 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004334
4335 <expr> Is a standard HAProxy expression formed by a sample-fetch
4336 followed by some converters.
4337
4338 Example:
4339
4340 http-response set-var(sess.last_redir) res.hdr(location)
4341
Christopher Faulet85d79c92016-11-09 16:54:56 +01004342 - unset-var(<var-name>) :
4343 Is used to unset a variable. See above for details about <var-name>.
4344
4345 Example:
4346
4347 http-response unset-var(sess.last_redir)
4348
Ruoshan Huange4edc6b2016-07-14 15:07:45 +08004349 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
4350 enables tracking of sticky counters from current response. Please refer to
4351 "http-request track-sc" for a complete description. The only difference
4352 from "http-request track-sc" is the <key> sample expression can only make
4353 use of samples in response (eg. res.*, status etc.) and samples below
4354 Layer 6 (eg. ssl related samples, see section 7.3.4). If the sample is
4355 not supported, haproxy will fail and warn while parsing the config.
4356
Thierry FOURNIER236657b2015-08-19 08:25:14 +02004357 - sc-set-gpt0(<sc-id>) <int> :
4358 This action sets the GPT0 tag according to the sticky counter designated
4359 by <sc-id> and the value of <int>. The expected result is a boolean. If
4360 an error occurs, this action silently fails and the actions evaluation
4361 continues.
4362
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02004363 - sc-inc-gpc0(<sc-id>):
4364 This action increments the GPC0 counter according with the sticky counter
4365 designated by <sc-id>. If an error occurs, this action silently fails and
4366 the actions evaluation continues.
4367
Willy Tarreau2d392c22015-08-24 01:43:45 +02004368 - "silent-drop" : this stops the evaluation of the rules and makes the
4369 client-facing connection suddenly disappear using a system-dependant way
4370 that tries to prevent the client from being notified. The effect it then
4371 that the client still sees an established connection while there's none
4372 on HAProxy. The purpose is to achieve a comparable effect to "tarpit"
4373 except that it doesn't use any local resource at all on the machine
4374 running HAProxy. It can resist much higher loads than "tarpit", and slow
4375 down stronger attackers. It is important to undestand the impact of using
4376 this mechanism. All stateful equipments placed between the client and
4377 HAProxy (firewalls, proxies, load balancers) will also keep the
4378 established connection for a long time and may suffer from this action.
4379 On modern Linux systems running with enough privileges, the TCP_REPAIR
4380 socket option is used to block the emission of a TCP reset. On other
4381 systems, the socket's TTL is reduced to 1 so that the TCP reset doesn't
4382 pass the first router, though it's still delivered to local networks. Do
4383 not use it unless you fully understand how it works.
4384
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004385 There is no limit to the number of http-response statements per instance.
4386
Godbach09250262013-07-02 01:19:15 +08004387 It is important to know that http-response rules are processed very early in
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08004388 the HTTP processing, before "rspdel" or "rsprep" or "rspadd" rules. That way,
4389 headers added by "add-header"/"set-header" are visible by almost all further ACL
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004390 rules.
4391
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08004392 Using "rspadd"/"rspdel"/"rsprep" to manipulate request headers is discouraged
4393 in newer versions (>= 1.5). But if you need to use regular expression to
4394 delete headers, you can still use "rspdel". Also please use
4395 "http-response deny" instead of "rspdeny".
4396
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004397 Example:
4398 acl key_acl res.hdr(X-Acl-Key) -m found
4399
4400 acl myhost hdr(Host) -f myhost.lst
4401
4402 http-response add-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
4403 http-response del-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
4404
4405 Example:
4406 acl value res.hdr(X-Value) -m found
4407
4408 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
4409
4410 http-response set-map(map.lst) %[src] %[res.hdr(X-Value)] if value
4411 http-response del-map(map.lst) %[src] if ! value
4412
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004413 See also : "http-request", section 3.4 about userlists and section 7 about
4414 ACL usage.
4415
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02004416
Willy Tarreau30631952015-08-06 15:05:24 +02004417http-reuse { never | safe | aggressive | always }
4418 Declare how idle HTTP connections may be shared between requests
4419
4420 May be used in sections: defaults | frontend | listen | backend
4421 yes | no | yes | yes
4422
4423 By default, a connection established between haproxy and the backend server
4424 belongs to the session that initiated it. The downside is that between the
4425 response and the next request, the connection remains idle and is not used.
4426 In many cases for performance reasons it is desirable to make it possible to
4427 reuse these idle connections to serve other requests from different sessions.
4428 This directive allows to tune this behaviour.
4429
4430 The argument indicates the desired connection reuse strategy :
4431
4432 - "never" : idle connections are never shared between sessions. This is
4433 the default choice. It may be enforced to cancel a different
4434 strategy inherited from a defaults section or for
4435 troubleshooting. For example, if an old bogus application
4436 considers that multiple requests over the same connection come
4437 from the same client and it is not possible to fix the
4438 application, it may be desirable to disable connection sharing
4439 in a single backend. An example of such an application could
4440 be an old haproxy using cookie insertion in tunnel mode and
4441 not checking any request past the first one.
4442
4443 - "safe" : this is the recommended strategy. The first request of a
4444 session is always sent over its own connection, and only
4445 subsequent requests may be dispatched over other existing
4446 connections. This ensures that in case the server closes the
4447 connection when the request is being sent, the browser can
4448 decide to silently retry it. Since it is exactly equivalent to
4449 regular keep-alive, there should be no side effects.
4450
4451 - "aggressive" : this mode may be useful in webservices environments where
4452 all servers are not necessarily known and where it would be
4453 appreciable to deliver most first requests over existing
4454 connections. In this case, first requests are only delivered
4455 over existing connections that have been reused at least once,
4456 proving that the server correctly supports connection reuse.
4457 It should only be used when it's sure that the client can
4458 retry a failed request once in a while and where the benefit
4459 of aggressive connection reuse significantly outweights the
4460 downsides of rare connection failures.
4461
4462 - "always" : this mode is only recommended when the path to the server is
4463 known for never breaking existing connections quickly after
4464 releasing them. It allows the first request of a session to be
4465 sent to an existing connection. This can provide a significant
4466 performance increase over the "safe" strategy when the backend
4467 is a cache farm, since such components tend to show a
4468 consistent behaviour and will benefit from the connection
4469 sharing. It is recommended that the "http-keep-alive" timeout
4470 remains low in this mode so that no dead connections remain
4471 usable. In most cases, this will lead to the same performance
4472 gains as "aggressive" but with more risks. It should only be
4473 used when it improves the situation over "aggressive".
4474
4475 When http connection sharing is enabled, a great care is taken to respect the
4476 connection properties and compatiblities. Specifically :
4477 - connections made with "usesrc" followed by a client-dependant value
4478 ("client", "clientip", "hdr_ip") are marked private and never shared ;
4479
4480 - connections sent to a server with a TLS SNI extension are marked private
4481 and are never shared ;
4482
4483 - connections receiving a status code 401 or 407 expect some authentication
4484 to be sent in return. Due to certain bogus authentication schemes (such
4485 as NTLM) relying on the connection, these connections are marked private
4486 and are never shared ;
4487
4488 No connection pool is involved, once a session dies, the last idle connection
4489 it was attached to is deleted at the same time. This ensures that connections
4490 may not last after all sessions are closed.
4491
4492 Note: connection reuse improves the accuracy of the "server maxconn" setting,
4493 because almost no new connection will be established while idle connections
4494 remain available. This is particularly true with the "always" strategy.
4495
4496 See also : "option http-keep-alive", "server maxconn"
4497
4498
Mark Lamourinec2247f02012-01-04 13:02:01 -05004499http-send-name-header [<header>]
4500 Add the server name to a request. Use the header string given by <header>
4501
4502 May be used in sections: defaults | frontend | listen | backend
4503 yes | no | yes | yes
4504
4505 Arguments :
4506
4507 <header> The header string to use to send the server name
4508
4509 The "http-send-name-header" statement causes the name of the target
4510 server to be added to the headers of an HTTP request. The name
4511 is added with the header string proved.
4512
4513 See also : "server"
4514
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01004515id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02004516 Set a persistent ID to a proxy.
4517 May be used in sections : defaults | frontend | listen | backend
4518 no | yes | yes | yes
4519 Arguments : none
4520
4521 Set a persistent ID for the proxy. This ID must be unique and positive.
4522 An unused ID will automatically be assigned if unset. The first assigned
4523 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01004524
4525
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004526ignore-persist { if | unless } <condition>
4527 Declare a condition to ignore persistence
4528 May be used in sections: defaults | frontend | listen | backend
4529 no | yes | yes | yes
4530
4531 By default, when cookie persistence is enabled, every requests containing
4532 the cookie are unconditionally persistent (assuming the target server is up
4533 and running).
4534
4535 The "ignore-persist" statement allows one to declare various ACL-based
4536 conditions which, when met, will cause a request to ignore persistence.
4537 This is sometimes useful to load balance requests for static files, which
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03004538 often don't require persistence. This can also be used to fully disable
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004539 persistence for a specific User-Agent (for example, some web crawler bots).
4540
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004541 The persistence is ignored when an "if" condition is met, or unless an
4542 "unless" condition is met.
4543
4544 See also : "force-persist", "cookie", and section 7 about ACL usage.
4545
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004546load-server-state-from-file { global | local | none }
4547 Allow seamless reload of HAProxy
4548 May be used in sections: defaults | frontend | listen | backend
4549 yes | no | yes | yes
4550
4551 This directive points HAProxy to a file where server state from previous
4552 running process has been saved. That way, when starting up, before handling
4553 traffic, the new process can apply old states to servers exactly has if no
4554 reload occured. The purpose of the "load-server-state-from-file" directive is
4555 to tell haproxy which file to use. For now, only 2 arguments to either prevent
4556 loading state or load states from a file containing all backends and servers.
4557 The state file can be generated by running the command "show servers state"
4558 over the stats socket and redirect output.
4559
4560 The format of the file is versionned and is very specific. To understand it,
4561 please read the documentation of the "show servers state" command (chapter
Kevin Decherf949c7202015-10-13 23:26:44 +02004562 9.2 of Management Guide).
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004563
4564 Arguments:
4565 global load the content of the file pointed by the global directive
4566 named "server-state-file".
4567
4568 local load the content of the file pointed by the directive
4569 "server-state-file-name" if set. If not set, then the backend
4570 name is used as a file name.
4571
4572 none don't load any stat for this backend
4573
4574 Notes:
Willy Tarreaue5a60682016-11-09 14:54:53 +01004575 - server's IP address is preserved across reloads by default, but the
4576 order can be changed thanks to the server's "init-addr" setting. This
4577 means that an IP address change performed on the CLI at run time will
4578 be preserved, and that any change to the local resolver (eg: /etc/hosts)
4579 will possibly not have any effect if the state file is in use.
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004580
4581 - server's weight is applied from previous running process unless it has
4582 has changed between previous and new configuration files.
4583
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004584 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004585
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004586 global
4587 stats socket /tmp/socket
4588 server-state-file /tmp/server_state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004589
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004590 defaults
4591 load-server-state-from-file global
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004592
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004593 backend bk
4594 server s1 127.0.0.1:22 check weight 11
4595 server s2 127.0.0.1:22 check weight 12
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004596
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004597
4598 Then one can run :
4599
4600 socat /tmp/socket - <<< "show servers state" > /tmp/server_state
4601
4602 Content of the file /tmp/server_state would be like this:
4603
4604 1
4605 # <field names skipped for the doc example>
4606 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
4607 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
4608
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004609 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004610
4611 global
4612 stats socket /tmp/socket
4613 server-state-base /etc/haproxy/states
4614
4615 defaults
4616 load-server-state-from-file local
4617
4618 backend bk
4619 server s1 127.0.0.1:22 check weight 11
4620 server s2 127.0.0.1:22 check weight 12
4621
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02004622
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004623 Then one can run :
4624
4625 socat /tmp/socket - <<< "show servers state bk" > /etc/haproxy/states/bk
4626
4627 Content of the file /etc/haproxy/states/bk would be like this:
4628
4629 1
4630 # <field names skipped for the doc example>
4631 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
4632 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
4633
4634 See also: "server-state-file", "server-state-file-name", and
4635 "show servers state"
4636
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004637
Willy Tarreau2769aa02007-12-27 18:26:09 +01004638log global
Willy Tarreau18324f52014-06-27 18:10:07 +02004639log <address> [len <length>] <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02004640no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01004641 Enable per-instance logging of events and traffic.
4642 May be used in sections : defaults | frontend | listen | backend
4643 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02004644
4645 Prefix :
4646 no should be used when the logger list must be flushed. For example,
4647 if you don't want to inherit from the default logger list. This
4648 prefix does not allow arguments.
4649
Willy Tarreau2769aa02007-12-27 18:26:09 +01004650 Arguments :
4651 global should be used when the instance's logging parameters are the
4652 same as the global ones. This is the most common usage. "global"
4653 replaces <address>, <facility> and <level> with those of the log
4654 entries found in the "global" section. Only one "log global"
4655 statement may be used per instance, and this form takes no other
4656 parameter.
4657
4658 <address> indicates where to send the logs. It takes the same format as
4659 for the "global" section's logs, and can be one of :
4660
4661 - An IPv4 address optionally followed by a colon (':') and a UDP
4662 port. If no port is specified, 514 is used by default (the
4663 standard syslog port).
4664
David du Colombier24bb5f52011-03-17 10:40:23 +01004665 - An IPv6 address followed by a colon (':') and optionally a UDP
4666 port. If no port is specified, 514 is used by default (the
4667 standard syslog port).
4668
Willy Tarreau2769aa02007-12-27 18:26:09 +01004669 - A filesystem path to a UNIX domain socket, keeping in mind
4670 considerations for chroot (be sure the path is accessible
4671 inside the chroot) and uid/gid (be sure the path is
4672 appropriately writeable).
4673
William Lallemandb2f07452015-05-12 14:27:13 +02004674 You may want to reference some environment variables in the
4675 address parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01004676
Willy Tarreau18324f52014-06-27 18:10:07 +02004677 <length> is an optional maximum line length. Log lines larger than this
4678 value will be truncated before being sent. The reason is that
4679 syslog servers act differently on log line length. All servers
4680 support the default value of 1024, but some servers simply drop
4681 larger lines while others do log them. If a server supports long
4682 lines, it may make sense to set this value here in order to avoid
4683 truncating long lines. Similarly, if a server drops long lines,
4684 it is preferable to truncate them before sending them. Accepted
4685 values are 80 to 65535 inclusive. The default value of 1024 is
4686 generally fine for all standard usages. Some specific cases of
4687 long captures or JSON-formated logs may require larger values.
4688
Willy Tarreau2769aa02007-12-27 18:26:09 +01004689 <facility> must be one of the 24 standard syslog facilities :
4690
4691 kern user mail daemon auth syslog lpr news
4692 uucp cron auth2 ftp ntp audit alert cron2
4693 local0 local1 local2 local3 local4 local5 local6 local7
4694
4695 <level> is optional and can be specified to filter outgoing messages. By
4696 default, all messages are sent. If a level is specified, only
4697 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02004698 will be sent. An optional minimum level can be specified. If it
4699 is set, logs emitted with a more severe level than this one will
4700 be capped to this level. This is used to avoid sending "emerg"
4701 messages on all terminals on some default syslog configurations.
4702 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01004703
4704 emerg alert crit err warning notice info debug
4705
William Lallemand0f99e342011-10-12 17:50:54 +02004706 It is important to keep in mind that it is the frontend which decides what to
4707 log from a connection, and that in case of content switching, the log entries
4708 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01004709
4710 However, backend log declaration define how and where servers status changes
4711 will be logged. Level "notice" will be used to indicate a server going up,
4712 "warning" will be used for termination signals and definitive service
4713 termination, and "alert" will be used for when a server goes down.
4714
4715 Note : According to RFC3164, messages are truncated to 1024 bytes before
4716 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004717
4718 Example :
4719 log global
Willy Tarreauf7edefa2009-05-10 17:20:05 +02004720 log 127.0.0.1:514 local0 notice # only send important events
4721 log 127.0.0.1:514 local0 notice notice # same but limit output level
William Lallemandb2f07452015-05-12 14:27:13 +02004722 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
Willy Tarreaudad36a32013-03-11 01:20:04 +01004723
Willy Tarreau2769aa02007-12-27 18:26:09 +01004724
William Lallemand48940402012-01-30 16:47:22 +01004725log-format <string>
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01004726 Specifies the log format string to use for traffic logs
4727 May be used in sections: defaults | frontend | listen | backend
4728 yes | yes | yes | no
William Lallemand48940402012-01-30 16:47:22 +01004729
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01004730 This directive specifies the log format string that will be used for all logs
4731 resulting from traffic passing through the frontend using this line. If the
4732 directive is used in a defaults section, all subsequent frontends will use
4733 the same log format. Please see section 8.2.4 which covers the log format
4734 string in depth.
William Lallemand48940402012-01-30 16:47:22 +01004735
Dragan Dosen7ad31542015-09-28 17:16:47 +02004736log-format-sd <string>
4737 Specifies the RFC5424 structured-data log format string
4738 May be used in sections: defaults | frontend | listen | backend
4739 yes | yes | yes | no
4740
4741 This directive specifies the RFC5424 structured-data log format string that
4742 will be used for all logs resulting from traffic passing through the frontend
4743 using this line. If the directive is used in a defaults section, all
4744 subsequent frontends will use the same log format. Please see section 8.2.4
4745 which covers the log format string in depth.
4746
4747 See https://tools.ietf.org/html/rfc5424#section-6.3 for more information
4748 about the RFC5424 structured-data part.
4749
4750 Note : This log format string will be used only for loggers that have set
4751 log format to "rfc5424".
4752
4753 Example :
4754 log-format-sd [exampleSDID@1234\ bytes=\"%B\"\ status=\"%ST\"]
4755
4756
Willy Tarreau094af4e2015-01-07 15:03:42 +01004757log-tag <string>
4758 Specifies the log tag to use for all outgoing logs
4759 May be used in sections: defaults | frontend | listen | backend
4760 yes | yes | yes | yes
4761
4762 Sets the tag field in the syslog header to this string. It defaults to the
4763 log-tag set in the global section, otherwise the program name as launched
4764 from the command line, which usually is "haproxy". Sometimes it can be useful
4765 to differentiate between multiple processes running on the same host, or to
4766 differentiate customer instances running in the same process. In the backend,
4767 logs about servers up/down will use this tag. As a hint, it can be convenient
4768 to set a log-tag related to a hosted customer in a defaults section then put
4769 all the frontends and backends for that customer, then start another customer
4770 in a new defaults section. See also the global "log-tag" directive.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004771
Willy Tarreauc35362a2014-04-25 13:58:37 +02004772max-keep-alive-queue <value>
4773 Set the maximum server queue size for maintaining keep-alive connections
4774 May be used in sections: defaults | frontend | listen | backend
4775 yes | no | yes | yes
4776
4777 HTTP keep-alive tries to reuse the same server connection whenever possible,
4778 but sometimes it can be counter-productive, for example if a server has a lot
4779 of connections while other ones are idle. This is especially true for static
4780 servers.
4781
4782 The purpose of this setting is to set a threshold on the number of queued
4783 connections at which haproxy stops trying to reuse the same server and prefers
4784 to find another one. The default value, -1, means there is no limit. A value
4785 of zero means that keep-alive requests will never be queued. For very close
4786 servers which can be reached with a low latency and which are not sensible to
4787 breaking keep-alive, a low value is recommended (eg: local static server can
4788 use a value of 10 or less). For remote servers suffering from a high latency,
4789 higher values might be needed to cover for the latency and/or the cost of
4790 picking a different server.
4791
4792 Note that this has no impact on responses which are maintained to the same
4793 server consecutively to a 401 response. They will still go to the same server
4794 even if they have to be queued.
4795
4796 See also : "option http-server-close", "option prefer-last-server", server
4797 "maxconn" and cookie persistence.
4798
4799
Willy Tarreau2769aa02007-12-27 18:26:09 +01004800maxconn <conns>
4801 Fix the maximum number of concurrent connections on a frontend
4802 May be used in sections : defaults | frontend | listen | backend
4803 yes | yes | yes | no
4804 Arguments :
4805 <conns> is the maximum number of concurrent connections the frontend will
4806 accept to serve. Excess connections will be queued by the system
4807 in the socket's listen queue and will be served once a connection
4808 closes.
4809
4810 If the system supports it, it can be useful on big sites to raise this limit
4811 very high so that haproxy manages connection queues, instead of leaving the
4812 clients with unanswered connection attempts. This value should not exceed the
4813 global maxconn. Also, keep in mind that a connection contains two buffers
Baptiste Assmann79fb45d2016-03-06 23:34:31 +01004814 of tune.bufsize (16kB by default) each, as well as some other data resulting
4815 in about 33 kB of RAM being consumed per established connection. That means
4816 that a medium system equipped with 1GB of RAM can withstand around
4817 20000-25000 concurrent connections if properly tuned.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004818
4819 Also, when <conns> is set to large values, it is possible that the servers
4820 are not sized to accept such loads, and for this reason it is generally wise
4821 to assign them some reasonable connection limits.
4822
Vincent Bernat6341be52012-06-27 17:18:30 +02004823 By default, this value is set to 2000.
4824
Willy Tarreau2769aa02007-12-27 18:26:09 +01004825 See also : "server", global section's "maxconn", "fullconn"
4826
4827
4828mode { tcp|http|health }
4829 Set the running mode or protocol of the instance
4830 May be used in sections : defaults | frontend | listen | backend
4831 yes | yes | yes | yes
4832 Arguments :
4833 tcp The instance will work in pure TCP mode. A full-duplex connection
4834 will be established between clients and servers, and no layer 7
4835 examination will be performed. This is the default mode. It
4836 should be used for SSL, SSH, SMTP, ...
4837
4838 http The instance will work in HTTP mode. The client request will be
4839 analyzed in depth before connecting to any server. Any request
4840 which is not RFC-compliant will be rejected. Layer 7 filtering,
4841 processing and switching will be possible. This is the mode which
4842 brings HAProxy most of its value.
4843
4844 health The instance will work in "health" mode. It will just reply "OK"
Willy Tarreau82569f92012-09-27 23:48:56 +02004845 to incoming connections and close the connection. Alternatively,
4846 If the "httpchk" option is set, "HTTP/1.0 200 OK" will be sent
4847 instead. Nothing will be logged in either case. This mode is used
4848 to reply to external components health checks. This mode is
4849 deprecated and should not be used anymore as it is possible to do
4850 the same and even better by combining TCP or HTTP modes with the
4851 "monitor" keyword.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004852
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004853 When doing content switching, it is mandatory that the frontend and the
4854 backend are in the same mode (generally HTTP), otherwise the configuration
4855 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004856
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004857 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01004858 defaults http_instances
4859 mode http
4860
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004861 See also : "monitor", "monitor-net"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004862
Willy Tarreau0ba27502007-12-24 16:55:16 +01004863
Cyril Bontéf0c60612010-02-06 14:44:47 +01004864monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01004865 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004866 May be used in sections : defaults | frontend | listen | backend
4867 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01004868 Arguments :
4869 if <cond> the monitor request will fail if the condition is satisfied,
4870 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004871 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01004872 are met, for instance a low number of servers both in a
4873 backend and its backup.
4874
4875 unless <cond> the monitor request will succeed only if the condition is
4876 satisfied, and will fail otherwise. Such a condition may be
4877 based on a test on the presence of a minimum number of active
4878 servers in a list of backends.
4879
4880 This statement adds a condition which can force the response to a monitor
4881 request to report a failure. By default, when an external component queries
4882 the URI dedicated to monitoring, a 200 response is returned. When one of the
4883 conditions above is met, haproxy will return 503 instead of 200. This is
4884 very useful to report a site failure to an external component which may base
4885 routing advertisements between multiple sites on the availability reported by
4886 haproxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02004887 criterion. Note that "monitor fail" only works in HTTP mode. Both status
4888 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004889
4890 Example:
4891 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01004892 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01004893 acl site_dead nbsrv(dynamic) lt 2
4894 acl site_dead nbsrv(static) lt 2
4895 monitor-uri /site_alive
4896 monitor fail if site_dead
4897
Willy Tarreauae94d4d2011-05-11 16:28:49 +02004898 See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004899
4900
4901monitor-net <source>
4902 Declare a source network which is limited to monitor requests
4903 May be used in sections : defaults | frontend | listen | backend
4904 yes | yes | yes | no
4905 Arguments :
4906 <source> is the source IPv4 address or network which will only be able to
4907 get monitor responses to any request. It can be either an IPv4
4908 address, a host name, or an address followed by a slash ('/')
4909 followed by a mask.
4910
4911 In TCP mode, any connection coming from a source matching <source> will cause
4912 the connection to be immediately closed without any log. This allows another
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004913 equipment to probe the port and verify that it is still listening, without
Willy Tarreau2769aa02007-12-27 18:26:09 +01004914 forwarding the connection to a remote server.
4915
4916 In HTTP mode, a connection coming from a source matching <source> will be
4917 accepted, the following response will be sent without waiting for a request,
4918 then the connection will be closed : "HTTP/1.0 200 OK". This is normally
4919 enough for any front-end HTTP probe to detect that the service is UP and
Willy Tarreau82569f92012-09-27 23:48:56 +02004920 running without forwarding the request to a backend server. Note that this
4921 response is sent in raw format, without any transformation. This is important
4922 as it means that it will not be SSL-encrypted on SSL listeners.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004923
Willy Tarreau82569f92012-09-27 23:48:56 +02004924 Monitor requests are processed very early, just after tcp-request connection
4925 ACLs which are the only ones able to block them. These connections are short
4926 lived and never wait for any data from the client. They cannot be logged, and
4927 it is the intended purpose. They are only used to report HAProxy's health to
4928 an upper component, nothing more. Please note that "monitor fail" rules do
4929 not apply to connections intercepted by "monitor-net".
Willy Tarreau2769aa02007-12-27 18:26:09 +01004930
Willy Tarreau95cd2832010-03-04 23:36:33 +01004931 Last, please note that only one "monitor-net" statement can be specified in
4932 a frontend. If more than one is found, only the last one will be considered.
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004933
Willy Tarreau2769aa02007-12-27 18:26:09 +01004934 Example :
4935 # addresses .252 and .253 are just probing us.
4936 frontend www
4937 monitor-net 192.168.0.252/31
4938
4939 See also : "monitor fail", "monitor-uri"
4940
4941
4942monitor-uri <uri>
4943 Intercept a URI used by external components' monitor requests
4944 May be used in sections : defaults | frontend | listen | backend
4945 yes | yes | yes | no
4946 Arguments :
4947 <uri> is the exact URI which we want to intercept to return HAProxy's
4948 health status instead of forwarding the request.
4949
4950 When an HTTP request referencing <uri> will be received on a frontend,
4951 HAProxy will not forward it nor log it, but instead will return either
4952 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
4953 conditions defined with "monitor fail". This is normally enough for any
4954 front-end HTTP probe to detect that the service is UP and running without
4955 forwarding the request to a backend server. Note that the HTTP method, the
4956 version and all headers are ignored, but the request must at least be valid
4957 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
4958
4959 Monitor requests are processed very early. It is not possible to block nor
4960 divert them using ACLs. They cannot be logged either, and it is the intended
4961 purpose. They are only used to report HAProxy's health to an upper component,
4962 nothing more. However, it is possible to add any number of conditions using
4963 "monitor fail" and ACLs so that the result can be adjusted to whatever check
4964 can be imagined (most often the number of available servers in a backend).
4965
4966 Example :
4967 # Use /haproxy_test to report haproxy's status
4968 frontend www
4969 mode http
4970 monitor-uri /haproxy_test
4971
4972 See also : "monitor fail", "monitor-net"
4973
Willy Tarreau0ba27502007-12-24 16:55:16 +01004974
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004975option abortonclose
4976no option abortonclose
4977 Enable or disable early dropping of aborted requests pending in queues.
4978 May be used in sections : defaults | frontend | listen | backend
4979 yes | no | yes | yes
4980 Arguments : none
4981
4982 In presence of very high loads, the servers will take some time to respond.
4983 The per-instance connection queue will inflate, and the response time will
4984 increase respective to the size of the queue times the average per-session
4985 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01004986 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004987 the queue, and slowing down other users, and the servers as well, because the
4988 request will eventually be served, then aborted at the first error
4989 encountered while delivering the response.
4990
4991 As there is no way to distinguish between a full STOP and a simple output
4992 close on the client side, HTTP agents should be conservative and consider
4993 that the client might only have closed its output channel while waiting for
4994 the response. However, this introduces risks of congestion when lots of users
4995 do the same, and is completely useless nowadays because probably no client at
4996 all will close the session while waiting for the response. Some HTTP agents
Willy Tarreaud72758d2010-01-12 10:42:19 +01004997 support this behaviour (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004998 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01004999 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005000 of being the single component to break rare but valid traffic is extremely
5001 low, which adds to the temptation to be able to abort a session early while
5002 still not served and not pollute the servers.
5003
5004 In HAProxy, the user can choose the desired behaviour using the option
5005 "abortonclose". By default (without the option) the behaviour is HTTP
5006 compliant and aborted requests will be served. But when the option is
5007 specified, a session with an incoming channel closed will be aborted while
5008 it is still possible, either pending in the queue for a connection slot, or
5009 during the connection establishment if the server has not yet acknowledged
5010 the connection request. This considerably reduces the queue size and the load
5011 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01005012 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005013
5014 If this option has been enabled in a "defaults" section, it can be disabled
5015 in a specific instance by prepending the "no" keyword before it.
5016
5017 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
5018
5019
Willy Tarreau4076a152009-04-02 15:18:36 +02005020option accept-invalid-http-request
5021no option accept-invalid-http-request
5022 Enable or disable relaxing of HTTP request parsing
5023 May be used in sections : defaults | frontend | listen | backend
5024 yes | yes | yes | no
5025 Arguments : none
5026
Willy Tarreau91852eb2015-05-01 13:26:00 +02005027 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02005028 means that invalid characters in header names are not permitted and cause an
5029 error to be returned to the client. This is the desired behaviour as such
5030 forbidden characters are essentially used to build attacks exploiting server
5031 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
5032 server will emit invalid header names for whatever reason (configuration,
5033 implementation) and the issue will not be immediately fixed. In such a case,
5034 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01005035 even if that does not make sense, by specifying this option. Similarly, the
5036 list of characters allowed to appear in a URI is well defined by RFC3986, and
5037 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
5038 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
5039 not allowed at all. Haproxy always blocks a number of them (0..32, 127). The
Willy Tarreau91852eb2015-05-01 13:26:00 +02005040 remaining ones are blocked by default unless this option is enabled. This
Willy Tarreau13317662015-05-01 13:47:08 +02005041 option also relaxes the test on the HTTP version, it allows HTTP/0.9 requests
5042 to pass through (no version specified) and multiple digits for both the major
5043 and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02005044
5045 This option should never be enabled by default as it hides application bugs
5046 and open security breaches. It should only be deployed after a problem has
5047 been confirmed.
5048
5049 When this option is enabled, erroneous header names will still be accepted in
5050 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01005051 analysis using the "show errors" request on the UNIX stats socket. Similarly,
5052 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02005053 also helps confirming that the issue has been solved.
5054
5055 If this option has been enabled in a "defaults" section, it can be disabled
5056 in a specific instance by prepending the "no" keyword before it.
5057
5058 See also : "option accept-invalid-http-response" and "show errors" on the
5059 stats socket.
5060
5061
5062option accept-invalid-http-response
5063no option accept-invalid-http-response
5064 Enable or disable relaxing of HTTP response parsing
5065 May be used in sections : defaults | frontend | listen | backend
5066 yes | no | yes | yes
5067 Arguments : none
5068
Willy Tarreau91852eb2015-05-01 13:26:00 +02005069 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02005070 means that invalid characters in header names are not permitted and cause an
5071 error to be returned to the client. This is the desired behaviour as such
5072 forbidden characters are essentially used to build attacks exploiting server
5073 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
5074 server will emit invalid header names for whatever reason (configuration,
5075 implementation) and the issue will not be immediately fixed. In such a case,
5076 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau91852eb2015-05-01 13:26:00 +02005077 even if that does not make sense, by specifying this option. This option also
5078 relaxes the test on the HTTP version format, it allows multiple digits for
5079 both the major and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02005080
5081 This option should never be enabled by default as it hides application bugs
5082 and open security breaches. It should only be deployed after a problem has
5083 been confirmed.
5084
5085 When this option is enabled, erroneous header names will still be accepted in
5086 responses, but the complete response will be captured in order to permit
5087 later analysis using the "show errors" request on the UNIX stats socket.
5088 Doing this also helps confirming that the issue has been solved.
5089
5090 If this option has been enabled in a "defaults" section, it can be disabled
5091 in a specific instance by prepending the "no" keyword before it.
5092
5093 See also : "option accept-invalid-http-request" and "show errors" on the
5094 stats socket.
5095
5096
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005097option allbackups
5098no option allbackups
5099 Use either all backup servers at a time or only the first one
5100 May be used in sections : defaults | frontend | listen | backend
5101 yes | no | yes | yes
5102 Arguments : none
5103
5104 By default, the first operational backup server gets all traffic when normal
5105 servers are all down. Sometimes, it may be preferred to use multiple backups
5106 at once, because one will not be enough. When "option allbackups" is enabled,
5107 the load balancing will be performed among all backup servers when all normal
5108 ones are unavailable. The same load balancing algorithm will be used and the
5109 servers' weights will be respected. Thus, there will not be any priority
5110 order between the backup servers anymore.
5111
5112 This option is mostly used with static server farms dedicated to return a
5113 "sorry" page when an application is completely offline.
5114
5115 If this option has been enabled in a "defaults" section, it can be disabled
5116 in a specific instance by prepending the "no" keyword before it.
5117
5118
5119option checkcache
5120no option checkcache
Godbach7056a352013-12-11 20:01:07 +08005121 Analyze all server responses and block responses with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005122 May be used in sections : defaults | frontend | listen | backend
5123 yes | no | yes | yes
5124 Arguments : none
5125
5126 Some high-level frameworks set application cookies everywhere and do not
5127 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005128 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005129 high risk of session crossing or stealing between users traversing the same
5130 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02005131 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005132
5133 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005134 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01005135 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005136 response to check if there's a risk of caching a cookie on a client-side
5137 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01005138 to the client are :
5139 - all those without "Set-Cookie" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005140 - all those with a return code other than 200, 203, 206, 300, 301, 410,
Willy Tarreau198a7442008-01-17 12:05:32 +01005141 provided that the server has not set a "Cache-control: public" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005142 - all those that come from a POST request, provided that the server has not
5143 set a 'Cache-Control: public' header ;
5144 - those with a 'Pragma: no-cache' header
5145 - those with a 'Cache-control: private' header
5146 - those with a 'Cache-control: no-store' header
5147 - those with a 'Cache-control: max-age=0' header
5148 - those with a 'Cache-control: s-maxage=0' header
5149 - those with a 'Cache-control: no-cache' header
5150 - those with a 'Cache-control: no-cache="set-cookie"' header
5151 - those with a 'Cache-control: no-cache="set-cookie,' header
5152 (allowing other fields after set-cookie)
5153
5154 If a response doesn't respect these requirements, then it will be blocked
Willy Tarreau198a7442008-01-17 12:05:32 +01005155 just as if it was from an "rspdeny" filter, with an "HTTP 502 bad gateway".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005156 The session state shows "PH--" meaning that the proxy blocked the response
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005157 during headers processing. Additionally, an alert will be sent in the logs so
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005158 that admins are informed that there's something to be fixed.
5159
5160 Due to the high impact on the application, the application should be tested
5161 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01005162 good practice to always activate it during tests, even if it is not used in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005163 production, as it will report potentially dangerous application behaviours.
5164
5165 If this option has been enabled in a "defaults" section, it can be disabled
5166 in a specific instance by prepending the "no" keyword before it.
5167
5168
5169option clitcpka
5170no option clitcpka
5171 Enable or disable the sending of TCP keepalive packets on the client side
5172 May be used in sections : defaults | frontend | listen | backend
5173 yes | yes | yes | no
5174 Arguments : none
5175
5176 When there is a firewall or any session-aware component between a client and
5177 a server, and when the protocol involves very long sessions with long idle
5178 periods (eg: remote desktops), there is a risk that one of the intermediate
5179 components decides to expire a session which has remained idle for too long.
5180
5181 Enabling socket-level TCP keep-alives makes the system regularly send packets
5182 to the other end of the connection, leaving it active. The delay between
5183 keep-alive probes is controlled by the system only and depends both on the
5184 operating system and its tuning parameters.
5185
5186 It is important to understand that keep-alive packets are neither emitted nor
5187 received at the application level. It is only the network stacks which sees
5188 them. For this reason, even if one side of the proxy already uses keep-alives
5189 to maintain its connection alive, those keep-alive packets will not be
5190 forwarded to the other side of the proxy.
5191
5192 Please note that this has nothing to do with HTTP keep-alive.
5193
5194 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
5195 client side of a connection, which should help when session expirations are
5196 noticed between HAProxy and a client.
5197
5198 If this option has been enabled in a "defaults" section, it can be disabled
5199 in a specific instance by prepending the "no" keyword before it.
5200
5201 See also : "option srvtcpka", "option tcpka"
5202
5203
Willy Tarreau0ba27502007-12-24 16:55:16 +01005204option contstats
5205 Enable continuous traffic statistics updates
5206 May be used in sections : defaults | frontend | listen | backend
5207 yes | yes | yes | no
5208 Arguments : none
5209
5210 By default, counters used for statistics calculation are incremented
5211 only when a session finishes. It works quite well when serving small
5212 objects, but with big ones (for example large images or archives) or
5213 with A/V streaming, a graph generated from haproxy counters looks like
Willy Tarreaudef0d222016-11-08 22:03:00 +01005214 a hedgehog. With this option enabled counters get incremented frequently
5215 along the session, typically every 5 seconds, which is often enough to
5216 produce clean graphs. Recounting touches a hotpath directly so it is not
5217 not enabled by default, as it can cause a lot of wakeups for very large
5218 session counts and cause a small performance drop.
Willy Tarreau0ba27502007-12-24 16:55:16 +01005219
5220
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005221option dontlog-normal
5222no option dontlog-normal
5223 Enable or disable logging of normal, successful connections
5224 May be used in sections : defaults | frontend | listen | backend
5225 yes | yes | yes | no
5226 Arguments : none
5227
5228 There are large sites dealing with several thousand connections per second
5229 and for which logging is a major pain. Some of them are even forced to turn
5230 logs off and cannot debug production issues. Setting this option ensures that
5231 normal connections, those which experience no error, no timeout, no retry nor
5232 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
5233 mode, the response status code is checked and return codes 5xx will still be
5234 logged.
5235
5236 It is strongly discouraged to use this option as most of the time, the key to
5237 complex issues is in the normal logs which will not be logged here. If you
5238 need to separate logs, see the "log-separate-errors" option instead.
5239
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005240 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005241 logging.
5242
5243
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005244option dontlognull
5245no option dontlognull
5246 Enable or disable logging of null connections
5247 May be used in sections : defaults | frontend | listen | backend
5248 yes | yes | yes | no
5249 Arguments : none
5250
5251 In certain environments, there are components which will regularly connect to
5252 various systems to ensure that they are still alive. It can be the case from
5253 another load balancer as well as from monitoring systems. By default, even a
5254 simple port probe or scan will produce a log. If those connections pollute
5255 the logs too much, it is possible to enable option "dontlognull" to indicate
5256 that a connection on which no data has been transferred will not be logged,
Willy Tarreau0f228a02015-05-01 15:37:53 +02005257 which typically corresponds to those probes. Note that errors will still be
5258 returned to the client and accounted for in the stats. If this is not what is
5259 desired, option http-ignore-probes can be used instead.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005260
5261 It is generally recommended not to use this option in uncontrolled
5262 environments (eg: internet), otherwise scans and other malicious activities
5263 would not be logged.
5264
5265 If this option has been enabled in a "defaults" section, it can be disabled
5266 in a specific instance by prepending the "no" keyword before it.
5267
Willy Tarreau0f228a02015-05-01 15:37:53 +02005268 See also : "log", "http-ignore-probes", "monitor-net", "monitor-uri", and
5269 section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005270
5271
5272option forceclose
5273no option forceclose
5274 Enable or disable active connection closing after response is transferred.
5275 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaua31e5df2009-12-30 01:10:35 +01005276 yes | yes | yes | yes
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005277 Arguments : none
5278
5279 Some HTTP servers do not necessarily close the connections when they receive
5280 the "Connection: close" set by "option httpclose", and if the client does not
5281 close either, then the connection remains open till the timeout expires. This
5282 causes high number of simultaneous connections on the servers and shows high
5283 global session times in the logs.
5284
5285 When this happens, it is possible to use "option forceclose". It will
Willy Tarreau82eeaf22009-12-29 12:09:05 +01005286 actively close the outgoing server channel as soon as the server has finished
Cyril Bonté653dcd62014-02-20 00:13:15 +01005287 to respond and release some resources earlier than with "option httpclose".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005288
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005289 This option may also be combined with "option http-pretend-keepalive", which
5290 will disable sending of the "Connection: close" header, but will still cause
5291 the connection to be closed once the whole response is received.
5292
Cyril Bonté653dcd62014-02-20 00:13:15 +01005293 This option disables and replaces any previous "option httpclose", "option
5294 http-server-close", "option http-keep-alive", or "option http-tunnel".
Willy Tarreau02bce8b2014-01-30 00:15:28 +01005295
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005296 If this option has been enabled in a "defaults" section, it can be disabled
5297 in a specific instance by prepending the "no" keyword before it.
5298
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005299 See also : "option httpclose" and "option http-pretend-keepalive"
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005300
5301
Willy Tarreau87cf5142011-08-19 22:57:24 +02005302option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01005303 Enable insertion of the X-Forwarded-For header to requests sent to servers
5304 May be used in sections : defaults | frontend | listen | backend
5305 yes | yes | yes | yes
5306 Arguments :
5307 <network> is an optional argument used to disable this option for sources
5308 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02005309 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01005310 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005311
5312 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
5313 their client address. This is sometimes annoying when the client's IP address
5314 is expected in server logs. To solve this problem, the well-known HTTP header
5315 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
5316 This header contains a value representing the client's IP address. Since this
5317 header is always appended at the end of the existing header list, the server
5318 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02005319 the server's manual to find how to enable use of this standard header. Note
5320 that only the last occurrence of the header must be used, since it is really
5321 possible that the client has already brought one.
5322
Willy Tarreaud72758d2010-01-12 10:42:19 +01005323 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02005324 the default "X-Forwarded-For". This can be useful where you might already
Willy Tarreaud72758d2010-01-12 10:42:19 +01005325 have a "X-Forwarded-For" header from a different application (eg: stunnel),
5326 and you need preserve it. Also if your backend server doesn't use the
Ross Westaf72a1d2008-08-03 10:51:45 +02005327 "X-Forwarded-For" header and requires different one (eg: Zeus Web Servers
5328 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01005329
5330 Sometimes, a same HAProxy instance may be shared between a direct client
5331 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
5332 used to decrypt HTTPS traffic). It is possible to disable the addition of the
5333 header for a known source address or network by adding the "except" keyword
5334 followed by the network address. In this case, any source IP matching the
5335 network will not cause an addition of this header. Most common uses are with
5336 private networks or 127.0.0.1.
5337
Willy Tarreau87cf5142011-08-19 22:57:24 +02005338 Alternatively, the keyword "if-none" states that the header will only be
5339 added if it is not present. This should only be used in perfectly trusted
5340 environment, as this might cause a security issue if headers reaching haproxy
5341 are under the control of the end-user.
5342
Willy Tarreauc27debf2008-01-06 08:57:02 +01005343 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02005344 least one of them uses it, the header will be added. Note that the backend's
5345 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02005346 both are defined. In the case of the "if-none" argument, if at least one of
5347 the frontend or the backend does not specify it, it wants the addition to be
5348 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005349
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005350 Example :
Willy Tarreauc27debf2008-01-06 08:57:02 +01005351 # Public HTTP address also used by stunnel on the same machine
5352 frontend www
5353 mode http
5354 option forwardfor except 127.0.0.1 # stunnel already adds the header
5355
Ross Westaf72a1d2008-08-03 10:51:45 +02005356 # Those servers want the IP Address in X-Client
5357 backend www
5358 mode http
5359 option forwardfor header X-Client
5360
Willy Tarreau87cf5142011-08-19 22:57:24 +02005361 See also : "option httpclose", "option http-server-close",
Willy Tarreau70dffda2014-01-30 03:07:23 +01005362 "option forceclose", "option http-keep-alive"
Willy Tarreauc27debf2008-01-06 08:57:02 +01005363
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005364
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005365option http-buffer-request
5366no option http-buffer-request
5367 Enable or disable waiting for whole HTTP request body before proceeding
5368 May be used in sections : defaults | frontend | listen | backend
5369 yes | yes | yes | yes
5370 Arguments : none
5371
5372 It is sometimes desirable to wait for the body of an HTTP request before
5373 taking a decision. This is what is being done by "balance url_param" for
5374 example. The first use case is to buffer requests from slow clients before
5375 connecting to the server. Another use case consists in taking the routing
5376 decision based on the request body's contents. This option placed in a
5377 frontend or backend forces the HTTP processing to wait until either the whole
5378 body is received, or the request buffer is full, or the first chunk is
5379 complete in case of chunked encoding. It can have undesired side effects with
5380 some applications abusing HTTP by expecting unbufferred transmissions between
5381 the frontend and the backend, so this should definitely not be used by
5382 default.
5383
Baptiste Assmanneccdf432015-10-28 13:49:01 +01005384 See also : "option http-no-delay", "timeout http-request"
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005385
5386
Willy Tarreau0f228a02015-05-01 15:37:53 +02005387option http-ignore-probes
5388no option http-ignore-probes
5389 Enable or disable logging of null connections and request timeouts
5390 May be used in sections : defaults | frontend | listen | backend
5391 yes | yes | yes | no
5392 Arguments : none
5393
5394 Recently some browsers started to implement a "pre-connect" feature
5395 consisting in speculatively connecting to some recently visited web sites
5396 just in case the user would like to visit them. This results in many
5397 connections being established to web sites, which end up in 408 Request
5398 Timeout if the timeout strikes first, or 400 Bad Request when the browser
5399 decides to close them first. These ones pollute the log and feed the error
5400 counters. There was already "option dontlognull" but it's insufficient in
5401 this case. Instead, this option does the following things :
5402 - prevent any 400/408 message from being sent to the client if nothing
5403 was received over a connection before it was closed ;
5404 - prevent any log from being emitted in this situation ;
5405 - prevent any error counter from being incremented
5406
5407 That way the empty connection is silently ignored. Note that it is better
5408 not to use this unless it is clear that it is needed, because it will hide
5409 real problems. The most common reason for not receiving a request and seeing
5410 a 408 is due to an MTU inconsistency between the client and an intermediary
5411 element such as a VPN, which blocks too large packets. These issues are
5412 generally seen with POST requests as well as GET with large cookies. The logs
5413 are often the only way to detect them.
5414
5415 If this option has been enabled in a "defaults" section, it can be disabled
5416 in a specific instance by prepending the "no" keyword before it.
5417
5418 See also : "log", "dontlognull", "errorfile", and section 8 about logging.
5419
5420
Willy Tarreau16bfb022010-01-16 19:48:41 +01005421option http-keep-alive
5422no option http-keep-alive
5423 Enable or disable HTTP keep-alive from client to server
5424 May be used in sections : defaults | frontend | listen | backend
5425 yes | yes | yes | yes
5426 Arguments : none
5427
Willy Tarreau70dffda2014-01-30 03:07:23 +01005428 By default HAProxy operates in keep-alive mode with regards to persistent
5429 connections: for each connection it processes each request and response, and
5430 leaves the connection idle on both sides between the end of a response and the
5431 start of a new request. This mode may be changed by several options such as
5432 "option http-server-close", "option forceclose", "option httpclose" or
5433 "option http-tunnel". This option allows to set back the keep-alive mode,
5434 which can be useful when another mode was used in a defaults section.
5435
5436 Setting "option http-keep-alive" enables HTTP keep-alive mode on the client-
5437 and server- sides. This provides the lowest latency on the client side (slow
Willy Tarreau16bfb022010-01-16 19:48:41 +01005438 network) and the fastest session reuse on the server side at the expense
5439 of maintaining idle connections to the servers. In general, it is possible
5440 with this option to achieve approximately twice the request rate that the
5441 "http-server-close" option achieves on small objects. There are mainly two
5442 situations where this option may be useful :
5443
5444 - when the server is non-HTTP compliant and authenticates the connection
5445 instead of requests (eg: NTLM authentication)
5446
5447 - when the cost of establishing the connection to the server is significant
5448 compared to the cost of retrieving the associated object from the server.
5449
5450 This last case can happen when the server is a fast static server of cache.
5451 In this case, the server will need to be properly tuned to support high enough
5452 connection counts because connections will last until the client sends another
5453 request.
5454
5455 If the client request has to go to another backend or another server due to
5456 content switching or the load balancing algorithm, the idle connection will
Willy Tarreau9420b122013-12-15 18:58:25 +01005457 immediately be closed and a new one re-opened. Option "prefer-last-server" is
5458 available to try optimize server selection so that if the server currently
5459 attached to an idle connection is usable, it will be used.
Willy Tarreau16bfb022010-01-16 19:48:41 +01005460
5461 In general it is preferred to use "option http-server-close" with application
5462 servers, and some static servers might benefit from "option http-keep-alive".
5463
5464 At the moment, logs will not indicate whether requests came from the same
5465 session or not. The accept date reported in the logs corresponds to the end
5466 of the previous request, and the request time corresponds to the time spent
5467 waiting for a new request. The keep-alive request time is still bound to the
5468 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
5469 not set.
5470
Cyril Bonté653dcd62014-02-20 00:13:15 +01005471 This option disables and replaces any previous "option httpclose", "option
5472 http-server-close", "option forceclose" or "option http-tunnel". When backend
Willy Tarreau70dffda2014-01-30 03:07:23 +01005473 and frontend options differ, all of these 4 options have precedence over
Cyril Bonté653dcd62014-02-20 00:13:15 +01005474 "option http-keep-alive".
Willy Tarreau16bfb022010-01-16 19:48:41 +01005475
5476 See also : "option forceclose", "option http-server-close",
Willy Tarreau9420b122013-12-15 18:58:25 +01005477 "option prefer-last-server", "option http-pretend-keepalive",
5478 "option httpclose", and "1.1. The HTTP transaction model".
Willy Tarreau16bfb022010-01-16 19:48:41 +01005479
5480
Willy Tarreau96e31212011-05-30 18:10:30 +02005481option http-no-delay
5482no option http-no-delay
5483 Instruct the system to favor low interactive delays over performance in HTTP
5484 May be used in sections : defaults | frontend | listen | backend
5485 yes | yes | yes | yes
5486 Arguments : none
5487
5488 In HTTP, each payload is unidirectional and has no notion of interactivity.
5489 Any agent is expected to queue data somewhat for a reasonably low delay.
5490 There are some very rare server-to-server applications that abuse the HTTP
5491 protocol and expect the payload phase to be highly interactive, with many
5492 interleaved data chunks in both directions within a single request. This is
5493 absolutely not supported by the HTTP specification and will not work across
5494 most proxies or servers. When such applications attempt to do this through
5495 haproxy, it works but they will experience high delays due to the network
5496 optimizations which favor performance by instructing the system to wait for
5497 enough data to be available in order to only send full packets. Typical
5498 delays are around 200 ms per round trip. Note that this only happens with
5499 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
5500 affected.
5501
5502 When "option http-no-delay" is present in either the frontend or the backend
5503 used by a connection, all such optimizations will be disabled in order to
5504 make the exchanges as fast as possible. Of course this offers no guarantee on
5505 the functionality, as it may break at any other place. But if it works via
5506 HAProxy, it will work as fast as possible. This option should never be used
5507 by default, and should never be used at all unless such a buggy application
5508 is discovered. The impact of using this option is an increase of bandwidth
5509 usage and CPU usage, which may significantly lower performance in high
5510 latency environments.
5511
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005512 See also : "option http-buffer-request"
5513
Willy Tarreau96e31212011-05-30 18:10:30 +02005514
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005515option http-pretend-keepalive
5516no option http-pretend-keepalive
5517 Define whether haproxy will announce keepalive to the server or not
5518 May be used in sections : defaults | frontend | listen | backend
5519 yes | yes | yes | yes
5520 Arguments : none
5521
5522 When running with "option http-server-close" or "option forceclose", haproxy
5523 adds a "Connection: close" header to the request forwarded to the server.
5524 Unfortunately, when some servers see this header, they automatically refrain
5525 from using the chunked encoding for responses of unknown length, while this
5526 is totally unrelated. The immediate effect is that this prevents haproxy from
5527 maintaining the client connection alive. A second effect is that a client or
5528 a cache could receive an incomplete response without being aware of it, and
5529 consider the response complete.
5530
5531 By setting "option http-pretend-keepalive", haproxy will make the server
5532 believe it will keep the connection alive. The server will then not fall back
5533 to the abnormal undesired above. When haproxy gets the whole response, it
5534 will close the connection with the server just as it would do with the
5535 "forceclose" option. That way the client gets a normal response and the
5536 connection is correctly closed on the server side.
5537
5538 It is recommended not to enable this option by default, because most servers
5539 will more efficiently close the connection themselves after the last packet,
5540 and release its buffers slightly earlier. Also, the added packet on the
5541 network could slightly reduce the overall peak performance. However it is
5542 worth noting that when this option is enabled, haproxy will have slightly
5543 less work to do. So if haproxy is the bottleneck on the whole architecture,
5544 enabling this option might save a few CPU cycles.
5545
5546 This option may be set both in a frontend and in a backend. It is enabled if
5547 at least one of the frontend or backend holding a connection has it enabled.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005548 This option may be combined with "option httpclose", which will cause
Willy Tarreau22a95342010-09-29 14:31:41 +02005549 keepalive to be announced to the server and close to be announced to the
5550 client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005551
5552 If this option has been enabled in a "defaults" section, it can be disabled
5553 in a specific instance by prepending the "no" keyword before it.
5554
Willy Tarreau16bfb022010-01-16 19:48:41 +01005555 See also : "option forceclose", "option http-server-close", and
5556 "option http-keep-alive"
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005557
Willy Tarreauc27debf2008-01-06 08:57:02 +01005558
Willy Tarreaub608feb2010-01-02 22:47:18 +01005559option http-server-close
5560no option http-server-close
5561 Enable or disable HTTP connection closing on the server side
5562 May be used in sections : defaults | frontend | listen | backend
5563 yes | yes | yes | yes
5564 Arguments : none
5565
Willy Tarreau70dffda2014-01-30 03:07:23 +01005566 By default HAProxy operates in keep-alive mode with regards to persistent
5567 connections: for each connection it processes each request and response, and
5568 leaves the connection idle on both sides between the end of a response and
5569 the start of a new request. This mode may be changed by several options such
5570 as "option http-server-close", "option forceclose", "option httpclose" or
5571 "option http-tunnel". Setting "option http-server-close" enables HTTP
5572 connection-close mode on the server side while keeping the ability to support
5573 HTTP keep-alive and pipelining on the client side. This provides the lowest
5574 latency on the client side (slow network) and the fastest session reuse on
5575 the server side to save server resources, similarly to "option forceclose".
5576 It also permits non-keepalive capable servers to be served in keep-alive mode
5577 to the clients if they conform to the requirements of RFC2616. Please note
5578 that some servers do not always conform to those requirements when they see
5579 "Connection: close" in the request. The effect will be that keep-alive will
5580 never be used. A workaround consists in enabling "option
5581 http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01005582
5583 At the moment, logs will not indicate whether requests came from the same
5584 session or not. The accept date reported in the logs corresponds to the end
5585 of the previous request, and the request time corresponds to the time spent
5586 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01005587 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
5588 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01005589
5590 This option may be set both in a frontend and in a backend. It is enabled if
5591 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01005592 It disables and replaces any previous "option httpclose", "option forceclose",
5593 "option http-tunnel" or "option http-keep-alive". Please check section 4
Willy Tarreau70dffda2014-01-30 03:07:23 +01005594 ("Proxies") to see how this option combines with others when frontend and
5595 backend options differ.
Willy Tarreaub608feb2010-01-02 22:47:18 +01005596
5597 If this option has been enabled in a "defaults" section, it can be disabled
5598 in a specific instance by prepending the "no" keyword before it.
5599
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02005600 See also : "option forceclose", "option http-pretend-keepalive",
Willy Tarreau16bfb022010-01-16 19:48:41 +01005601 "option httpclose", "option http-keep-alive", and
5602 "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01005603
5604
Willy Tarreau02bce8b2014-01-30 00:15:28 +01005605option http-tunnel
5606no option http-tunnel
5607 Disable or enable HTTP connection processing after first transaction
5608 May be used in sections : defaults | frontend | listen | backend
5609 yes | yes | yes | yes
5610 Arguments : none
5611
Willy Tarreau70dffda2014-01-30 03:07:23 +01005612 By default HAProxy operates in keep-alive mode with regards to persistent
5613 connections: for each connection it processes each request and response, and
5614 leaves the connection idle on both sides between the end of a response and
5615 the start of a new request. This mode may be changed by several options such
5616 as "option http-server-close", "option forceclose", "option httpclose" or
5617 "option http-tunnel".
5618
5619 Option "http-tunnel" disables any HTTP processing past the first request and
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005620 the first response. This is the mode which was used by default in versions
Willy Tarreau70dffda2014-01-30 03:07:23 +01005621 1.0 to 1.5-dev21. It is the mode with the lowest processing overhead, which
5622 is normally not needed anymore unless in very specific cases such as when
5623 using an in-house protocol that looks like HTTP but is not compatible, or
5624 just to log one request per client in order to reduce log size. Note that
5625 everything which works at the HTTP level, including header parsing/addition,
5626 cookie processing or content switching will only work for the first request
5627 and will be ignored after the first response.
Willy Tarreau02bce8b2014-01-30 00:15:28 +01005628
5629 If this option has been enabled in a "defaults" section, it can be disabled
5630 in a specific instance by prepending the "no" keyword before it.
5631
5632 See also : "option forceclose", "option http-server-close",
5633 "option httpclose", "option http-keep-alive", and
5634 "1.1. The HTTP transaction model".
5635
5636
Willy Tarreau88d349d2010-01-25 12:15:43 +01005637option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01005638no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01005639 Make use of non-standard Proxy-Connection header instead of Connection
5640 May be used in sections : defaults | frontend | listen | backend
5641 yes | yes | yes | no
5642 Arguments : none
5643
5644 While RFC2616 explicitly states that HTTP/1.1 agents must use the
5645 Connection header to indicate their wish of persistent or non-persistent
5646 connections, both browsers and proxies ignore this header for proxied
5647 connections and make use of the undocumented, non-standard Proxy-Connection
5648 header instead. The issue begins when trying to put a load balancer between
5649 browsers and such proxies, because there will be a difference between what
5650 haproxy understands and what the client and the proxy agree on.
5651
5652 By setting this option in a frontend, haproxy can automatically switch to use
5653 that non-standard header if it sees proxied requests. A proxied request is
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01005654 defined here as one where the URI begins with neither a '/' nor a '*'. This
5655 is incompatible with the HTTP tunnel mode. Note that this option can only be
5656 specified in a frontend and will affect the request along its whole life.
Willy Tarreau88d349d2010-01-25 12:15:43 +01005657
Willy Tarreau844a7e72010-01-31 21:46:18 +01005658 Also, when this option is set, a request which requires authentication will
5659 automatically switch to use proxy authentication headers if it is itself a
5660 proxied request. That makes it possible to check or enforce authentication in
5661 front of an existing proxy.
5662
Willy Tarreau88d349d2010-01-25 12:15:43 +01005663 This option should normally never be used, except in front of a proxy.
5664
5665 See also : "option httpclose", "option forceclose" and "option
5666 http-server-close".
5667
5668
Willy Tarreaud63335a2010-02-26 12:56:52 +01005669option httpchk
5670option httpchk <uri>
5671option httpchk <method> <uri>
5672option httpchk <method> <uri> <version>
5673 Enable HTTP protocol to check on the servers health
5674 May be used in sections : defaults | frontend | listen | backend
5675 yes | no | yes | yes
5676 Arguments :
5677 <method> is the optional HTTP method used with the requests. When not set,
5678 the "OPTIONS" method is used, as it generally requires low server
5679 processing and is easy to filter out from the logs. Any method
5680 may be used, though it is not recommended to invent non-standard
5681 ones.
5682
5683 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
5684 which is accessible by default on almost any server, but may be
5685 changed to any other URI. Query strings are permitted.
5686
5687 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
5688 but some servers might behave incorrectly in HTTP 1.0, so turning
5689 it to HTTP/1.1 may sometimes help. Note that the Host field is
5690 mandatory in HTTP/1.1, and as a trick, it is possible to pass it
5691 after "\r\n" following the version string.
5692
5693 By default, server health checks only consist in trying to establish a TCP
5694 connection. When "option httpchk" is specified, a complete HTTP request is
5695 sent once the TCP connection is established, and responses 2xx and 3xx are
5696 considered valid, while all other ones indicate a server failure, including
5697 the lack of any response.
5698
5699 The port and interval are specified in the server configuration.
5700
5701 This option does not necessarily require an HTTP backend, it also works with
5702 plain TCP backends. This is particularly useful to check simple scripts bound
5703 to some dedicated ports using the inetd daemon.
5704
5705 Examples :
5706 # Relay HTTPS traffic to Apache instance and check service availability
5707 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
5708 backend https_relay
5709 mode tcp
5710 option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www
5711 server apache1 192.168.1.1:443 check port 80
5712
Simon Hormanafc47ee2013-11-25 10:46:35 +09005713 See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",
5714 "option pgsql-check", "http-check" and the "check", "port" and
5715 "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01005716
5717
Willy Tarreauc27debf2008-01-06 08:57:02 +01005718option httpclose
5719no option httpclose
5720 Enable or disable passive HTTP connection closing
5721 May be used in sections : defaults | frontend | listen | backend
5722 yes | yes | yes | yes
5723 Arguments : none
5724
Willy Tarreau70dffda2014-01-30 03:07:23 +01005725 By default HAProxy operates in keep-alive mode with regards to persistent
5726 connections: for each connection it processes each request and response, and
5727 leaves the connection idle on both sides between the end of a response and
5728 the start of a new request. This mode may be changed by several options such
Cyril Bonté653dcd62014-02-20 00:13:15 +01005729 as "option http-server-close", "option forceclose", "option httpclose" or
Willy Tarreau70dffda2014-01-30 03:07:23 +01005730 "option http-tunnel".
5731
5732 If "option httpclose" is set, HAProxy will work in HTTP tunnel mode and check
5733 if a "Connection: close" header is already set in each direction, and will
5734 add one if missing. Each end should react to this by actively closing the TCP
5735 connection after each transfer, thus resulting in a switch to the HTTP close
5736 mode. Any "Connection" header different from "close" will also be removed.
5737 Note that this option is deprecated since what it does is very cheap but not
5738 reliable. Using "option http-server-close" or "option forceclose" is strongly
5739 recommended instead.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005740
5741 It seldom happens that some servers incorrectly ignore this header and do not
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005742 close the connection even though they reply "Connection: close". For this
Willy Tarreau0dfdf192010-01-05 11:33:11 +01005743 reason, they are not compatible with older HTTP 1.0 browsers. If this happens
5744 it is possible to use the "option forceclose" which actively closes the
5745 request connection once the server responds. Option "forceclose" also
5746 releases the server connection earlier because it does not have to wait for
5747 the client to acknowledge it.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005748
5749 This option may be set both in a frontend and in a backend. It is enabled if
5750 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01005751 It disables and replaces any previous "option http-server-close",
5752 "option forceclose", "option http-keep-alive" or "option http-tunnel". Please
Willy Tarreau70dffda2014-01-30 03:07:23 +01005753 check section 4 ("Proxies") to see how this option combines with others when
5754 frontend and backend options differ.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005755
5756 If this option has been enabled in a "defaults" section, it can be disabled
5757 in a specific instance by prepending the "no" keyword before it.
5758
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02005759 See also : "option forceclose", "option http-server-close" and
5760 "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01005761
5762
Emeric Brun3a058f32009-06-30 18:26:00 +02005763option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01005764 Enable logging of HTTP request, session state and timers
5765 May be used in sections : defaults | frontend | listen | backend
5766 yes | yes | yes | yes
Emeric Brun3a058f32009-06-30 18:26:00 +02005767 Arguments :
5768 clf if the "clf" argument is added, then the output format will be
5769 the CLF format instead of HAProxy's default HTTP format. You can
5770 use this when you need to feed HAProxy's logs through a specific
5771 log analyser which only support the CLF format and which is not
5772 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005773
5774 By default, the log output format is very poor, as it only contains the
5775 source and destination addresses, and the instance name. By specifying
5776 "option httplog", each log line turns into a much richer format including,
5777 but not limited to, the HTTP request, the connection timers, the session
5778 status, the connections numbers, the captured headers and cookies, the
5779 frontend, backend and server name, and of course the source address and
5780 ports.
5781
5782 This option may be set either in the frontend or the backend.
5783
PiBa-NLbd556bf2014-12-11 21:31:54 +01005784 Specifying only "option httplog" will automatically clear the 'clf' mode
5785 if it was set by default.
Emeric Brun3a058f32009-06-30 18:26:00 +02005786
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005787 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005788
Willy Tarreau55165fe2009-05-10 12:02:55 +02005789
5790option http_proxy
5791no option http_proxy
5792 Enable or disable plain HTTP proxy mode
5793 May be used in sections : defaults | frontend | listen | backend
5794 yes | yes | yes | yes
5795 Arguments : none
5796
5797 It sometimes happens that people need a pure HTTP proxy which understands
5798 basic proxy requests without caching nor any fancy feature. In this case,
5799 it may be worth setting up an HAProxy instance with the "option http_proxy"
5800 set. In this mode, no server is declared, and the connection is forwarded to
5801 the IP address and port found in the URL after the "http://" scheme.
5802
5803 No host address resolution is performed, so this only works when pure IP
5804 addresses are passed. Since this option's usage perimeter is rather limited,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01005805 it will probably be used only by experts who know they need exactly it. This
5806 is incompatible with the HTTP tunnel mode.
Willy Tarreau55165fe2009-05-10 12:02:55 +02005807
5808 If this option has been enabled in a "defaults" section, it can be disabled
5809 in a specific instance by prepending the "no" keyword before it.
5810
5811 Example :
5812 # this backend understands HTTP proxy requests and forwards them directly.
5813 backend direct_forward
5814 option httpclose
5815 option http_proxy
5816
5817 See also : "option httpclose"
5818
Willy Tarreau211ad242009-10-03 21:45:07 +02005819
Jamie Gloudon801a0a32012-08-25 00:18:33 -04005820option independent-streams
5821no option independent-streams
5822 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02005823 May be used in sections : defaults | frontend | listen | backend
5824 yes | yes | yes | yes
5825 Arguments : none
5826
5827 By default, when data is sent over a socket, both the write timeout and the
5828 read timeout for that socket are refreshed, because we consider that there is
5829 activity on that socket, and we have no other means of guessing if we should
5830 receive data or not.
5831
5832 While this default behaviour is desirable for almost all applications, there
5833 exists a situation where it is desirable to disable it, and only refresh the
5834 read timeout if there are incoming data. This happens on sessions with large
5835 timeouts and low amounts of exchanged data such as telnet session. If the
5836 server suddenly disappears, the output data accumulates in the system's
5837 socket buffers, both timeouts are correctly refreshed, and there is no way
5838 to know the server does not receive them, so we don't timeout. However, when
5839 the underlying protocol always echoes sent data, it would be enough by itself
5840 to detect the issue using the read timeout. Note that this problem does not
5841 happen with more verbose protocols because data won't accumulate long in the
5842 socket buffers.
5843
5844 When this option is set on the frontend, it will disable read timeout updates
5845 on data sent to the client. There probably is little use of this case. When
5846 the option is set on the backend, it will disable read timeout updates on
5847 data sent to the server. Doing so will typically break large HTTP posts from
5848 slow lines, so use it with caution.
5849
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005850 Note: older versions used to call this setting "option independent-streams"
Jamie Gloudon801a0a32012-08-25 00:18:33 -04005851 with a spelling mistake. This spelling is still supported but
5852 deprecated.
5853
Willy Tarreauce887fd2012-05-12 12:50:00 +02005854 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02005855
5856
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02005857option ldap-check
5858 Use LDAPv3 health checks for server testing
5859 May be used in sections : defaults | frontend | listen | backend
5860 yes | no | yes | yes
5861 Arguments : none
5862
5863 It is possible to test that the server correctly talks LDAPv3 instead of just
5864 testing that it accepts the TCP connection. When this option is set, an
5865 LDAPv3 anonymous simple bind message is sent to the server, and the response
5866 is analyzed to find an LDAPv3 bind response message.
5867
5868 The server is considered valid only when the LDAP response contains success
5869 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
5870
5871 Logging of bind requests is server dependent see your documentation how to
5872 configure it.
5873
5874 Example :
5875 option ldap-check
5876
5877 See also : "option httpchk"
5878
5879
Simon Horman98637e52014-06-20 12:30:16 +09005880option external-check
5881 Use external processes for server health checks
5882 May be used in sections : defaults | frontend | listen | backend
5883 yes | no | yes | yes
5884
5885 It is possible to test the health of a server using an external command.
5886 This is achieved by running the executable set using "external-check
5887 command".
5888
5889 Requires the "external-check" global to be set.
5890
5891 See also : "external-check", "external-check command", "external-check path"
5892
5893
Willy Tarreau211ad242009-10-03 21:45:07 +02005894option log-health-checks
5895no option log-health-checks
Willy Tarreaubef1b322014-05-13 21:01:39 +02005896 Enable or disable logging of health checks status updates
Willy Tarreau211ad242009-10-03 21:45:07 +02005897 May be used in sections : defaults | frontend | listen | backend
5898 yes | no | yes | yes
5899 Arguments : none
5900
Willy Tarreaubef1b322014-05-13 21:01:39 +02005901 By default, failed health check are logged if server is UP and successful
5902 health checks are logged if server is DOWN, so the amount of additional
5903 information is limited.
Willy Tarreau211ad242009-10-03 21:45:07 +02005904
Willy Tarreaubef1b322014-05-13 21:01:39 +02005905 When this option is enabled, any change of the health check status or to
5906 the server's health will be logged, so that it becomes possible to know
5907 that a server was failing occasional checks before crashing, or exactly when
5908 it failed to respond a valid HTTP status, then when the port started to
5909 reject connections, then when the server stopped responding at all.
5910
5911 Note that status changes not caused by health checks (eg: enable/disable on
5912 the CLI) are intentionally not logged by this option.
Willy Tarreau211ad242009-10-03 21:45:07 +02005913
Willy Tarreaubef1b322014-05-13 21:01:39 +02005914 See also: "option httpchk", "option ldap-check", "option mysql-check",
5915 "option pgsql-check", "option redis-check", "option smtpchk",
5916 "option tcp-check", "log" and section 8 about logging.
Willy Tarreau211ad242009-10-03 21:45:07 +02005917
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005918
5919option log-separate-errors
5920no option log-separate-errors
5921 Change log level for non-completely successful connections
5922 May be used in sections : defaults | frontend | listen | backend
5923 yes | yes | yes | no
5924 Arguments : none
5925
5926 Sometimes looking for errors in logs is not easy. This option makes haproxy
5927 raise the level of logs containing potentially interesting information such
5928 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
5929 level changes from "info" to "err". This makes it possible to log them
5930 separately to a different file with most syslog daemons. Be careful not to
5931 remove them from the original file, otherwise you would lose ordering which
5932 provides very important information.
5933
5934 Using this option, large sites dealing with several thousand connections per
5935 second may log normal traffic to a rotating buffer and only archive smaller
5936 error logs.
5937
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005938 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005939 logging.
5940
Willy Tarreauc27debf2008-01-06 08:57:02 +01005941
5942option logasap
5943no option logasap
5944 Enable or disable early logging of HTTP requests
5945 May be used in sections : defaults | frontend | listen | backend
5946 yes | yes | yes | no
5947 Arguments : none
5948
5949 By default, HTTP requests are logged upon termination so that the total
5950 transfer time and the number of bytes appear in the logs. When large objects
5951 are being transferred, it may take a while before the request appears in the
5952 logs. Using "option logasap", the request gets logged as soon as the server
5953 sends the complete headers. The only missing information in the logs will be
5954 the total number of bytes which will indicate everything except the amount
5955 of data transferred, and the total time which will not take the transfer
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01005956 time into account. In such a situation, it's a good practice to capture the
Willy Tarreauc27debf2008-01-06 08:57:02 +01005957 "Content-Length" response header so that the logs at least indicate how many
5958 bytes are expected to be transferred.
5959
Willy Tarreaucc6c8912009-02-22 10:53:55 +01005960 Examples :
5961 listen http_proxy 0.0.0.0:80
5962 mode http
5963 option httplog
5964 option logasap
5965 log 192.168.2.200 local3
5966
5967 >>> Feb 6 12:14:14 localhost \
5968 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
5969 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
5970 "GET /image.iso HTTP/1.0"
5971
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005972 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01005973 logging.
5974
5975
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02005976option mysql-check [ user <username> [ post-41 ] ]
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005977 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005978 May be used in sections : defaults | frontend | listen | backend
5979 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005980 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005981 <username> This is the username which will be used when connecting to MySQL
5982 server.
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02005983 post-41 Send post v4.1 client compatible checks
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005984
5985 If you specify a username, the check consists of sending two MySQL packet,
5986 one Client Authentication packet, and one QUIT packet, to correctly close
5987 MySQL session. We then parse the MySQL Handshake Initialisation packet and/or
5988 Error packet. It is a basic but useful test which does not produce error nor
5989 aborted connect on the server. However, it requires adding an authorization
5990 in the MySQL table, like this :
5991
5992 USE mysql;
5993 INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');
5994 FLUSH PRIVILEGES;
5995
5996 If you don't specify a username (it is deprecated and not recommended), the
5997 check only consists in parsing the Mysql Handshake Initialisation packet or
5998 Error packet, we don't send anything in this mode. It was reported that it
5999 can generate lockout if check is too frequent and/or if there is not enough
6000 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
6001 value as if a connection is established successfully within fewer than MySQL
6002 "max_connect_errors" attempts after a previous connection was interrupted,
6003 the error count for the host is cleared to zero. If HAProxy's server get
6004 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
6005
6006 Remember that this does not check database presence nor database consistency.
6007 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006008
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02006009 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006010
6011 Most often, an incoming MySQL server needs to see the client's IP address for
6012 various purposes, including IP privilege matching and connection logging.
6013 When possible, it is often wise to masquerade the client's IP address when
6014 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02006015 which requires the transparent proxy feature to be compiled in, and the MySQL
6016 server to route the client via the machine hosting haproxy.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006017
6018 See also: "option httpchk"
6019
6020
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006021option nolinger
6022no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006023 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006024 May be used in sections: defaults | frontend | listen | backend
6025 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006026 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006027
6028 When clients or servers abort connections in a dirty way (eg: they are
6029 physically disconnected), the session timeouts triggers and the session is
6030 closed. But it will remain in FIN_WAIT1 state for some time in the system,
6031 using some resources and possibly limiting the ability to establish newer
6032 connections.
6033
6034 When this happens, it is possible to activate "option nolinger" which forces
6035 the system to immediately remove any socket's pending data on close. Thus,
6036 the session is instantly purged from the system's tables. This usually has
6037 side effects such as increased number of TCP resets due to old retransmits
6038 getting immediately rejected. Some firewalls may sometimes complain about
6039 this too.
6040
6041 For this reason, it is not recommended to use this option when not absolutely
6042 needed. You know that you need it when you have thousands of FIN_WAIT1
6043 sessions on your system (TIME_WAIT ones do not count).
6044
6045 This option may be used both on frontends and backends, depending on the side
6046 where it is required. Use it on the frontend for clients, and on the backend
6047 for servers.
6048
6049 If this option has been enabled in a "defaults" section, it can be disabled
6050 in a specific instance by prepending the "no" keyword before it.
6051
6052
Willy Tarreau55165fe2009-05-10 12:02:55 +02006053option originalto [ except <network> ] [ header <name> ]
6054 Enable insertion of the X-Original-To header to requests sent to servers
6055 May be used in sections : defaults | frontend | listen | backend
6056 yes | yes | yes | yes
6057 Arguments :
6058 <network> is an optional argument used to disable this option for sources
6059 matching <network>
6060 <name> an optional argument to specify a different "X-Original-To"
6061 header name.
6062
6063 Since HAProxy can work in transparent mode, every request from a client can
6064 be redirected to the proxy and HAProxy itself can proxy every request to a
6065 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
6066 be lost. This is annoying when you want access rules based on destination ip
6067 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
6068 added by HAProxy to all requests sent to the server. This header contains a
6069 value representing the original destination IP address. Since this must be
6070 configured to always use the last occurrence of this header only. Note that
6071 only the last occurrence of the header must be used, since it is really
6072 possible that the client has already brought one.
6073
6074 The keyword "header" may be used to supply a different header name to replace
6075 the default "X-Original-To". This can be useful where you might already
6076 have a "X-Original-To" header from a different application, and you need
6077 preserve it. Also if your backend server doesn't use the "X-Original-To"
6078 header and requires different one.
6079
6080 Sometimes, a same HAProxy instance may be shared between a direct client
6081 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
6082 used to decrypt HTTPS traffic). It is possible to disable the addition of the
6083 header for a known source address or network by adding the "except" keyword
6084 followed by the network address. In this case, any source IP matching the
6085 network will not cause an addition of this header. Most common uses are with
6086 private networks or 127.0.0.1.
6087
6088 This option may be specified either in the frontend or in the backend. If at
6089 least one of them uses it, the header will be added. Note that the backend's
6090 setting of the header subargument takes precedence over the frontend's if
6091 both are defined.
6092
Willy Tarreau55165fe2009-05-10 12:02:55 +02006093 Examples :
6094 # Original Destination address
6095 frontend www
6096 mode http
6097 option originalto except 127.0.0.1
6098
6099 # Those servers want the IP Address in X-Client-Dst
6100 backend www
6101 mode http
6102 option originalto header X-Client-Dst
6103
Willy Tarreau87cf5142011-08-19 22:57:24 +02006104 See also : "option httpclose", "option http-server-close",
6105 "option forceclose"
Willy Tarreau55165fe2009-05-10 12:02:55 +02006106
6107
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006108option persist
6109no option persist
6110 Enable or disable forced persistence on down servers
6111 May be used in sections: defaults | frontend | listen | backend
6112 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006113 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006114
6115 When an HTTP request reaches a backend with a cookie which references a dead
6116 server, by default it is redispatched to another server. It is possible to
6117 force the request to be sent to the dead server first using "option persist"
6118 if absolutely needed. A common use case is when servers are under extreme
6119 load and spend their time flapping. In this case, the users would still be
6120 directed to the server they opened the session on, in the hope they would be
6121 correctly served. It is recommended to use "option redispatch" in conjunction
6122 with this option so that in the event it would not be possible to connect to
6123 the server at all (server definitely dead), the client would finally be
6124 redirected to another valid server.
6125
6126 If this option has been enabled in a "defaults" section, it can be disabled
6127 in a specific instance by prepending the "no" keyword before it.
6128
Willy Tarreau4de91492010-01-22 19:10:05 +01006129 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006130
6131
Willy Tarreau0c122822013-12-15 18:49:01 +01006132option pgsql-check [ user <username> ]
6133 Use PostgreSQL health checks for server testing
6134 May be used in sections : defaults | frontend | listen | backend
6135 yes | no | yes | yes
6136 Arguments :
6137 <username> This is the username which will be used when connecting to
6138 PostgreSQL server.
6139
6140 The check sends a PostgreSQL StartupMessage and waits for either
6141 Authentication request or ErrorResponse message. It is a basic but useful
6142 test which does not produce error nor aborted connect on the server.
6143 This check is identical with the "mysql-check".
6144
6145 See also: "option httpchk"
6146
6147
Willy Tarreau9420b122013-12-15 18:58:25 +01006148option prefer-last-server
6149no option prefer-last-server
6150 Allow multiple load balanced requests to remain on the same server
6151 May be used in sections: defaults | frontend | listen | backend
6152 yes | no | yes | yes
6153 Arguments : none
6154
6155 When the load balancing algorithm in use is not deterministic, and a previous
6156 request was sent to a server to which haproxy still holds a connection, it is
6157 sometimes desirable that subsequent requests on a same session go to the same
6158 server as much as possible. Note that this is different from persistence, as
6159 we only indicate a preference which haproxy tries to apply without any form
6160 of warranty. The real use is for keep-alive connections sent to servers. When
6161 this option is used, haproxy will try to reuse the same connection that is
6162 attached to the server instead of rebalancing to another server, causing a
6163 close of the connection. This can make sense for static file servers. It does
Willy Tarreau068621e2013-12-23 15:11:25 +01006164 not make much sense to use this in combination with hashing algorithms. Note,
6165 haproxy already automatically tries to stick to a server which sends a 401 or
6166 to a proxy which sends a 407 (authentication required). This is mandatory for
6167 use with the broken NTLM authentication challenge, and significantly helps in
6168 troubleshooting some faulty applications. Option prefer-last-server might be
6169 desirable in these environments as well, to avoid redistributing the traffic
6170 after every other response.
Willy Tarreau9420b122013-12-15 18:58:25 +01006171
6172 If this option has been enabled in a "defaults" section, it can be disabled
6173 in a specific instance by prepending the "no" keyword before it.
6174
6175 See also: "option http-keep-alive"
6176
6177
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006178option redispatch
Joseph Lynch726ab712015-05-11 23:25:34 -07006179option redispatch <interval>
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006180no option redispatch
6181 Enable or disable session redistribution in case of connection failure
6182 May be used in sections: defaults | frontend | listen | backend
6183 yes | no | yes | yes
Joseph Lynch726ab712015-05-11 23:25:34 -07006184 Arguments :
6185 <interval> The optional integer value that controls how often redispatches
6186 occur when retrying connections. Positive value P indicates a
6187 redispatch is desired on every Pth retry, and negative value
6188 N indicate a redispath is desired on the Nth retry prior to the
6189 last retry. For example, the default of -1 preserves the
6190 historical behaviour of redispatching on the last retry, a
6191 positive value of 1 would indicate a redispatch on every retry,
6192 and a positive value of 3 would indicate a redispatch on every
6193 third retry. You can disable redispatches with a value of 0.
6194
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006195
6196 In HTTP mode, if a server designated by a cookie is down, clients may
6197 definitely stick to it because they cannot flush the cookie, so they will not
6198 be able to access the service anymore.
6199
6200 Specifying "option redispatch" will allow the proxy to break their
6201 persistence and redistribute them to a working server.
6202
Joseph Lynch726ab712015-05-11 23:25:34 -07006203 It also allows to retry connections to another server in case of multiple
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006204 connection failures. Of course, it requires having "retries" set to a nonzero
6205 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01006206
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006207 This form is the preferred form, which replaces both the "redispatch" and
6208 "redisp" keywords.
6209
6210 If this option has been enabled in a "defaults" section, it can be disabled
6211 in a specific instance by prepending the "no" keyword before it.
6212
Willy Tarreau4de91492010-01-22 19:10:05 +01006213 See also : "redispatch", "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006214
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006215
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02006216option redis-check
6217 Use redis health checks for server testing
6218 May be used in sections : defaults | frontend | listen | backend
6219 yes | no | yes | yes
6220 Arguments : none
6221
6222 It is possible to test that the server correctly talks REDIS protocol instead
6223 of just testing that it accepts the TCP connection. When this option is set,
6224 a PING redis command is sent to the server, and the response is analyzed to
6225 find the "+PONG" response message.
6226
6227 Example :
6228 option redis-check
6229
6230 See also : "option httpchk"
6231
6232
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006233option smtpchk
6234option smtpchk <hello> <domain>
6235 Use SMTP health checks for server testing
6236 May be used in sections : defaults | frontend | listen | backend
6237 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01006238 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006239 <hello> is an optional argument. It is the "hello" command to use. It can
6240 be either "HELO" (for SMTP) or "EHLO" (for ESTMP). All other
6241 values will be turned into the default command ("HELO").
6242
6243 <domain> is the domain name to present to the server. It may only be
6244 specified (and is mandatory) if the hello command has been
6245 specified. By default, "localhost" is used.
6246
6247 When "option smtpchk" is set, the health checks will consist in TCP
6248 connections followed by an SMTP command. By default, this command is
6249 "HELO localhost". The server's return code is analyzed and only return codes
6250 starting with a "2" will be considered as valid. All other responses,
6251 including a lack of response will constitute an error and will indicate a
6252 dead server.
6253
6254 This test is meant to be used with SMTP servers or relays. Depending on the
6255 request, it is possible that some servers do not log each connection attempt,
6256 so you may want to experiment to improve the behaviour. Using telnet on port
6257 25 is often easier than adjusting the configuration.
6258
6259 Most often, an incoming SMTP server needs to see the client's IP address for
6260 various purposes, including spam filtering, anti-spoofing and logging. When
6261 possible, it is often wise to masquerade the client's IP address when
6262 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02006263 which requires the transparent proxy feature to be compiled in.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006264
6265 Example :
6266 option smtpchk HELO mydomain.org
6267
6268 See also : "option httpchk", "source"
6269
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006270
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02006271option socket-stats
6272no option socket-stats
6273
6274 Enable or disable collecting & providing separate statistics for each socket.
6275 May be used in sections : defaults | frontend | listen | backend
6276 yes | yes | yes | no
6277
6278 Arguments : none
6279
6280
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006281option splice-auto
6282no option splice-auto
6283 Enable or disable automatic kernel acceleration on sockets in both directions
6284 May be used in sections : defaults | frontend | listen | backend
6285 yes | yes | yes | yes
6286 Arguments : none
6287
6288 When this option is enabled either on a frontend or on a backend, haproxy
6289 will automatically evaluate the opportunity to use kernel tcp splicing to
6290 forward data between the client and the server, in either direction. Haproxy
6291 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006292 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006293 are not much aggressive in order to limit excessive use of splicing. This
6294 option requires splicing to be enabled at compile time, and may be globally
6295 disabled with the global option "nosplice". Since splice uses pipes, using it
6296 requires that there are enough spare pipes.
6297
6298 Important note: kernel-based TCP splicing is a Linux-specific feature which
6299 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
6300 transfer data between sockets without copying these data to user-space, thus
6301 providing noticeable performance gains and CPU cycles savings. Since many
6302 early implementations are buggy, corrupt data and/or are inefficient, this
6303 feature is not enabled by default, and it should be used with extreme care.
6304 While it is not possible to detect the correctness of an implementation,
6305 2.6.29 is the first version offering a properly working implementation. In
6306 case of doubt, splicing may be globally disabled using the global "nosplice"
6307 keyword.
6308
6309 Example :
6310 option splice-auto
6311
6312 If this option has been enabled in a "defaults" section, it can be disabled
6313 in a specific instance by prepending the "no" keyword before it.
6314
6315 See also : "option splice-request", "option splice-response", and global
6316 options "nosplice" and "maxpipes"
6317
6318
6319option splice-request
6320no option splice-request
6321 Enable or disable automatic kernel acceleration on sockets for requests
6322 May be used in sections : defaults | frontend | listen | backend
6323 yes | yes | yes | yes
6324 Arguments : none
6325
6326 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04006327 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006328 the client to the server. It might still use the recv/send scheme if there
6329 are no spare pipes left. This option requires splicing to be enabled at
6330 compile time, and may be globally disabled with the global option "nosplice".
6331 Since splice uses pipes, using it requires that there are enough spare pipes.
6332
6333 Important note: see "option splice-auto" for usage limitations.
6334
6335 Example :
6336 option splice-request
6337
6338 If this option has been enabled in a "defaults" section, it can be disabled
6339 in a specific instance by prepending the "no" keyword before it.
6340
6341 See also : "option splice-auto", "option splice-response", and global options
6342 "nosplice" and "maxpipes"
6343
6344
6345option splice-response
6346no option splice-response
6347 Enable or disable automatic kernel acceleration on sockets for responses
6348 May be used in sections : defaults | frontend | listen | backend
6349 yes | yes | yes | yes
6350 Arguments : none
6351
6352 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04006353 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006354 the server to the client. It might still use the recv/send scheme if there
6355 are no spare pipes left. This option requires splicing to be enabled at
6356 compile time, and may be globally disabled with the global option "nosplice".
6357 Since splice uses pipes, using it requires that there are enough spare pipes.
6358
6359 Important note: see "option splice-auto" for usage limitations.
6360
6361 Example :
6362 option splice-response
6363
6364 If this option has been enabled in a "defaults" section, it can be disabled
6365 in a specific instance by prepending the "no" keyword before it.
6366
6367 See also : "option splice-auto", "option splice-request", and global options
6368 "nosplice" and "maxpipes"
6369
6370
Christopher Fauletba7bc162016-11-07 21:07:38 +01006371option spop-check
6372 Use SPOP health checks for server testing
6373 May be used in sections : defaults | frontend | listen | backend
6374 no | no | no | yes
6375 Arguments : none
6376
6377 It is possible to test that the server correctly talks SPOP protocol instead
6378 of just testing that it accepts the TCP connection. When this option is set,
6379 a HELLO handshake is performed between HAProxy and the server, and the
6380 response is analyzed to check no error is reported.
6381
6382 Example :
6383 option spop-check
6384
6385 See also : "option httpchk"
6386
6387
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006388option srvtcpka
6389no option srvtcpka
6390 Enable or disable the sending of TCP keepalive packets on the server side
6391 May be used in sections : defaults | frontend | listen | backend
6392 yes | no | yes | yes
6393 Arguments : none
6394
6395 When there is a firewall or any session-aware component between a client and
6396 a server, and when the protocol involves very long sessions with long idle
6397 periods (eg: remote desktops), there is a risk that one of the intermediate
6398 components decides to expire a session which has remained idle for too long.
6399
6400 Enabling socket-level TCP keep-alives makes the system regularly send packets
6401 to the other end of the connection, leaving it active. The delay between
6402 keep-alive probes is controlled by the system only and depends both on the
6403 operating system and its tuning parameters.
6404
6405 It is important to understand that keep-alive packets are neither emitted nor
6406 received at the application level. It is only the network stacks which sees
6407 them. For this reason, even if one side of the proxy already uses keep-alives
6408 to maintain its connection alive, those keep-alive packets will not be
6409 forwarded to the other side of the proxy.
6410
6411 Please note that this has nothing to do with HTTP keep-alive.
6412
6413 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
6414 server side of a connection, which should help when session expirations are
6415 noticed between HAProxy and a server.
6416
6417 If this option has been enabled in a "defaults" section, it can be disabled
6418 in a specific instance by prepending the "no" keyword before it.
6419
6420 See also : "option clitcpka", "option tcpka"
6421
6422
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006423option ssl-hello-chk
6424 Use SSLv3 client hello health checks for server testing
6425 May be used in sections : defaults | frontend | listen | backend
6426 yes | no | yes | yes
6427 Arguments : none
6428
6429 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
6430 possible to test that the server correctly talks SSL instead of just testing
6431 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
6432 SSLv3 client hello messages are sent once the connection is established to
6433 the server, and the response is analyzed to find an SSL server hello message.
6434 The server is considered valid only when the response contains this server
6435 hello message.
6436
6437 All servers tested till there correctly reply to SSLv3 client hello messages,
6438 and most servers tested do not even log the requests containing only hello
6439 messages, which is appreciable.
6440
Willy Tarreau763a95b2012-10-04 23:15:39 +02006441 Note that this check works even when SSL support was not built into haproxy
6442 because it forges the SSL message. When SSL support is available, it is best
6443 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006444
Willy Tarreau763a95b2012-10-04 23:15:39 +02006445 See also: "option httpchk", "check-ssl"
6446
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006447
Willy Tarreaued179852013-12-16 01:07:00 +01006448option tcp-check
6449 Perform health checks using tcp-check send/expect sequences
6450 May be used in sections: defaults | frontend | listen | backend
6451 yes | no | yes | yes
6452
6453 This health check method is intended to be combined with "tcp-check" command
6454 lists in order to support send/expect types of health check sequences.
6455
6456 TCP checks currently support 4 modes of operations :
6457 - no "tcp-check" directive : the health check only consists in a connection
6458 attempt, which remains the default mode.
6459
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006460 - "tcp-check send" or "tcp-check send-binary" only is mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01006461 used to send a string along with a connection opening. With some
6462 protocols, it helps sending a "QUIT" message for example that prevents
6463 the server from logging a connection error for each health check. The
6464 check result will still be based on the ability to open the connection
6465 only.
6466
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006467 - "tcp-check expect" only is mentioned : this is used to test a banner.
Willy Tarreaued179852013-12-16 01:07:00 +01006468 The connection is opened and haproxy waits for the server to present some
6469 contents which must validate some rules. The check result will be based
6470 on the matching between the contents and the rules. This is suited for
6471 POP, IMAP, SMTP, FTP, SSH, TELNET.
6472
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006473 - both "tcp-check send" and "tcp-check expect" are mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01006474 used to test a hello-type protocol. Haproxy sends a message, the server
6475 responds and its response is analysed. the check result will be based on
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006476 the matching between the response contents and the rules. This is often
Willy Tarreaued179852013-12-16 01:07:00 +01006477 suited for protocols which require a binding or a request/response model.
6478 LDAP, MySQL, Redis and SSL are example of such protocols, though they
6479 already all have their dedicated checks with a deeper understanding of
6480 the respective protocols.
6481 In this mode, many questions may be sent and many answers may be
6482 analysed.
6483
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006484 A fifth mode can be used to insert comments in different steps of the
6485 script.
6486
6487 For each tcp-check rule you create, you can add a "comment" directive,
6488 followed by a string. This string will be reported in the log and stderr
6489 in debug mode. It is useful to make user-friendly error reporting.
6490 The "comment" is of course optional.
6491
6492
Willy Tarreaued179852013-12-16 01:07:00 +01006493 Examples :
6494 # perform a POP check (analyse only server's banner)
6495 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006496 tcp-check expect string +OK\ POP3\ ready comment POP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01006497
6498 # perform an IMAP check (analyse only server's banner)
6499 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006500 tcp-check expect string *\ OK\ IMAP4\ ready comment IMAP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01006501
6502 # look for the redis master server after ensuring it speaks well
6503 # redis protocol, then it exits properly.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006504 # (send a command then analyse the response 3 times)
Willy Tarreaued179852013-12-16 01:07:00 +01006505 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006506 tcp-check comment PING\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01006507 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02006508 tcp-check expect string +PONG
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006509 tcp-check comment role\ check
Willy Tarreaued179852013-12-16 01:07:00 +01006510 tcp-check send info\ replication\r\n
6511 tcp-check expect string role:master
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006512 tcp-check comment QUIT\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01006513 tcp-check send QUIT\r\n
6514 tcp-check expect string +OK
6515
6516 forge a HTTP request, then analyse the response
6517 (send many headers before analyzing)
6518 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006519 tcp-check comment forge\ and\ send\ HTTP\ request
Willy Tarreaued179852013-12-16 01:07:00 +01006520 tcp-check send HEAD\ /\ HTTP/1.1\r\n
6521 tcp-check send Host:\ www.mydomain.com\r\n
6522 tcp-check send User-Agent:\ HAProxy\ tcpcheck\r\n
6523 tcp-check send \r\n
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006524 tcp-check expect rstring HTTP/1\..\ (2..|3..) comment check\ HTTP\ response
Willy Tarreaued179852013-12-16 01:07:00 +01006525
6526
6527 See also : "tcp-check expect", "tcp-check send"
6528
6529
Willy Tarreau9ea05a72009-06-14 12:07:01 +02006530option tcp-smart-accept
6531no option tcp-smart-accept
6532 Enable or disable the saving of one ACK packet during the accept sequence
6533 May be used in sections : defaults | frontend | listen | backend
6534 yes | yes | yes | no
6535 Arguments : none
6536
6537 When an HTTP connection request comes in, the system acknowledges it on
6538 behalf of HAProxy, then the client immediately sends its request, and the
6539 system acknowledges it too while it is notifying HAProxy about the new
6540 connection. HAProxy then reads the request and responds. This means that we
6541 have one TCP ACK sent by the system for nothing, because the request could
6542 very well be acknowledged by HAProxy when it sends its response.
6543
6544 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
6545 sending this useless ACK on platforms which support it (currently at least
6546 Linux). It must not cause any problem, because the system will send it anyway
6547 after 40 ms if the response takes more time than expected to come.
6548
6549 During complex network debugging sessions, it may be desirable to disable
6550 this optimization because delayed ACKs can make troubleshooting more complex
6551 when trying to identify where packets are delayed. It is then possible to
6552 fall back to normal behaviour by specifying "no option tcp-smart-accept".
6553
6554 It is also possible to force it for non-HTTP proxies by simply specifying
6555 "option tcp-smart-accept". For instance, it can make sense with some services
6556 such as SMTP where the server speaks first.
6557
6558 It is recommended to avoid forcing this option in a defaults section. In case
6559 of doubt, consider setting it back to automatic values by prepending the
6560 "default" keyword before it, or disabling it using the "no" keyword.
6561
Willy Tarreaud88edf22009-06-14 15:48:17 +02006562 See also : "option tcp-smart-connect"
6563
6564
6565option tcp-smart-connect
6566no option tcp-smart-connect
6567 Enable or disable the saving of one ACK packet during the connect sequence
6568 May be used in sections : defaults | frontend | listen | backend
6569 yes | no | yes | yes
6570 Arguments : none
6571
6572 On certain systems (at least Linux), HAProxy can ask the kernel not to
6573 immediately send an empty ACK upon a connection request, but to directly
6574 send the buffer request instead. This saves one packet on the network and
6575 thus boosts performance. It can also be useful for some servers, because they
6576 immediately get the request along with the incoming connection.
6577
6578 This feature is enabled when "option tcp-smart-connect" is set in a backend.
6579 It is not enabled by default because it makes network troubleshooting more
6580 complex.
6581
6582 It only makes sense to enable it with protocols where the client speaks first
6583 such as HTTP. In other situations, if there is no data to send in place of
6584 the ACK, a normal ACK is sent.
6585
6586 If this option has been enabled in a "defaults" section, it can be disabled
6587 in a specific instance by prepending the "no" keyword before it.
6588
6589 See also : "option tcp-smart-accept"
6590
Willy Tarreau9ea05a72009-06-14 12:07:01 +02006591
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006592option tcpka
6593 Enable or disable the sending of TCP keepalive packets on both sides
6594 May be used in sections : defaults | frontend | listen | backend
6595 yes | yes | yes | yes
6596 Arguments : none
6597
6598 When there is a firewall or any session-aware component between a client and
6599 a server, and when the protocol involves very long sessions with long idle
6600 periods (eg: remote desktops), there is a risk that one of the intermediate
6601 components decides to expire a session which has remained idle for too long.
6602
6603 Enabling socket-level TCP keep-alives makes the system regularly send packets
6604 to the other end of the connection, leaving it active. The delay between
6605 keep-alive probes is controlled by the system only and depends both on the
6606 operating system and its tuning parameters.
6607
6608 It is important to understand that keep-alive packets are neither emitted nor
6609 received at the application level. It is only the network stacks which sees
6610 them. For this reason, even if one side of the proxy already uses keep-alives
6611 to maintain its connection alive, those keep-alive packets will not be
6612 forwarded to the other side of the proxy.
6613
6614 Please note that this has nothing to do with HTTP keep-alive.
6615
6616 Using option "tcpka" enables the emission of TCP keep-alive probes on both
6617 the client and server sides of a connection. Note that this is meaningful
6618 only in "defaults" or "listen" sections. If this option is used in a
6619 frontend, only the client side will get keep-alives, and if this option is
6620 used in a backend, only the server side will get keep-alives. For this
6621 reason, it is strongly recommended to explicitly use "option clitcpka" and
6622 "option srvtcpka" when the configuration is split between frontends and
6623 backends.
6624
6625 See also : "option clitcpka", "option srvtcpka"
6626
Willy Tarreau844e3c52008-01-11 16:28:18 +01006627
6628option tcplog
6629 Enable advanced logging of TCP connections with session state and timers
6630 May be used in sections : defaults | frontend | listen | backend
6631 yes | yes | yes | yes
6632 Arguments : none
6633
6634 By default, the log output format is very poor, as it only contains the
6635 source and destination addresses, and the instance name. By specifying
6636 "option tcplog", each log line turns into a much richer format including, but
6637 not limited to, the connection timers, the session status, the connections
6638 numbers, the frontend, backend and server name, and of course the source
6639 address and ports. This option is useful for pure TCP proxies in order to
6640 find which of the client or server disconnects or times out. For normal HTTP
6641 proxies, it's better to use "option httplog" which is even more complete.
6642
6643 This option may be set either in the frontend or the backend.
6644
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006645 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01006646
6647
Willy Tarreau844e3c52008-01-11 16:28:18 +01006648option transparent
6649no option transparent
6650 Enable client-side transparent proxying
6651 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01006652 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01006653 Arguments : none
6654
6655 This option was introduced in order to provide layer 7 persistence to layer 3
6656 load balancers. The idea is to use the OS's ability to redirect an incoming
6657 connection for a remote address to a local process (here HAProxy), and let
6658 this process know what address was initially requested. When this option is
6659 used, sessions without cookies will be forwarded to the original destination
6660 IP address of the incoming request (which should match that of another
6661 equipment), while requests with cookies will still be forwarded to the
6662 appropriate server.
6663
6664 Note that contrary to a common belief, this option does NOT make HAProxy
6665 present the client's IP to the server when establishing the connection.
6666
Willy Tarreaua1146052011-03-01 09:51:54 +01006667 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006668 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01006669
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006670
Simon Horman98637e52014-06-20 12:30:16 +09006671external-check command <command>
6672 Executable to run when performing an external-check
6673 May be used in sections : defaults | frontend | listen | backend
6674 yes | no | yes | yes
6675
6676 Arguments :
6677 <command> is the external command to run
6678
Simon Horman98637e52014-06-20 12:30:16 +09006679 The arguments passed to the to the command are:
6680
Cyril Bonté777be862014-12-02 21:21:35 +01006681 <proxy_address> <proxy_port> <server_address> <server_port>
Simon Horman98637e52014-06-20 12:30:16 +09006682
Cyril Bonté777be862014-12-02 21:21:35 +01006683 The <proxy_address> and <proxy_port> are derived from the first listener
6684 that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
6685 listener the proxy_address will be the path of the socket and the
6686 <proxy_port> will be the string "NOT_USED". In a backend section, it's not
6687 possible to determine a listener, and both <proxy_address> and <proxy_port>
6688 will have the string value "NOT_USED".
Simon Horman98637e52014-06-20 12:30:16 +09006689
Cyril Bonté72cda2a2014-12-27 22:28:39 +01006690 Some values are also provided through environment variables.
6691
6692 Environment variables :
6693 HAPROXY_PROXY_ADDR The first bind address if available (or empty if not
6694 applicable, for example in a "backend" section).
6695
6696 HAPROXY_PROXY_ID The backend id.
6697
6698 HAPROXY_PROXY_NAME The backend name.
6699
6700 HAPROXY_PROXY_PORT The first bind port if available (or empty if not
6701 applicable, for example in a "backend" section or
6702 for a UNIX socket).
6703
6704 HAPROXY_SERVER_ADDR The server address.
6705
6706 HAPROXY_SERVER_CURCONN The current number of connections on the server.
6707
6708 HAPROXY_SERVER_ID The server id.
6709
6710 HAPROXY_SERVER_MAXCONN The server max connections.
6711
6712 HAPROXY_SERVER_NAME The server name.
6713
6714 HAPROXY_SERVER_PORT The server port if available (or empty for a UNIX
6715 socket).
6716
6717 PATH The PATH environment variable used when executing
6718 the command may be set using "external-check path".
6719
Simon Horman98637e52014-06-20 12:30:16 +09006720 If the command executed and exits with a zero status then the check is
6721 considered to have passed, otherwise the check is considered to have
6722 failed.
6723
6724 Example :
6725 external-check command /bin/true
6726
6727 See also : "external-check", "option external-check", "external-check path"
6728
6729
6730external-check path <path>
6731 The value of the PATH environment variable used when running an external-check
6732 May be used in sections : defaults | frontend | listen | backend
6733 yes | no | yes | yes
6734
6735 Arguments :
6736 <path> is the path used when executing external command to run
6737
6738 The default path is "".
6739
6740 Example :
6741 external-check path "/usr/bin:/bin"
6742
6743 See also : "external-check", "option external-check",
6744 "external-check command"
6745
6746
Emeric Brun647caf12009-06-30 17:57:00 +02006747persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02006748persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02006749 Enable RDP cookie-based persistence
6750 May be used in sections : defaults | frontend | listen | backend
6751 yes | no | yes | yes
6752 Arguments :
6753 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02006754 default cookie name "msts" will be used. There currently is no
6755 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02006756
6757 This statement enables persistence based on an RDP cookie. The RDP cookie
6758 contains all information required to find the server in the list of known
6759 servers. So when this option is set in the backend, the request is analysed
6760 and if an RDP cookie is found, it is decoded. If it matches a known server
6761 which is still UP (or if "option persist" is set), then the connection is
6762 forwarded to this server.
6763
6764 Note that this only makes sense in a TCP backend, but for this to work, the
6765 frontend must have waited long enough to ensure that an RDP cookie is present
6766 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006767 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02006768 a single "listen" section.
6769
Willy Tarreau61e28f22010-05-16 22:31:05 +02006770 Also, it is important to understand that the terminal server will emit this
6771 RDP cookie only if it is configured for "token redirection mode", which means
6772 that the "IP address redirection" option is disabled.
6773
Emeric Brun647caf12009-06-30 17:57:00 +02006774 Example :
6775 listen tse-farm
6776 bind :3389
6777 # wait up to 5s for an RDP cookie in the request
6778 tcp-request inspect-delay 5s
6779 tcp-request content accept if RDP_COOKIE
6780 # apply RDP cookie persistence
6781 persist rdp-cookie
6782 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02006783 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02006784 balance rdp-cookie
6785 server srv1 1.1.1.1:3389
6786 server srv2 1.1.1.2:3389
6787
Simon Hormanab814e02011-06-24 14:50:20 +09006788 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
6789 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02006790
6791
Willy Tarreau3a7d2072009-03-05 23:48:25 +01006792rate-limit sessions <rate>
6793 Set a limit on the number of new sessions accepted per second on a frontend
6794 May be used in sections : defaults | frontend | listen | backend
6795 yes | yes | yes | no
6796 Arguments :
6797 <rate> The <rate> parameter is an integer designating the maximum number
6798 of new sessions per second to accept on the frontend.
6799
6800 When the frontend reaches the specified number of new sessions per second, it
6801 stops accepting new connections until the rate drops below the limit again.
6802 During this time, the pending sessions will be kept in the socket's backlog
6803 (in system buffers) and haproxy will not even be aware that sessions are
6804 pending. When applying very low limit on a highly loaded service, it may make
6805 sense to increase the socket's backlog using the "backlog" keyword.
6806
6807 This feature is particularly efficient at blocking connection-based attacks
6808 or service abuse on fragile servers. Since the session rate is measured every
6809 millisecond, it is extremely accurate. Also, the limit applies immediately,
6810 no delay is needed at all to detect the threshold.
6811
6812 Example : limit the connection rate on SMTP to 10 per second max
6813 listen smtp
6814 mode tcp
6815 bind :25
6816 rate-limit sessions 10
Panagiotis Panagiotopoulos7282d8e2016-02-11 16:37:15 +02006817 server smtp1 127.0.0.1:1025
Willy Tarreau3a7d2072009-03-05 23:48:25 +01006818
Willy Tarreaua17c2d92011-07-25 08:16:20 +02006819 Note : when the maximum rate is reached, the frontend's status is not changed
6820 but its sockets appear as "WAITING" in the statistics if the
6821 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +01006822
6823 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
6824
6825
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006826redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
6827redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
6828redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006829 Return an HTTP redirection if/unless a condition is matched
6830 May be used in sections : defaults | frontend | listen | backend
6831 no | yes | yes | yes
6832
6833 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +01006834 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006835
Willy Tarreau0140f252008-11-19 21:07:09 +01006836 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006837 <loc> With "redirect location", the exact value in <loc> is placed into
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006838 the HTTP "Location" header. When used in an "http-request" rule,
6839 <loc> value follows the log-format rules and can include some
6840 dynamic values (see Custom Log Format in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006841
6842 <pfx> With "redirect prefix", the "Location" header is built from the
6843 concatenation of <pfx> and the complete URI path, including the
6844 query string, unless the "drop-query" option is specified (see
6845 below). As a special case, if <pfx> equals exactly "/", then
6846 nothing is inserted before the original URI. It allows one to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006847 redirect to the same URL (for instance, to insert a cookie). When
6848 used in an "http-request" rule, <pfx> value follows the log-format
6849 rules and can include some dynamic values (see Custom Log Format
6850 in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006851
6852 <sch> With "redirect scheme", then the "Location" header is built by
6853 concatenating <sch> with "://" then the first occurrence of the
6854 "Host" header, and then the URI path, including the query string
6855 unless the "drop-query" option is specified (see below). If no
6856 path is found or if the path is "*", then "/" is used instead. If
6857 no "Host" header is found, then an empty host component will be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006858 returned, which most recent browsers interpret as redirecting to
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006859 the same host. This directive is mostly used to redirect HTTP to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006860 HTTPS. When used in an "http-request" rule, <sch> value follows
6861 the log-format rules and can include some dynamic values (see
6862 Custom Log Format in section 8.2.4).
Willy Tarreau0140f252008-11-19 21:07:09 +01006863
6864 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +01006865 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
6866 with 302 used by default if no code is specified. 301 means
6867 "Moved permanently", and a browser may cache the Location. 302
Baptiste Assmannea849c02015-08-03 11:42:50 +02006868 means "Moved temporarily" and means that the browser should not
Willy Tarreaub67fdc42013-03-29 19:28:11 +01006869 cache the redirection. 303 is equivalent to 302 except that the
6870 browser will fetch the location with a GET method. 307 is just
6871 like 302 but makes it clear that the same method must be reused.
6872 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +01006873
6874 <option> There are several options which can be specified to adjust the
6875 expected behaviour of a redirection :
6876
6877 - "drop-query"
6878 When this keyword is used in a prefix-based redirection, then the
6879 location will be set without any possible query-string, which is useful
6880 for directing users to a non-secure page for instance. It has no effect
6881 with a location-type redirect.
6882
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01006883 - "append-slash"
6884 This keyword may be used in conjunction with "drop-query" to redirect
6885 users who use a URL not ending with a '/' to the same one with the '/'.
6886 It can be useful to ensure that search engines will only see one URL.
6887 For this, a return code 301 is preferred.
6888
Willy Tarreau0140f252008-11-19 21:07:09 +01006889 - "set-cookie NAME[=value]"
6890 A "Set-Cookie" header will be added with NAME (and optionally "=value")
6891 to the response. This is sometimes used to indicate that a user has
6892 been seen, for instance to protect against some types of DoS. No other
6893 cookie option is added, so the cookie will be a session cookie. Note
6894 that for a browser, a sole cookie name without an equal sign is
6895 different from a cookie with an equal sign.
6896
6897 - "clear-cookie NAME[=]"
6898 A "Set-Cookie" header will be added with NAME (and optionally "="), but
6899 with the "Max-Age" attribute set to zero. This will tell the browser to
6900 delete this cookie. It is useful for instance on logout pages. It is
6901 important to note that clearing the cookie "NAME" will not remove a
6902 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
6903 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006904
6905 Example: move the login URL only to HTTPS.
6906 acl clear dst_port 80
6907 acl secure dst_port 8080
6908 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +01006909 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +01006910 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +01006911 acl cookie_set hdr_sub(cookie) SEEN=1
6912
6913 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +01006914 redirect prefix https://mysite.com if login_page !secure
6915 redirect prefix http://mysite.com drop-query if login_page !uid_given
6916 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +01006917 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006918
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01006919 Example: send redirects for request for articles without a '/'.
6920 acl missing_slash path_reg ^/article/[^/]*$
6921 redirect code 301 prefix / drop-query append-slash if missing_slash
6922
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006923 Example: redirect all HTTP traffic to HTTPS when SSL is handled by haproxy.
David BERARDe7153042012-11-03 00:11:31 +01006924 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006925
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006926 Example: append 'www.' prefix in front of all hosts not having it
Coen Rosdorff596659b2016-04-11 11:33:49 +02006927 http-request redirect code 301 location \
6928 http://www.%[hdr(host)]%[capture.req.uri] \
6929 unless { hdr_beg(host) -i www }
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006930
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006931 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006932
6933
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006934redisp (deprecated)
6935redispatch (deprecated)
6936 Enable or disable session redistribution in case of connection failure
6937 May be used in sections: defaults | frontend | listen | backend
6938 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006939 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006940
6941 In HTTP mode, if a server designated by a cookie is down, clients may
6942 definitely stick to it because they cannot flush the cookie, so they will not
6943 be able to access the service anymore.
6944
6945 Specifying "redispatch" will allow the proxy to break their persistence and
6946 redistribute them to a working server.
6947
6948 It also allows to retry last connection to another server in case of multiple
6949 connection failures. Of course, it requires having "retries" set to a nonzero
6950 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01006951
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006952 This form is deprecated, do not use it in any new configuration, use the new
6953 "option redispatch" instead.
6954
6955 See also : "option redispatch"
6956
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006957
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006958reqadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01006959 Add a header at the end of the HTTP request
6960 May be used in sections : defaults | frontend | listen | backend
6961 no | yes | yes | yes
6962 Arguments :
6963 <string> is the complete line to be added. Any space or known delimiter
6964 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006965 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006966
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006967 <cond> is an optional matching condition built from ACLs. It makes it
6968 possible to ignore this rule when other conditions are not met.
6969
Willy Tarreau303c0352008-01-17 19:01:39 +01006970 A new line consisting in <string> followed by a line feed will be added after
6971 the last header of an HTTP request.
6972
6973 Header transformations only apply to traffic which passes through HAProxy,
6974 and not to traffic generated by HAProxy, such as health-checks or error
6975 responses.
6976
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006977 Example : add "X-Proto: SSL" to requests coming via port 81
6978 acl is-ssl dst_port 81
6979 reqadd X-Proto:\ SSL if is-ssl
6980
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006981 See also: "rspadd", "http-request", section 6 about HTTP header manipulation,
6982 and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006983
6984
Willy Tarreau5321c422010-01-28 20:35:13 +01006985reqallow <search> [{if | unless} <cond>]
6986reqiallow <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006987 Definitely allow an HTTP request if a line matches a regular expression
6988 May be used in sections : defaults | frontend | listen | backend
6989 no | yes | yes | yes
6990 Arguments :
6991 <search> is the regular expression applied to HTTP headers and to the
6992 request line. This is an extended regular expression. Parenthesis
6993 grouping is supported and no preliminary backslash is required.
6994 Any space or known delimiter must be escaped using a backslash
6995 ('\'). The pattern applies to a full line at a time. The
6996 "reqallow" keyword strictly matches case while "reqiallow"
6997 ignores case.
6998
Willy Tarreau5321c422010-01-28 20:35:13 +01006999 <cond> is an optional matching condition built from ACLs. It makes it
7000 possible to ignore this rule when other conditions are not met.
7001
Willy Tarreau303c0352008-01-17 19:01:39 +01007002 A request containing any line which matches extended regular expression
7003 <search> will mark the request as allowed, even if any later test would
7004 result in a deny. The test applies both to the request line and to request
7005 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01007006 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01007007
7008 It is easier, faster and more powerful to use ACLs to write access policies.
7009 Reqdeny, reqallow and reqpass should be avoided in new designs.
7010
7011 Example :
7012 # allow www.* but refuse *.local
7013 reqiallow ^Host:\ www\.
7014 reqideny ^Host:\ .*\.local
7015
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007016 See also: "reqdeny", "block", "http-request", section 6 about HTTP header
7017 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007018
7019
Willy Tarreau5321c422010-01-28 20:35:13 +01007020reqdel <search> [{if | unless} <cond>]
7021reqidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007022 Delete all headers matching a regular expression in an HTTP request
7023 May be used in sections : defaults | frontend | listen | backend
7024 no | yes | yes | yes
7025 Arguments :
7026 <search> is the regular expression applied to HTTP headers and to the
7027 request line. This is an extended regular expression. Parenthesis
7028 grouping is supported and no preliminary backslash is required.
7029 Any space or known delimiter must be escaped using a backslash
7030 ('\'). The pattern applies to a full line at a time. The "reqdel"
7031 keyword strictly matches case while "reqidel" ignores case.
7032
Willy Tarreau5321c422010-01-28 20:35:13 +01007033 <cond> is an optional matching condition built from ACLs. It makes it
7034 possible to ignore this rule when other conditions are not met.
7035
Willy Tarreau303c0352008-01-17 19:01:39 +01007036 Any header line matching extended regular expression <search> in the request
7037 will be completely deleted. Most common use of this is to remove unwanted
7038 and/or dangerous headers or cookies from a request before passing it to the
7039 next servers.
7040
7041 Header transformations only apply to traffic which passes through HAProxy,
7042 and not to traffic generated by HAProxy, such as health-checks or error
7043 responses. Keep in mind that header names are not case-sensitive.
7044
7045 Example :
7046 # remove X-Forwarded-For header and SERVER cookie
7047 reqidel ^X-Forwarded-For:.*
7048 reqidel ^Cookie:.*SERVER=
7049
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007050 See also: "reqadd", "reqrep", "rspdel", "http-request", section 6 about
7051 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007052
7053
Willy Tarreau5321c422010-01-28 20:35:13 +01007054reqdeny <search> [{if | unless} <cond>]
7055reqideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007056 Deny an HTTP request if a line matches a regular expression
7057 May be used in sections : defaults | frontend | listen | backend
7058 no | yes | yes | yes
7059 Arguments :
7060 <search> is the regular expression applied to HTTP headers and to the
7061 request line. This is an extended regular expression. Parenthesis
7062 grouping is supported and no preliminary backslash is required.
7063 Any space or known delimiter must be escaped using a backslash
7064 ('\'). The pattern applies to a full line at a time. The
7065 "reqdeny" keyword strictly matches case while "reqideny" ignores
7066 case.
7067
Willy Tarreau5321c422010-01-28 20:35:13 +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 request containing any line which matches extended regular expression
7072 <search> will mark the request as denied, even if any later test would
7073 result in an allow. The test applies both to the request line and to request
7074 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01007075 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01007076
Willy Tarreauced27012008-01-17 20:35:34 +01007077 A denied request will generate an "HTTP 403 forbidden" response once the
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007078 complete request has been parsed. This is consistent with what is practiced
Willy Tarreaud72758d2010-01-12 10:42:19 +01007079 using ACLs.
Willy Tarreauced27012008-01-17 20:35:34 +01007080
Willy Tarreau303c0352008-01-17 19:01:39 +01007081 It is easier, faster and more powerful to use ACLs to write access policies.
7082 Reqdeny, reqallow and reqpass should be avoided in new designs.
7083
7084 Example :
7085 # refuse *.local, then allow www.*
7086 reqideny ^Host:\ .*\.local
7087 reqiallow ^Host:\ www\.
7088
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007089 See also: "reqallow", "rspdeny", "block", "http-request", section 6 about
7090 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007091
7092
Willy Tarreau5321c422010-01-28 20:35:13 +01007093reqpass <search> [{if | unless} <cond>]
7094reqipass <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007095 Ignore any HTTP request line matching a regular expression in next rules
7096 May be used in sections : defaults | frontend | listen | backend
7097 no | yes | yes | yes
7098 Arguments :
7099 <search> is the regular expression applied to HTTP headers and to the
7100 request line. This is an extended regular expression. Parenthesis
7101 grouping is supported and no preliminary backslash is required.
7102 Any space or known delimiter must be escaped using a backslash
7103 ('\'). The pattern applies to a full line at a time. The
7104 "reqpass" keyword strictly matches case while "reqipass" ignores
7105 case.
7106
Willy Tarreau5321c422010-01-28 20:35:13 +01007107 <cond> is an optional matching condition built from ACLs. It makes it
7108 possible to ignore this rule when other conditions are not met.
7109
Willy Tarreau303c0352008-01-17 19:01:39 +01007110 A request containing any line which matches extended regular expression
7111 <search> will skip next rules, without assigning any deny or allow verdict.
7112 The test applies both to the request line and to request headers. Keep in
7113 mind that URLs in request line are case-sensitive while header names are not.
7114
7115 It is easier, faster and more powerful to use ACLs to write access policies.
7116 Reqdeny, reqallow and reqpass should be avoided in new designs.
7117
7118 Example :
7119 # refuse *.local, then allow www.*, but ignore "www.private.local"
7120 reqipass ^Host:\ www.private\.local
7121 reqideny ^Host:\ .*\.local
7122 reqiallow ^Host:\ www\.
7123
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007124 See also: "reqallow", "reqdeny", "block", "http-request", section 6 about
7125 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007126
7127
Willy Tarreau5321c422010-01-28 20:35:13 +01007128reqrep <search> <string> [{if | unless} <cond>]
7129reqirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007130 Replace a regular expression with a string in an HTTP request line
7131 May be used in sections : defaults | frontend | listen | backend
7132 no | yes | yes | yes
7133 Arguments :
7134 <search> is the regular expression applied to HTTP headers and to the
7135 request line. This is an extended regular expression. Parenthesis
7136 grouping is supported and no preliminary backslash is required.
7137 Any space or known delimiter must be escaped using a backslash
7138 ('\'). The pattern applies to a full line at a time. The "reqrep"
7139 keyword strictly matches case while "reqirep" ignores case.
7140
7141 <string> is the complete line to be added. Any space or known delimiter
7142 must be escaped using a backslash ('\'). References to matched
7143 pattern groups are possible using the common \N form, with N
7144 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007145 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01007146
Willy Tarreau5321c422010-01-28 20:35:13 +01007147 <cond> is an optional matching condition built from ACLs. It makes it
7148 possible to ignore this rule when other conditions are not met.
7149
Willy Tarreau303c0352008-01-17 19:01:39 +01007150 Any line matching extended regular expression <search> in the request (both
7151 the request line and header lines) will be completely replaced with <string>.
7152 Most common use of this is to rewrite URLs or domain names in "Host" headers.
7153
7154 Header transformations only apply to traffic which passes through HAProxy,
7155 and not to traffic generated by HAProxy, such as health-checks or error
7156 responses. Note that for increased readability, it is suggested to add enough
7157 spaces between the request and the response. Keep in mind that URLs in
7158 request line are case-sensitive while header names are not.
7159
7160 Example :
7161 # replace "/static/" with "/" at the beginning of any request path.
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007162 reqrep ^([^\ :]*)\ /static/(.*) \1\ /\2
Willy Tarreau303c0352008-01-17 19:01:39 +01007163 # replace "www.mydomain.com" with "www" in the host name.
7164 reqirep ^Host:\ www.mydomain.com Host:\ www
7165
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007166 See also: "reqadd", "reqdel", "rsprep", "tune.bufsize", "http-request",
7167 section 6 about HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007168
7169
Willy Tarreau5321c422010-01-28 20:35:13 +01007170reqtarpit <search> [{if | unless} <cond>]
7171reqitarpit <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007172 Tarpit an HTTP request containing a line matching a regular expression
7173 May be used in sections : defaults | frontend | listen | backend
7174 no | yes | yes | yes
7175 Arguments :
7176 <search> is the regular expression applied to HTTP headers and to the
7177 request line. This is an extended regular expression. Parenthesis
7178 grouping is supported and no preliminary backslash is required.
7179 Any space or known delimiter must be escaped using a backslash
7180 ('\'). The pattern applies to a full line at a time. The
7181 "reqtarpit" keyword strictly matches case while "reqitarpit"
7182 ignores case.
7183
Willy Tarreau5321c422010-01-28 20:35:13 +01007184 <cond> is an optional matching condition built from ACLs. It makes it
7185 possible to ignore this rule when other conditions are not met.
7186
Willy Tarreau303c0352008-01-17 19:01:39 +01007187 A request containing any line which matches extended regular expression
7188 <search> will be tarpitted, which means that it will connect to nowhere, will
Willy Tarreauced27012008-01-17 20:35:34 +01007189 be kept open for a pre-defined time, then will return an HTTP error 500 so
7190 that the attacker does not suspect it has been tarpitted. The status 500 will
7191 be reported in the logs, but the completion flags will indicate "PT". The
Willy Tarreau303c0352008-01-17 19:01:39 +01007192 delay is defined by "timeout tarpit", or "timeout connect" if the former is
7193 not set.
7194
7195 The goal of the tarpit is to slow down robots attacking servers with
7196 identifiable requests. Many robots limit their outgoing number of connections
7197 and stay connected waiting for a reply which can take several minutes to
7198 come. Depending on the environment and attack, it may be particularly
7199 efficient at reducing the load on the network and firewalls.
7200
Willy Tarreau5321c422010-01-28 20:35:13 +01007201 Examples :
Willy Tarreau303c0352008-01-17 19:01:39 +01007202 # ignore user-agents reporting any flavour of "Mozilla" or "MSIE", but
7203 # block all others.
7204 reqipass ^User-Agent:\.*(Mozilla|MSIE)
7205 reqitarpit ^User-Agent:
7206
Willy Tarreau5321c422010-01-28 20:35:13 +01007207 # block bad guys
7208 acl badguys src 10.1.0.3 172.16.13.20/28
7209 reqitarpit . if badguys
7210
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007211 See also: "reqallow", "reqdeny", "reqpass", "http-request", section 6
7212 about HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007213
7214
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02007215retries <value>
7216 Set the number of retries to perform on a server after a connection failure
7217 May be used in sections: defaults | frontend | listen | backend
7218 yes | no | yes | yes
7219 Arguments :
7220 <value> is the number of times a connection attempt should be retried on
7221 a server when a connection either is refused or times out. The
7222 default value is 3.
7223
7224 It is important to understand that this value applies to the number of
7225 connection attempts, not full requests. When a connection has effectively
7226 been established to a server, there will be no more retry.
7227
7228 In order to avoid immediate reconnections to a server which is restarting,
Joseph Lynch726ab712015-05-11 23:25:34 -07007229 a turn-around timer of min("timeout connect", one second) is applied before
7230 a retry occurs.
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02007231
7232 When "option redispatch" is set, the last retry may be performed on another
7233 server even if a cookie references a different server.
7234
7235 See also : "option redispatch"
7236
7237
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007238rspadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01007239 Add a header at the end of the HTTP response
7240 May be used in sections : defaults | frontend | listen | backend
7241 no | yes | yes | yes
7242 Arguments :
7243 <string> is the complete line to be added. Any space or known delimiter
7244 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007245 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01007246
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007247 <cond> is an optional matching condition built from ACLs. It makes it
7248 possible to ignore this rule when other conditions are not met.
7249
Willy Tarreau303c0352008-01-17 19:01:39 +01007250 A new line consisting in <string> followed by a line feed will be added after
7251 the last header of an HTTP response.
7252
7253 Header transformations only apply to traffic which passes through HAProxy,
7254 and not to traffic generated by HAProxy, such as health-checks or error
7255 responses.
7256
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007257 See also: "rspdel" "reqadd", "http-response", section 6 about HTTP header
7258 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007259
7260
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007261rspdel <search> [{if | unless} <cond>]
7262rspidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007263 Delete all headers matching a regular expression in an HTTP response
7264 May be used in sections : defaults | frontend | listen | backend
7265 no | yes | yes | yes
7266 Arguments :
7267 <search> is the regular expression applied to HTTP headers and to the
7268 response line. This is an extended regular expression, so
7269 parenthesis grouping is supported and no preliminary backslash
7270 is required. Any space or known delimiter must be escaped using
7271 a backslash ('\'). The pattern applies to a full line at a time.
7272 The "rspdel" keyword strictly matches case while "rspidel"
7273 ignores case.
7274
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007275 <cond> is an optional matching condition built from ACLs. It makes it
7276 possible to ignore this rule when other conditions are not met.
7277
Willy Tarreau303c0352008-01-17 19:01:39 +01007278 Any header line matching extended regular expression <search> in the response
7279 will be completely deleted. Most common use of this is to remove unwanted
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02007280 and/or sensitive headers or cookies from a response before passing it to the
Willy Tarreau303c0352008-01-17 19:01:39 +01007281 client.
7282
7283 Header transformations only apply to traffic which passes through HAProxy,
7284 and not to traffic generated by HAProxy, such as health-checks or error
7285 responses. Keep in mind that header names are not case-sensitive.
7286
7287 Example :
7288 # remove the Server header from responses
Willy Tarreau5e80e022013-05-25 08:31:25 +02007289 rspidel ^Server:.*
Willy Tarreau303c0352008-01-17 19:01:39 +01007290
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007291 See also: "rspadd", "rsprep", "reqdel", "http-response", section 6 about
7292 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007293
7294
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007295rspdeny <search> [{if | unless} <cond>]
7296rspideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007297 Block an HTTP response if a line matches a regular expression
7298 May be used in sections : defaults | frontend | listen | backend
7299 no | yes | yes | yes
7300 Arguments :
7301 <search> is the regular expression applied to HTTP headers and to the
7302 response line. This is an extended regular expression, so
7303 parenthesis grouping is supported and no preliminary backslash
7304 is required. Any space or known delimiter must be escaped using
7305 a backslash ('\'). The pattern applies to a full line at a time.
7306 The "rspdeny" keyword strictly matches case while "rspideny"
7307 ignores case.
7308
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007309 <cond> is an optional matching condition built from ACLs. It makes it
7310 possible to ignore this rule when other conditions are not met.
7311
Willy Tarreau303c0352008-01-17 19:01:39 +01007312 A response containing any line which matches extended regular expression
7313 <search> will mark the request as denied. The test applies both to the
7314 response line and to response headers. Keep in mind that header names are not
7315 case-sensitive.
7316
7317 Main use of this keyword is to prevent sensitive information leak and to
Willy Tarreauced27012008-01-17 20:35:34 +01007318 block the response before it reaches the client. If a response is denied, it
7319 will be replaced with an HTTP 502 error so that the client never retrieves
7320 any sensitive data.
Willy Tarreau303c0352008-01-17 19:01:39 +01007321
7322 It is easier, faster and more powerful to use ACLs to write access policies.
7323 Rspdeny should be avoided in new designs.
7324
7325 Example :
7326 # Ensure that no content type matching ms-word will leak
7327 rspideny ^Content-type:\.*/ms-word
7328
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007329 See also: "reqdeny", "acl", "block", "http-response", section 6 about
7330 HTTP header manipulation and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007331
7332
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007333rsprep <search> <string> [{if | unless} <cond>]
7334rspirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007335 Replace a regular expression with a string in an HTTP response line
7336 May be used in sections : defaults | frontend | listen | backend
7337 no | yes | yes | yes
7338 Arguments :
7339 <search> is the regular expression applied to HTTP headers and to the
7340 response line. This is an extended regular expression, so
7341 parenthesis grouping is supported and no preliminary backslash
7342 is required. Any space or known delimiter must be escaped using
7343 a backslash ('\'). The pattern applies to a full line at a time.
7344 The "rsprep" keyword strictly matches case while "rspirep"
7345 ignores case.
7346
7347 <string> is the complete line to be added. Any space or known delimiter
7348 must be escaped using a backslash ('\'). References to matched
7349 pattern groups are possible using the common \N form, with N
7350 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007351 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01007352
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007353 <cond> is an optional matching condition built from ACLs. It makes it
7354 possible to ignore this rule when other conditions are not met.
7355
Willy Tarreau303c0352008-01-17 19:01:39 +01007356 Any line matching extended regular expression <search> in the response (both
7357 the response line and header lines) will be completely replaced with
7358 <string>. Most common use of this is to rewrite Location headers.
7359
7360 Header transformations only apply to traffic which passes through HAProxy,
7361 and not to traffic generated by HAProxy, such as health-checks or error
7362 responses. Note that for increased readability, it is suggested to add enough
7363 spaces between the request and the response. Keep in mind that header names
7364 are not case-sensitive.
7365
7366 Example :
7367 # replace "Location: 127.0.0.1:8080" with "Location: www.mydomain.com"
7368 rspirep ^Location:\ 127.0.0.1:8080 Location:\ www.mydomain.com
7369
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007370 See also: "rspadd", "rspdel", "reqrep", "http-response", section 6 about
7371 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007372
7373
David du Colombier486df472011-03-17 10:40:26 +01007374server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007375 Declare a server in a backend
7376 May be used in sections : defaults | frontend | listen | backend
7377 no | no | yes | yes
7378 Arguments :
7379 <name> is the internal name assigned to this server. This name will
Cyril Bonté941a0c62012-10-15 19:44:24 +02007380 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -05007381 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007382
David du Colombier486df472011-03-17 10:40:26 +01007383 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
7384 resolvable hostname is supported, but this name will be resolved
7385 during start-up. Address "0.0.0.0" or "*" has a special meaning.
7386 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +02007387 address as the one from the client connection. This is useful in
7388 transparent proxy architectures where the client's connection is
7389 intercepted and haproxy must forward to the original destination
7390 address. This is more or less what the "transparent" keyword does
7391 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +01007392 to report statistics. Optionally, an address family prefix may be
7393 used before the address to force the family regardless of the
7394 address format, which can be useful to specify a path to a unix
7395 socket with no slash ('/'). Currently supported prefixes are :
7396 - 'ipv4@' -> address is always IPv4
7397 - 'ipv6@' -> address is always IPv6
7398 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02007399 - 'abns@' -> address is in abstract namespace (Linux only)
William Lallemandb2f07452015-05-12 14:27:13 +02007400 You may want to reference some environment variables in the
7401 address parameter, see section 2.3 about environment
Willy Tarreau6a031d12016-11-07 19:42:35 +01007402 variables. The "init-addr" setting can be used to modify the way
7403 IP addresses should be resolved upon startup.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007404
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007405 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007406 be sent to this port. If unset, the same port the client
7407 connected to will be used. The port may also be prefixed by a "+"
7408 or a "-". In this case, the server's port will be determined by
7409 adding this value to the client's port.
7410
7411 <param*> is a list of parameters for this server. The "server" keywords
7412 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007413 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007414
7415 Examples :
7416 server first 10.1.1.1:1080 cookie first check inter 1000
7417 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +01007418 server transp ipv4@
William Lallemandb2f07452015-05-12 14:27:13 +02007419 server backup "${SRV_BACKUP}:1080" backup
7420 server www1_dc1 "${LAN_DC1}.101:80"
7421 server www1_dc2 "${LAN_DC2}.101:80"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007422
Willy Tarreau55dcaf62015-09-27 15:03:15 +02007423 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
7424 sun_path length is used for the address length. Some other programs
7425 such as socat use the string length only by default. Pass the option
7426 ",unix-tightsocklen=0" to any abstract socket definition in socat to
7427 make it compatible with HAProxy's.
7428
Mark Lamourinec2247f02012-01-04 13:02:01 -05007429 See also: "default-server", "http-send-name-header" and section 5 about
7430 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007431
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007432server-state-file-name [<file>]
7433 Set the server state file to read, load and apply to servers available in
7434 this backend. It only applies when the directive "load-server-state-from-file"
7435 is set to "local". When <file> is not provided or if this directive is not
7436 set, then backend name is used. If <file> starts with a slash '/', then it is
7437 considered as an absolute path. Otherwise, <file> is concatenated to the
7438 global directive "server-state-file-base".
7439
7440 Example: the minimal configuration below would make HAProxy look for the
7441 state server file '/etc/haproxy/states/bk':
7442
7443 global
7444 server-state-file-base /etc/haproxy/states
7445
7446 backend bk
7447 load-server-state-from-file
7448
7449 See also: "server-state-file-base", "load-server-state-from-file", and
7450 "show servers state"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007451
7452source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02007453source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +01007454source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007455 Set the source address for outgoing connections
7456 May be used in sections : defaults | frontend | listen | backend
7457 yes | no | yes | yes
7458 Arguments :
7459 <addr> is the IPv4 address HAProxy will bind to before connecting to a
7460 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +01007461
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007462 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +01007463 the most appropriate address to reach its destination. Optionally
7464 an address family prefix may be used before the address to force
7465 the family regardless of the address format, which can be useful
7466 to specify a path to a unix socket with no slash ('/'). Currently
7467 supported prefixes are :
7468 - 'ipv4@' -> address is always IPv4
7469 - 'ipv6@' -> address is always IPv6
7470 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02007471 - 'abns@' -> address is in abstract namespace (Linux only)
Cyril Bonté307ee1e2015-09-28 23:16:06 +02007472 You may want to reference some environment variables in the
7473 address parameter, see section 2.3 about environment variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007474
7475 <port> is an optional port. It is normally not needed but may be useful
7476 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02007477 the system will select a free port. Note that port ranges are not
7478 supported in the backend. If you want to force port ranges, you
7479 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007480
7481 <addr2> is the IP address to present to the server when connections are
7482 forwarded in full transparent proxy mode. This is currently only
7483 supported on some patched Linux kernels. When this address is
7484 specified, clients connecting to the server will be presented
7485 with this address, while health checks will still use the address
7486 <addr>.
7487
7488 <port2> is the optional port to present to the server when connections
7489 are forwarded in full transparent proxy mode (see <addr2> above).
7490 The default value of zero means the system will select a free
7491 port.
7492
Willy Tarreaubce70882009-09-07 11:51:47 +02007493 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
7494 This is the name of a comma-separated header list which can
7495 contain multiple IP addresses. By default, the last occurrence is
7496 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +01007497 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +02007498 by previous proxy, typically Stunnel. In order to use another
7499 occurrence from the last one, please see the <occ> parameter
7500 below. When the header (or occurrence) is not found, no binding
7501 is performed so that the proxy's default IP address is used. Also
7502 keep in mind that the header name is case insensitive, as for any
7503 HTTP header.
7504
7505 <occ> is the occurrence number of a value to be used in a multi-value
7506 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007507 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +02007508 address. Positive values indicate a position from the first
7509 occurrence, 1 being the first one. Negative values indicate
7510 positions relative to the last one, -1 being the last one. This
7511 is helpful for situations where an X-Forwarded-For header is set
7512 at the entry point of an infrastructure and must be used several
7513 proxy layers away. When this value is not specified, -1 is
7514 assumed. Passing a zero here disables the feature.
7515
Willy Tarreaud53f96b2009-02-04 18:46:54 +01007516 <name> is an optional interface name to which to bind to for outgoing
7517 traffic. On systems supporting this features (currently, only
7518 Linux), this allows one to bind all traffic to the server to
7519 this interface even if it is not the one the system would select
7520 based on routing tables. This should be used with extreme care.
7521 Note that using this option requires root privileges.
7522
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007523 The "source" keyword is useful in complex environments where a specific
7524 address only is allowed to connect to the servers. It may be needed when a
7525 private address must be used through a public gateway for instance, and it is
7526 known that the system cannot determine the adequate source address by itself.
7527
7528 An extension which is available on certain patched Linux kernels may be used
7529 through the "usesrc" optional keyword. It makes it possible to connect to the
7530 servers with an IP address which does not belong to the system itself. This
7531 is called "full transparent proxy mode". For this to work, the destination
7532 servers have to route their traffic back to this address through the machine
7533 running HAProxy, and IP forwarding must generally be enabled on this machine.
7534
7535 In this "full transparent proxy" mode, it is possible to force a specific IP
7536 address to be presented to the servers. This is not much used in fact. A more
7537 common use is to tell HAProxy to present the client's IP address. For this,
7538 there are two methods :
7539
7540 - present the client's IP and port addresses. This is the most transparent
7541 mode, but it can cause problems when IP connection tracking is enabled on
7542 the machine, because a same connection may be seen twice with different
7543 states. However, this solution presents the huge advantage of not
7544 limiting the system to the 64k outgoing address+port couples, because all
7545 of the client ranges may be used.
7546
7547 - present only the client's IP address and select a spare port. This
7548 solution is still quite elegant but slightly less transparent (downstream
7549 firewalls logs will not match upstream's). It also presents the downside
7550 of limiting the number of concurrent connections to the usual 64k ports.
7551 However, since the upstream and downstream ports are different, local IP
7552 connection tracking on the machine will not be upset by the reuse of the
7553 same session.
7554
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007555 This option sets the default source for all servers in the backend. It may
7556 also be specified in a "defaults" section. Finer source address specification
7557 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007558 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007559
Baptiste Assmann91bd3372015-07-17 21:59:42 +02007560 In order to work, "usesrc" requires root privileges.
7561
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007562 Examples :
7563 backend private
7564 # Connect to the servers using our 192.168.1.200 source address
7565 source 192.168.1.200
7566
7567 backend transparent_ssl1
7568 # Connect to the SSL farm from the client's source address
7569 source 192.168.1.200 usesrc clientip
7570
7571 backend transparent_ssl2
7572 # Connect to the SSL farm from the client's source address and port
7573 # not recommended if IP conntrack is present on the local machine.
7574 source 192.168.1.200 usesrc client
7575
7576 backend transparent_ssl3
7577 # Connect to the SSL farm from the client's source address. It
7578 # is more conntrack-friendly.
7579 source 192.168.1.200 usesrc clientip
7580
7581 backend transparent_smtp
7582 # Connect to the SMTP farm from the client's source address/port
7583 # with Tproxy version 4.
7584 source 0.0.0.0 usesrc clientip
7585
Willy Tarreaubce70882009-09-07 11:51:47 +02007586 backend transparent_http
7587 # Connect to the servers using the client's IP as seen by previous
7588 # proxy.
7589 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
7590
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007591 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007592 the Linux kernel on www.balabit.com, the "bind" keyword.
7593
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007594
Willy Tarreau844e3c52008-01-11 16:28:18 +01007595srvtimeout <timeout> (deprecated)
7596 Set the maximum inactivity time on the server side.
7597 May be used in sections : defaults | frontend | listen | backend
7598 yes | no | yes | yes
7599 Arguments :
7600 <timeout> is the timeout value specified in milliseconds by default, but
7601 can be in any other unit if the number is suffixed by the unit,
7602 as explained at the top of this document.
7603
7604 The inactivity timeout applies when the server is expected to acknowledge or
7605 send data. In HTTP mode, this timeout is particularly important to consider
7606 during the first phase of the server's response, when it has to send the
7607 headers, as it directly represents the server's processing time for the
7608 request. To find out what value to put there, it's often good to start with
7609 what would be considered as unacceptable response times, then check the logs
7610 to observe the response time distribution, and adjust the value accordingly.
7611
7612 The value is specified in milliseconds by default, but can be in any other
7613 unit if the number is suffixed by the unit, as specified at the top of this
7614 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
7615 recommended that the client timeout remains equal to the server timeout in
7616 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007617 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01007618 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreaud72758d2010-01-12 10:42:19 +01007619 seconds (eg: 4 or 5 seconds minimum).
Willy Tarreau844e3c52008-01-11 16:28:18 +01007620
7621 This parameter is specific to backends, but can be specified once for all in
7622 "defaults" sections. This is in fact one of the easiest solutions not to
7623 forget about it. An unspecified timeout results in an infinite timeout, which
7624 is not recommended. Such a usage is accepted and works but reports a warning
7625 during startup because it may results in accumulation of expired sessions in
7626 the system if the system's timeouts are not configured either.
7627
7628 This parameter is provided for compatibility but is currently deprecated.
7629 Please use "timeout server" instead.
7630
Willy Tarreauce887fd2012-05-12 12:50:00 +02007631 See also : "timeout server", "timeout tunnel", "timeout client" and
7632 "clitimeout".
Willy Tarreau844e3c52008-01-11 16:28:18 +01007633
7634
Cyril Bonté66c327d2010-10-12 00:14:37 +02007635stats admin { if | unless } <cond>
7636 Enable statistics admin level if/unless a condition is matched
7637 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007638 no | yes | yes | yes
Cyril Bonté66c327d2010-10-12 00:14:37 +02007639
7640 This statement enables the statistics admin level if/unless a condition is
7641 matched.
7642
7643 The admin level allows to enable/disable servers from the web interface. By
7644 default, statistics page is read-only for security reasons.
7645
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007646 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7647 unless you know what you do : memory is not shared between the
7648 processes, which can result in random behaviours.
7649
Cyril Bonté23b39d92011-02-10 22:54:44 +01007650 Currently, the POST request is limited to the buffer size minus the reserved
7651 buffer space, which means that if the list of servers is too long, the
7652 request won't be processed. It is recommended to alter few servers at a
7653 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +02007654
7655 Example :
7656 # statistics admin level only for localhost
7657 backend stats_localhost
7658 stats enable
7659 stats admin if LOCALHOST
7660
7661 Example :
7662 # statistics admin level always enabled because of the authentication
7663 backend stats_auth
7664 stats enable
7665 stats auth admin:AdMiN123
7666 stats admin if TRUE
7667
7668 Example :
7669 # statistics admin level depends on the authenticated user
7670 userlist stats-auth
7671 group admin users admin
7672 user admin insecure-password AdMiN123
7673 group readonly users haproxy
7674 user haproxy insecure-password haproxy
7675
7676 backend stats_auth
7677 stats enable
7678 acl AUTH http_auth(stats-auth)
7679 acl AUTH_ADMIN http_auth_group(stats-auth) admin
7680 stats http-request auth unless AUTH
7681 stats admin if AUTH_ADMIN
7682
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007683 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
7684 "bind-process", section 3.4 about userlists and section 7 about
7685 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +02007686
7687
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007688stats auth <user>:<passwd>
7689 Enable statistics with authentication and grant access to an account
7690 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007691 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007692 Arguments :
7693 <user> is a user name to grant access to
7694
7695 <passwd> is the cleartext password associated to this user
7696
7697 This statement enables statistics with default settings, and restricts access
7698 to declared users only. It may be repeated as many times as necessary to
7699 allow as many users as desired. When a user tries to access the statistics
7700 without a valid account, a "401 Forbidden" response will be returned so that
7701 the browser asks the user to provide a valid user and password. The real
7702 which will be returned to the browser is configurable using "stats realm".
7703
7704 Since the authentication method is HTTP Basic Authentication, the passwords
7705 circulate in cleartext on the network. Thus, it was decided that the
7706 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02007707 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007708
7709 It is also possible to reduce the scope of the proxies which appear in the
7710 report using "stats scope".
7711
7712 Though this statement alone is enough to enable statistics reporting, it is
7713 recommended to set all other settings in order to avoid relying on default
7714 unobvious parameters.
7715
7716 Example :
7717 # public access (limited to this backend only)
7718 backend public_www
7719 server srv1 192.168.0.1:80
7720 stats enable
7721 stats hide-version
7722 stats scope .
7723 stats uri /admin?stats
7724 stats realm Haproxy\ Statistics
7725 stats auth admin1:AdMiN123
7726 stats auth admin2:AdMiN321
7727
7728 # internal monitoring access (unlimited)
7729 backend private_monitoring
7730 stats enable
7731 stats uri /admin?stats
7732 stats refresh 5s
7733
7734 See also : "stats enable", "stats realm", "stats scope", "stats uri"
7735
7736
7737stats enable
7738 Enable statistics reporting with default settings
7739 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007740 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007741 Arguments : none
7742
7743 This statement enables statistics reporting with default settings defined
7744 at build time. Unless stated otherwise, these settings are used :
7745 - stats uri : /haproxy?stats
7746 - stats realm : "HAProxy Statistics"
7747 - stats auth : no authentication
7748 - stats scope : no restriction
7749
7750 Though this statement alone is enough to enable statistics reporting, it is
7751 recommended to set all other settings in order to avoid relying on default
7752 unobvious parameters.
7753
7754 Example :
7755 # public access (limited to this backend only)
7756 backend public_www
7757 server srv1 192.168.0.1:80
7758 stats enable
7759 stats hide-version
7760 stats scope .
7761 stats uri /admin?stats
7762 stats realm Haproxy\ Statistics
7763 stats auth admin1:AdMiN123
7764 stats auth admin2:AdMiN321
7765
7766 # internal monitoring access (unlimited)
7767 backend private_monitoring
7768 stats enable
7769 stats uri /admin?stats
7770 stats refresh 5s
7771
7772 See also : "stats auth", "stats realm", "stats uri"
7773
7774
Willy Tarreaud63335a2010-02-26 12:56:52 +01007775stats hide-version
7776 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007777 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007778 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01007779 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007780
Willy Tarreaud63335a2010-02-26 12:56:52 +01007781 By default, the stats page reports some useful status information along with
7782 the statistics. Among them is HAProxy's version. However, it is generally
7783 considered dangerous to report precise version to anyone, as it can help them
7784 target known weaknesses with specific attacks. The "stats hide-version"
7785 statement removes the version from the statistics report. This is recommended
7786 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007787
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02007788 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
7790 unobvious parameters.
7791
Willy Tarreaud63335a2010-02-26 12:56:52 +01007792 Example :
7793 # public access (limited to this backend only)
7794 backend public_www
7795 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02007796 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01007797 stats hide-version
7798 stats scope .
7799 stats uri /admin?stats
7800 stats realm Haproxy\ Statistics
7801 stats auth admin1:AdMiN123
7802 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007803
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007804 # internal monitoring access (unlimited)
7805 backend private_monitoring
7806 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01007807 stats uri /admin?stats
7808 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +01007809
Willy Tarreaud63335a2010-02-26 12:56:52 +01007810 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007811
Willy Tarreau983e01e2010-01-11 18:42:06 +01007812
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02007813stats http-request { allow | deny | auth [realm <realm>] }
7814 [ { if | unless } <condition> ]
7815 Access control for statistics
7816
7817 May be used in sections: defaults | frontend | listen | backend
7818 no | no | yes | yes
7819
7820 As "http-request", these set of options allow to fine control access to
7821 statistics. Each option may be followed by if/unless and acl.
7822 First option with matched condition (or option without condition) is final.
7823 For "deny" a 403 error will be returned, for "allow" normal processing is
7824 performed, for "auth" a 401/407 error code is returned so the client
7825 should be asked to enter a username and password.
7826
7827 There is no fixed limit to the number of http-request statements per
7828 instance.
7829
7830 See also : "http-request", section 3.4 about userlists and section 7
7831 about ACL usage.
7832
7833
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007834stats realm <realm>
7835 Enable statistics and set authentication realm
7836 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007837 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007838 Arguments :
7839 <realm> is the name of the HTTP Basic Authentication realm reported to
7840 the browser. The browser uses it to display it in the pop-up
7841 inviting the user to enter a valid username and password.
7842
7843 The realm is read as a single word, so any spaces in it should be escaped
7844 using a backslash ('\').
7845
7846 This statement is useful only in conjunction with "stats auth" since it is
7847 only related to authentication.
7848
7849 Though this statement alone is enough to enable statistics reporting, it is
7850 recommended to set all other settings in order to avoid relying on default
7851 unobvious parameters.
7852
7853 Example :
7854 # public access (limited to this backend only)
7855 backend public_www
7856 server srv1 192.168.0.1:80
7857 stats enable
7858 stats hide-version
7859 stats scope .
7860 stats uri /admin?stats
7861 stats realm Haproxy\ Statistics
7862 stats auth admin1:AdMiN123
7863 stats auth admin2:AdMiN321
7864
7865 # internal monitoring access (unlimited)
7866 backend private_monitoring
7867 stats enable
7868 stats uri /admin?stats
7869 stats refresh 5s
7870
7871 See also : "stats auth", "stats enable", "stats uri"
7872
7873
7874stats refresh <delay>
7875 Enable statistics with automatic refresh
7876 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007877 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007878 Arguments :
7879 <delay> is the suggested refresh delay, specified in seconds, which will
7880 be returned to the browser consulting the report page. While the
7881 browser is free to apply any delay, it will generally respect it
7882 and refresh the page this every seconds. The refresh interval may
7883 be specified in any other non-default time unit, by suffixing the
7884 unit after the value, as explained at the top of this document.
7885
7886 This statement is useful on monitoring displays with a permanent page
7887 reporting the load balancer's activity. When set, the HTML report page will
7888 include a link "refresh"/"stop refresh" so that the user can select whether
7889 he wants automatic refresh of the page or not.
7890
7891 Though this statement alone is enough to enable statistics reporting, it is
7892 recommended to set all other settings in order to avoid relying on default
7893 unobvious parameters.
7894
7895 Example :
7896 # public access (limited to this backend only)
7897 backend public_www
7898 server srv1 192.168.0.1:80
7899 stats enable
7900 stats hide-version
7901 stats scope .
7902 stats uri /admin?stats
7903 stats realm Haproxy\ Statistics
7904 stats auth admin1:AdMiN123
7905 stats auth admin2:AdMiN321
7906
7907 # internal monitoring access (unlimited)
7908 backend private_monitoring
7909 stats enable
7910 stats uri /admin?stats
7911 stats refresh 5s
7912
7913 See also : "stats auth", "stats enable", "stats realm", "stats uri"
7914
7915
7916stats scope { <name> | "." }
7917 Enable statistics and limit access scope
7918 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007919 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007920 Arguments :
7921 <name> is the name of a listen, frontend or backend section to be
7922 reported. The special name "." (a single dot) designates the
7923 section in which the statement appears.
7924
7925 When this statement is specified, only the sections enumerated with this
7926 statement will appear in the report. All other ones will be hidden. This
7927 statement may appear as many times as needed if multiple sections need to be
7928 reported. Please note that the name checking is performed as simple string
7929 comparisons, and that it is never checked that a give section name really
7930 exists.
7931
7932 Though this statement alone is enough to enable statistics reporting, it is
7933 recommended to set all other settings in order to avoid relying on default
7934 unobvious parameters.
7935
7936 Example :
7937 # public access (limited to this backend only)
7938 backend public_www
7939 server srv1 192.168.0.1:80
7940 stats enable
7941 stats hide-version
7942 stats scope .
7943 stats uri /admin?stats
7944 stats realm Haproxy\ Statistics
7945 stats auth admin1:AdMiN123
7946 stats auth admin2:AdMiN321
7947
7948 # internal monitoring access (unlimited)
7949 backend private_monitoring
7950 stats enable
7951 stats uri /admin?stats
7952 stats refresh 5s
7953
7954 See also : "stats auth", "stats enable", "stats realm", "stats uri"
7955
Willy Tarreaud63335a2010-02-26 12:56:52 +01007956
Willy Tarreauc9705a12010-07-27 20:05:50 +02007957stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +01007958 Enable reporting of a description on the statistics page.
7959 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007960 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01007961
Willy Tarreauc9705a12010-07-27 20:05:50 +02007962 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +01007963 description from global section is automatically used instead.
7964
7965 This statement is useful for users that offer shared services to their
7966 customers, where node or description should be different for each customer.
7967
7968 Though this statement alone is enough to enable statistics reporting, it is
7969 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007970 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +01007971
7972 Example :
7973 # internal monitoring access (unlimited)
7974 backend private_monitoring
7975 stats enable
7976 stats show-desc Master node for Europe, Asia, Africa
7977 stats uri /admin?stats
7978 stats refresh 5s
7979
7980 See also: "show-node", "stats enable", "stats uri" and "description" in
7981 global section.
7982
7983
7984stats show-legends
Willy Tarreaued2119c2014-04-24 22:10:39 +02007985 Enable reporting additional information on the statistics page
7986 May be used in sections : defaults | frontend | listen | backend
7987 yes | yes | yes | yes
7988 Arguments : none
7989
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007990 Enable reporting additional information on the statistics page :
Willy Tarreaud63335a2010-02-26 12:56:52 +01007991 - cap: capabilities (proxy)
7992 - mode: one of tcp, http or health (proxy)
7993 - id: SNMP ID (proxy, socket, server)
7994 - IP (socket, server)
7995 - cookie (backend, server)
7996
7997 Though this statement alone is enough to enable statistics reporting, it is
7998 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007999 unobvious parameters. Default behaviour is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008000
8001 See also: "stats enable", "stats uri".
8002
8003
8004stats show-node [ <name> ]
8005 Enable reporting of a host name on the statistics page.
8006 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008007 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008008 Arguments:
8009 <name> is an optional name to be reported. If unspecified, the
8010 node name from global section is automatically used instead.
8011
8012 This statement is useful for users that offer shared services to their
8013 customers, where node or description might be different on a stats page
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04008014 provided for each customer. Default behaviour is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008015
8016 Though this statement alone is enough to enable statistics reporting, it is
8017 recommended to set all other settings in order to avoid relying on default
8018 unobvious parameters.
8019
8020 Example:
8021 # internal monitoring access (unlimited)
8022 backend private_monitoring
8023 stats enable
8024 stats show-node Europe-1
8025 stats uri /admin?stats
8026 stats refresh 5s
8027
8028 See also: "show-desc", "stats enable", "stats uri", and "node" in global
8029 section.
8030
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008031
8032stats uri <prefix>
8033 Enable statistics and define the URI prefix to access them
8034 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008035 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008036 Arguments :
8037 <prefix> is the prefix of any URI which will be redirected to stats. This
8038 prefix may contain a question mark ('?') to indicate part of a
8039 query string.
8040
8041 The statistics URI is intercepted on the relayed traffic, so it appears as a
8042 page within the normal application. It is strongly advised to ensure that the
8043 selected URI will never appear in the application, otherwise it will never be
8044 possible to reach it in the application.
8045
8046 The default URI compiled in haproxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008047 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008048 It is generally a good idea to include a question mark in the URI so that
8049 intermediate proxies refrain from caching the results. Also, since any string
8050 beginning with the prefix will be accepted as a stats request, the question
8051 mark helps ensuring that no valid URI will begin with the same words.
8052
8053 It is sometimes very convenient to use "/" as the URI prefix, and put that
8054 statement in a "listen" instance of its own. That makes it easy to dedicate
8055 an address or a port to statistics only.
8056
8057 Though this statement alone is enough to enable statistics reporting, it is
8058 recommended to set all other settings in order to avoid relying on default
8059 unobvious parameters.
8060
8061 Example :
8062 # public access (limited to this backend only)
8063 backend public_www
8064 server srv1 192.168.0.1:80
8065 stats enable
8066 stats hide-version
8067 stats scope .
8068 stats uri /admin?stats
8069 stats realm Haproxy\ Statistics
8070 stats auth admin1:AdMiN123
8071 stats auth admin2:AdMiN321
8072
8073 # internal monitoring access (unlimited)
8074 backend private_monitoring
8075 stats enable
8076 stats uri /admin?stats
8077 stats refresh 5s
8078
8079 See also : "stats auth", "stats enable", "stats realm"
8080
8081
Willy Tarreaud63335a2010-02-26 12:56:52 +01008082stick match <pattern> [table <table>] [{if | unless} <cond>]
8083 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008084 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +01008085 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008086
8087 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008088 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008089 describes what elements of the incoming request or connection
8090 will be analysed in the hope to find a matching entry in a
8091 stickiness table. This rule is mandatory.
8092
8093 <table> is an optional stickiness table name. If unspecified, the same
8094 backend's table is used. A stickiness table is declared using
8095 the "stick-table" statement.
8096
8097 <cond> is an optional matching condition. It makes it possible to match
8098 on a certain criterion only when other conditions are met (or
8099 not met). For instance, it could be used to match on a source IP
8100 address except when a request passes through a known proxy, in
8101 which case we'd match on a header containing that IP address.
8102
8103 Some protocols or applications require complex stickiness rules and cannot
8104 always simply rely on cookies nor hashing. The "stick match" statement
8105 describes a rule to extract the stickiness criterion from an incoming request
8106 or connection. See section 7 for a complete list of possible patterns and
8107 transformation rules.
8108
8109 The table has to be declared using the "stick-table" statement. It must be of
8110 a type compatible with the pattern. By default it is the one which is present
8111 in the same backend. It is possible to share a table with other backends by
8112 referencing it using the "table" keyword. If another table is referenced,
8113 the server's ID inside the backends are used. By default, all server IDs
8114 start at 1 in each backend, so the server ordering is enough. But in case of
8115 doubt, it is highly recommended to force server IDs using their "id" setting.
8116
8117 It is possible to restrict the conditions where a "stick match" statement
8118 will apply, using "if" or "unless" followed by a condition. See section 7 for
8119 ACL based conditions.
8120
8121 There is no limit on the number of "stick match" statements. The first that
8122 applies and matches will cause the request to be directed to the same server
8123 as was used for the request which created the entry. That way, multiple
8124 matches can be used as fallbacks.
8125
8126 The stick rules are checked after the persistence cookies, so they will not
8127 affect stickiness if a cookie has already been used to select a server. That
8128 way, it becomes very easy to insert cookies and match on IP addresses in
8129 order to maintain stickiness between HTTP and HTTPS.
8130
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008131 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8132 unless you know what you do : memory is not shared between the
8133 processes, which can result in random behaviours.
8134
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008135 Example :
8136 # forward SMTP users to the same server they just used for POP in the
8137 # last 30 minutes
8138 backend pop
8139 mode tcp
8140 balance roundrobin
8141 stick store-request src
8142 stick-table type ip size 200k expire 30m
8143 server s1 192.168.1.1:110
8144 server s2 192.168.1.1:110
8145
8146 backend smtp
8147 mode tcp
8148 balance roundrobin
8149 stick match src table pop
8150 server s1 192.168.1.1:25
8151 server s2 192.168.1.1:25
8152
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008153 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02008154 about ACLs and samples fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008155
8156
8157stick on <pattern> [table <table>] [{if | unless} <condition>]
8158 Define a request pattern to associate a user to a server
8159 May be used in sections : defaults | frontend | listen | backend
8160 no | no | yes | yes
8161
8162 Note : This form is exactly equivalent to "stick match" followed by
8163 "stick store-request", all with the same arguments. Please refer
8164 to both keywords for details. It is only provided as a convenience
8165 for writing more maintainable configurations.
8166
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008167 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8168 unless you know what you do : memory is not shared between the
8169 processes, which can result in random behaviours.
8170
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008171 Examples :
8172 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +01008173 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008174
8175 # ...is strictly equivalent to this one :
8176 stick match src table pop if !localhost
8177 stick store-request src table pop if !localhost
8178
8179
8180 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
8181 # well as HTTP without cookie. Share the same table between both accesses.
8182 backend http
8183 mode http
8184 balance roundrobin
8185 stick on src table https
8186 cookie SRV insert indirect nocache
8187 server s1 192.168.1.1:80 cookie s1
8188 server s2 192.168.1.1:80 cookie s2
8189
8190 backend https
8191 mode tcp
8192 balance roundrobin
8193 stick-table type ip size 200k expire 30m
8194 stick on src
8195 server s1 192.168.1.1:443
8196 server s2 192.168.1.1:443
8197
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008198 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008199
8200
8201stick store-request <pattern> [table <table>] [{if | unless} <condition>]
8202 Define a request pattern used to create an entry in a stickiness table
8203 May be used in sections : defaults | frontend | listen | backend
8204 no | no | yes | yes
8205
8206 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008207 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008208 describes what elements of the incoming request or connection
8209 will be analysed, extracted and stored in the table once a
8210 server is selected.
8211
8212 <table> is an optional stickiness table name. If unspecified, the same
8213 backend's table is used. A stickiness table is declared using
8214 the "stick-table" statement.
8215
8216 <cond> is an optional storage condition. It makes it possible to store
8217 certain criteria only when some conditions are met (or not met).
8218 For instance, it could be used to store the source IP address
8219 except when the request passes through a known proxy, in which
8220 case we'd store a converted form of a header containing that IP
8221 address.
8222
8223 Some protocols or applications require complex stickiness rules and cannot
8224 always simply rely on cookies nor hashing. The "stick store-request" statement
8225 describes a rule to decide what to extract from the request and when to do
8226 it, in order to store it into a stickiness table for further requests to
8227 match it using the "stick match" statement. Obviously the extracted part must
8228 make sense and have a chance to be matched in a further request. Storing a
8229 client's IP address for instance often makes sense. Storing an ID found in a
8230 URL parameter also makes sense. Storing a source port will almost never make
8231 any sense because it will be randomly matched. See section 7 for a complete
8232 list of possible patterns and transformation rules.
8233
8234 The table has to be declared using the "stick-table" statement. It must be of
8235 a type compatible with the pattern. By default it is the one which is present
8236 in the same backend. It is possible to share a table with other backends by
8237 referencing it using the "table" keyword. If another table is referenced,
8238 the server's ID inside the backends are used. By default, all server IDs
8239 start at 1 in each backend, so the server ordering is enough. But in case of
8240 doubt, it is highly recommended to force server IDs using their "id" setting.
8241
8242 It is possible to restrict the conditions where a "stick store-request"
8243 statement will apply, using "if" or "unless" followed by a condition. This
8244 condition will be evaluated while parsing the request, so any criteria can be
8245 used. See section 7 for ACL based conditions.
8246
8247 There is no limit on the number of "stick store-request" statements, but
8248 there is a limit of 8 simultaneous stores per request or response. This
8249 makes it possible to store up to 8 criteria, all extracted from either the
8250 request or the response, regardless of the number of rules. Only the 8 first
8251 ones which match will be kept. Using this, it is possible to feed multiple
8252 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01008253 another protocol or access method. Using multiple store-request rules with
8254 the same table is possible and may be used to find the best criterion to rely
8255 on, by arranging the rules by decreasing preference order. Only the first
8256 extracted criterion for a given table will be stored. All subsequent store-
8257 request rules referencing the same table will be skipped and their ACLs will
8258 not be evaluated.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008259
8260 The "store-request" rules are evaluated once the server connection has been
8261 established, so that the table will contain the real server that processed
8262 the request.
8263
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008264 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8265 unless you know what you do : memory is not shared between the
8266 processes, which can result in random behaviours.
8267
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008268 Example :
8269 # forward SMTP users to the same server they just used for POP in the
8270 # last 30 minutes
8271 backend pop
8272 mode tcp
8273 balance roundrobin
8274 stick store-request src
8275 stick-table type ip size 200k expire 30m
8276 server s1 192.168.1.1:110
8277 server s2 192.168.1.1:110
8278
8279 backend smtp
8280 mode tcp
8281 balance roundrobin
8282 stick match src table pop
8283 server s1 192.168.1.1:25
8284 server s2 192.168.1.1:25
8285
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008286 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02008287 about ACLs and sample fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008288
8289
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008290stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Emeric Brunf099e792010-09-27 12:05:28 +02008291 size <size> [expire <expire>] [nopurge] [peers <peersect>]
8292 [store <data_type>]*
Godbach64cef792013-12-04 16:08:22 +08008293 Configure the stickiness table for the current section
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008294 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +02008295 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008296
8297 Arguments :
8298 ip a table declared with "type ip" will only store IPv4 addresses.
8299 This form is very compact (about 50 bytes per entry) and allows
8300 very fast entry lookup and stores with almost no overhead. This
8301 is mainly used to store client source IP addresses.
8302
David du Colombier9a6d3c92011-03-17 10:40:24 +01008303 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
8304 This form is very compact (about 60 bytes per entry) and allows
8305 very fast entry lookup and stores with almost no overhead. This
8306 is mainly used to store client source IP addresses.
8307
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008308 integer a table declared with "type integer" will store 32bit integers
8309 which can represent a client identifier found in a request for
8310 instance.
8311
8312 string a table declared with "type string" will store substrings of up
8313 to <len> characters. If the string provided by the pattern
8314 extractor is larger than <len>, it will be truncated before
8315 being stored. During matching, at most <len> characters will be
8316 compared between the string in the table and the extracted
8317 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008318 to 32 characters.
8319
8320 binary a table declared with "type binary" will store binary blocks
8321 of <len> bytes. If the block provided by the pattern
8322 extractor is larger than <len>, it will be truncated before
Willy Tarreaube722a22014-06-13 16:31:59 +02008323 being stored. If the block provided by the sample expression
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008324 is shorter than <len>, it will be padded by 0. When not
8325 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008326
8327 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008328 "string" type table (See type "string" above). Or the number
8329 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008330 changing this parameter as memory usage will proportionally
8331 increase.
8332
8333 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +01008334 value directly impacts memory usage. Count approximately
8335 50 bytes per entry, plus the size of a string if any. The size
8336 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008337
8338 [nopurge] indicates that we refuse to purge older entries when the table
8339 is full. When not specified and the table is full when haproxy
8340 wants to store an entry in it, it will flush a few of the oldest
8341 entries in order to release some space for the new ones. This is
8342 most often the desired behaviour. In some specific cases, it
8343 be desirable to refuse new entries instead of purging the older
8344 ones. That may be the case when the amount of data to store is
8345 far above the hardware limits and we prefer not to offer access
8346 to new clients than to reject the ones already connected. When
8347 using this parameter, be sure to properly set the "expire"
8348 parameter (see below).
8349
Emeric Brunf099e792010-09-27 12:05:28 +02008350 <peersect> is the name of the peers section to use for replication. Entries
8351 which associate keys to server IDs are kept synchronized with
8352 the remote peers declared in this section. All entries are also
8353 automatically learned from the local peer (old process) during a
8354 soft restart.
8355
Willy Tarreau1abc6732015-05-01 19:21:02 +02008356 NOTE : each peers section may be referenced only by tables
8357 belonging to the same unique process.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008358
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008359 <expire> defines the maximum duration of an entry in the table since it
8360 was last created, refreshed or matched. The expiration delay is
8361 defined using the standard time format, similarly as the various
8362 timeouts. The maximum duration is slightly above 24 days. See
8363 section 2.2 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +02008364 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008365 be removed once full. Be sure not to use the "nopurge" parameter
8366 if not expiration delay is specified.
8367
Willy Tarreau08d5f982010-06-06 13:34:54 +02008368 <data_type> is used to store additional information in the stick-table. This
8369 may be used by ACLs in order to control various criteria related
8370 to the activity of the client matching the stick-table. For each
8371 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +02008372 that the additional data can fit. Several data types may be
8373 stored with an entry. Multiple data types may be specified after
8374 the "store" keyword, as a comma-separated list. Alternatively,
8375 it is possible to repeat the "store" keyword followed by one or
8376 several data types. Except for the "server_id" type which is
8377 automatically detected and enabled, all data types must be
8378 explicitly declared to be stored. If an ACL references a data
8379 type which is not stored, the ACL will simply not match. Some
8380 data types require an argument which must be passed just after
8381 the type between parenthesis. See below for the supported data
8382 types and their arguments.
8383
8384 The data types that can be stored with an entry are the following :
8385 - server_id : this is an integer which holds the numeric ID of the server a
8386 request was assigned to. It is used by the "stick match", "stick store",
8387 and "stick on" rules. It is automatically enabled when referenced.
8388
8389 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
8390 integer which may be used for anything. Most of the time it will be used
8391 to put a special tag on some entries, for instance to note that a
8392 specific behaviour was detected and must be known for future matches.
8393
Willy Tarreauba2ffd12013-05-29 15:54:14 +02008394 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
8395 over a period. It is a positive 32-bit integer integer which may be used
8396 for anything. Just like <gpc0>, it counts events, but instead of keeping
8397 a cumulative count, it maintains the rate at which the counter is
8398 incremented. Most of the time it will be used to measure the frequency of
8399 occurrence of certain events (eg: requests to a specific URL).
8400
Willy Tarreauc9705a12010-07-27 20:05:50 +02008401 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
8402 the absolute number of connections received from clients which matched
8403 this entry. It does not mean the connections were accepted, just that
8404 they were received.
8405
8406 - conn_cur : Current Connections. It is a positive 32-bit integer which
8407 stores the concurrent connection counts for the entry. It is incremented
8408 once an incoming connection matches the entry, and decremented once the
8409 connection leaves. That way it is possible to know at any time the exact
8410 number of concurrent connections for an entry.
8411
8412 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8413 integer parameter <period> which indicates in milliseconds the length
8414 of the period over which the average is measured. It reports the average
8415 incoming connection rate over that period, in connections per period. The
8416 result is an integer which can be matched using ACLs.
8417
8418 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
8419 the absolute number of sessions received from clients which matched this
8420 entry. A session is a connection that was accepted by the layer 4 rules.
8421
8422 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8423 integer parameter <period> which indicates in milliseconds the length
8424 of the period over which the average is measured. It reports the average
8425 incoming session rate over that period, in sessions per period. The
8426 result is an integer which can be matched using ACLs.
8427
8428 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
8429 counts the absolute number of HTTP requests received from clients which
8430 matched this entry. It does not matter whether they are valid requests or
8431 not. Note that this is different from sessions when keep-alive is used on
8432 the client side.
8433
8434 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8435 integer parameter <period> which indicates in milliseconds the length
8436 of the period over which the average is measured. It reports the average
8437 HTTP request rate over that period, in requests per period. The result is
8438 an integer which can be matched using ACLs. It does not matter whether
8439 they are valid requests or not. Note that this is different from sessions
8440 when keep-alive is used on the client side.
8441
8442 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
8443 counts the absolute number of HTTP requests errors induced by clients
8444 which matched this entry. Errors are counted on invalid and truncated
8445 requests, as well as on denied or tarpitted requests, and on failed
8446 authentications. If the server responds with 4xx, then the request is
8447 also counted as an error since it's an error triggered by the client
8448 (eg: vulnerability scan).
8449
8450 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8451 integer parameter <period> which indicates in milliseconds the length
8452 of the period over which the average is measured. It reports the average
8453 HTTP request error rate over that period, in requests per period (see
8454 http_err_cnt above for what is accounted as an error). The result is an
8455 integer which can be matched using ACLs.
8456
8457 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
8458 integer which counts the cumulated amount of bytes received from clients
8459 which matched this entry. Headers are included in the count. This may be
8460 used to limit abuse of upload features on photo or video servers.
8461
8462 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8463 integer parameter <period> which indicates in milliseconds the length
8464 of the period over which the average is measured. It reports the average
8465 incoming bytes rate over that period, in bytes per period. It may be used
8466 to detect users which upload too much and too fast. Warning: with large
8467 uploads, it is possible that the amount of uploaded data will be counted
8468 once upon termination, thus causing spikes in the average transfer speed
8469 instead of having a smooth one. This may partially be smoothed with
8470 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
8471 recommended for better fairness.
8472
8473 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
8474 integer which counts the cumulated amount of bytes sent to clients which
8475 matched this entry. Headers are included in the count. This may be used
8476 to limit abuse of bots sucking the whole site.
8477
8478 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
8479 an integer parameter <period> which indicates in milliseconds the length
8480 of the period over which the average is measured. It reports the average
8481 outgoing bytes rate over that period, in bytes per period. It may be used
8482 to detect users which download too much and too fast. Warning: with large
8483 transfers, it is possible that the amount of transferred data will be
8484 counted once upon termination, thus causing spikes in the average
8485 transfer speed instead of having a smooth one. This may partially be
8486 smoothed with "option contstats" though this is not perfect yet. Use of
8487 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +02008488
Willy Tarreauc00cdc22010-06-06 16:48:26 +02008489 There is only one stick-table per proxy. At the moment of writing this doc,
8490 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008491 to be required, simply create a dummy backend with a stick-table in it and
8492 reference it.
8493
8494 It is important to understand that stickiness based on learning information
8495 has some limitations, including the fact that all learned associations are
Baptiste Assmann123ff042016-03-06 23:29:28 +01008496 lost upon restart unless peers are properly configured to transfer such
8497 information upon restart (recommended). In general it can be good as a
8498 complement but not always as an exclusive stickiness.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008499
Willy Tarreauc9705a12010-07-27 20:05:50 +02008500 Last, memory requirements may be important when storing many data types.
8501 Indeed, storing all indicators above at once in each entry requires 116 bytes
8502 per entry, or 116 MB for a 1-million entries table. This is definitely not
8503 something that can be ignored.
8504
8505 Example:
8506 # Keep track of counters of up to 1 million IP addresses over 5 minutes
8507 # and store a general purpose counter and the average connection rate
8508 # computed over a sliding window of 30 seconds.
8509 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
8510
8511 See also : "stick match", "stick on", "stick store-request", section 2.2
David du Colombiera13d1b92011-03-17 10:40:22 +01008512 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008513
8514
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008515stick store-response <pattern> [table <table>] [{if | unless} <condition>]
Baptiste Assmann2f2d2ec2016-03-06 23:27:24 +01008516 Define a response pattern used to create an entry in a stickiness table
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008517 May be used in sections : defaults | frontend | listen | backend
8518 no | no | yes | yes
8519
8520 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008521 <pattern> is a sample expression rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008522 describes what elements of the response or connection will
8523 be analysed, extracted and stored in the table once a
8524 server is selected.
8525
8526 <table> is an optional stickiness table name. If unspecified, the same
8527 backend's table is used. A stickiness table is declared using
8528 the "stick-table" statement.
8529
8530 <cond> is an optional storage condition. It makes it possible to store
8531 certain criteria only when some conditions are met (or not met).
8532 For instance, it could be used to store the SSL session ID only
8533 when the response is a SSL server hello.
8534
8535 Some protocols or applications require complex stickiness rules and cannot
8536 always simply rely on cookies nor hashing. The "stick store-response"
8537 statement describes a rule to decide what to extract from the response and
8538 when to do it, in order to store it into a stickiness table for further
8539 requests to match it using the "stick match" statement. Obviously the
8540 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008541 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008542 See section 7 for a complete list of possible patterns and transformation
8543 rules.
8544
8545 The table has to be declared using the "stick-table" statement. It must be of
8546 a type compatible with the pattern. By default it is the one which is present
8547 in the same backend. It is possible to share a table with other backends by
8548 referencing it using the "table" keyword. If another table is referenced,
8549 the server's ID inside the backends are used. By default, all server IDs
8550 start at 1 in each backend, so the server ordering is enough. But in case of
8551 doubt, it is highly recommended to force server IDs using their "id" setting.
8552
8553 It is possible to restrict the conditions where a "stick store-response"
8554 statement will apply, using "if" or "unless" followed by a condition. This
8555 condition will be evaluated while parsing the response, so any criteria can
8556 be used. See section 7 for ACL based conditions.
8557
8558 There is no limit on the number of "stick store-response" statements, but
8559 there is a limit of 8 simultaneous stores per request or response. This
8560 makes it possible to store up to 8 criteria, all extracted from either the
8561 request or the response, regardless of the number of rules. Only the 8 first
8562 ones which match will be kept. Using this, it is possible to feed multiple
8563 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01008564 another protocol or access method. Using multiple store-response rules with
8565 the same table is possible and may be used to find the best criterion to rely
8566 on, by arranging the rules by decreasing preference order. Only the first
8567 extracted criterion for a given table will be stored. All subsequent store-
8568 response rules referencing the same table will be skipped and their ACLs will
8569 not be evaluated. However, even if a store-request rule references a table, a
8570 store-response rule may also use the same table. This means that each table
8571 may learn exactly one element from the request and one element from the
8572 response at once.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008573
8574 The table will contain the real server that processed the request.
8575
8576 Example :
8577 # Learn SSL session ID from both request and response and create affinity.
8578 backend https
8579 mode tcp
8580 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +02008581 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008582 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008583
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008584 acl clienthello req_ssl_hello_type 1
8585 acl serverhello rep_ssl_hello_type 2
8586
8587 # use tcp content accepts to detects ssl client and server hello.
8588 tcp-request inspect-delay 5s
8589 tcp-request content accept if clienthello
8590
8591 # no timeout on response inspect delay by default.
8592 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008593
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008594 # SSL session ID (SSLID) may be present on a client or server hello.
8595 # Its length is coded on 1 byte at offset 43 and its value starts
8596 # at offset 44.
8597
8598 # Match and learn on request if client hello.
8599 stick on payload_lv(43,1) if clienthello
8600
8601 # Learn on response if server hello.
8602 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +02008603
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008604 server s1 192.168.1.1:443
8605 server s2 192.168.1.1:443
8606
8607 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
8608 extraction.
8609
8610
Willy Tarreau938c7fe2014-04-25 14:21:39 +02008611tcp-check connect [params*]
8612 Opens a new connection
8613 May be used in sections: defaults | frontend | listen | backend
8614 no | no | yes | yes
8615
8616 When an application lies on more than a single TCP port or when HAProxy
8617 load-balance many services in a single backend, it makes sense to probe all
8618 the services individually before considering a server as operational.
8619
8620 When there are no TCP port configured on the server line neither server port
8621 directive, then the 'tcp-check connect port <port>' must be the first step
8622 of the sequence.
8623
8624 In a tcp-check ruleset a 'connect' is required, it is also mandatory to start
8625 the ruleset with a 'connect' rule. Purpose is to ensure admin know what they
8626 do.
8627
8628 Parameters :
8629 They are optional and can be used to describe how HAProxy should open and
8630 use the TCP connection.
8631
8632 port if not set, check port or server port is used.
8633 It tells HAProxy where to open the connection to.
8634 <port> must be a valid TCP port source integer, from 1 to 65535.
8635
8636 send-proxy send a PROXY protocol string
8637
8638 ssl opens a ciphered connection
8639
8640 Examples:
8641 # check HTTP and HTTPs services on a server.
8642 # first open port 80 thanks to server line port directive, then
8643 # tcp-check opens port 443, ciphered and run a request on it:
8644 option tcp-check
8645 tcp-check connect
8646 tcp-check send GET\ /\ HTTP/1.0\r\n
8647 tcp-check send Host:\ haproxy.1wt.eu\r\n
8648 tcp-check send \r\n
8649 tcp-check expect rstring (2..|3..)
8650 tcp-check connect port 443 ssl
8651 tcp-check send GET\ /\ HTTP/1.0\r\n
8652 tcp-check send Host:\ haproxy.1wt.eu\r\n
8653 tcp-check send \r\n
8654 tcp-check expect rstring (2..|3..)
8655 server www 10.0.0.1 check port 80
8656
8657 # check both POP and IMAP from a single server:
8658 option tcp-check
8659 tcp-check connect port 110
8660 tcp-check expect string +OK\ POP3\ ready
8661 tcp-check connect port 143
8662 tcp-check expect string *\ OK\ IMAP4\ ready
8663 server mail 10.0.0.1 check
8664
8665 See also : "option tcp-check", "tcp-check send", "tcp-check expect"
8666
8667
8668tcp-check expect [!] <match> <pattern>
8669 Specify data to be collected and analysed during a generic health check
8670 May be used in sections: defaults | frontend | listen | backend
8671 no | no | yes | yes
8672
8673 Arguments :
8674 <match> is a keyword indicating how to look for a specific pattern in the
8675 response. The keyword may be one of "string", "rstring" or
8676 binary.
8677 The keyword may be preceded by an exclamation mark ("!") to negate
8678 the match. Spaces are allowed between the exclamation mark and the
8679 keyword. See below for more details on the supported keywords.
8680
8681 <pattern> is the pattern to look for. It may be a string or a regular
8682 expression. If the pattern contains spaces, they must be escaped
8683 with the usual backslash ('\').
8684 If the match is set to binary, then the pattern must be passed as
8685 a serie of hexadecimal digits in an even number. Each sequence of
8686 two digits will represent a byte. The hexadecimal digits may be
8687 used upper or lower case.
8688
8689
8690 The available matches are intentionally similar to their http-check cousins :
8691
8692 string <string> : test the exact string matches in the response buffer.
8693 A health check response will be considered valid if the
8694 response's buffer contains this exact string. If the
8695 "string" keyword is prefixed with "!", then the response
8696 will be considered invalid if the body contains this
8697 string. This can be used to look for a mandatory pattern
8698 in a protocol response, or to detect a failure when a
8699 specific error appears in a protocol banner.
8700
8701 rstring <regex> : test a regular expression on the response buffer.
8702 A health check response will be considered valid if the
8703 response's buffer matches this expression. If the
8704 "rstring" keyword is prefixed with "!", then the response
8705 will be considered invalid if the body matches the
8706 expression.
8707
8708 binary <hexstring> : test the exact string in its hexadecimal form matches
8709 in the response buffer. A health check response will
8710 be considered valid if the response's buffer contains
8711 this exact hexadecimal string.
8712 Purpose is to match data on binary protocols.
8713
8714 It is important to note that the responses will be limited to a certain size
8715 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
8716 Thus, too large responses may not contain the mandatory pattern when using
8717 "string", "rstring" or binary. If a large response is absolutely required, it
8718 is possible to change the default max size by setting the global variable.
8719 However, it is worth keeping in mind that parsing very large responses can
8720 waste some CPU cycles, especially when regular expressions are used, and that
8721 it is always better to focus the checks on smaller resources. Also, in its
8722 current state, the check will not find any string nor regex past a null
8723 character in the response. Similarly it is not possible to request matching
8724 the null character.
8725
8726 Examples :
8727 # perform a POP check
8728 option tcp-check
8729 tcp-check expect string +OK\ POP3\ ready
8730
8731 # perform an IMAP check
8732 option tcp-check
8733 tcp-check expect string *\ OK\ IMAP4\ ready
8734
8735 # look for the redis master server
8736 option tcp-check
8737 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02008738 tcp-check expect string +PONG
Willy Tarreau938c7fe2014-04-25 14:21:39 +02008739 tcp-check send info\ replication\r\n
8740 tcp-check expect string role:master
8741 tcp-check send QUIT\r\n
8742 tcp-check expect string +OK
8743
8744
8745 See also : "option tcp-check", "tcp-check connect", "tcp-check send",
8746 "tcp-check send-binary", "http-check expect", tune.chksize
8747
8748
8749tcp-check send <data>
8750 Specify a string to be sent as a question during a generic health check
8751 May be used in sections: defaults | frontend | listen | backend
8752 no | no | yes | yes
8753
8754 <data> : the data to be sent as a question during a generic health check
8755 session. For now, <data> must be a string.
8756
8757 Examples :
8758 # look for the redis master server
8759 option tcp-check
8760 tcp-check send info\ replication\r\n
8761 tcp-check expect string role:master
8762
8763 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
8764 "tcp-check send-binary", tune.chksize
8765
8766
8767tcp-check send-binary <hexastring>
8768 Specify an hexa digits string to be sent as a binary question during a raw
8769 tcp health check
8770 May be used in sections: defaults | frontend | listen | backend
8771 no | no | yes | yes
8772
8773 <data> : the data to be sent as a question during a generic health check
8774 session. For now, <data> must be a string.
8775 <hexastring> : test the exact string in its hexadecimal form matches in the
8776 response buffer. A health check response will be considered
8777 valid if the response's buffer contains this exact
8778 hexadecimal string.
8779 Purpose is to send binary data to ask on binary protocols.
8780
8781 Examples :
8782 # redis check in binary
8783 option tcp-check
8784 tcp-check send-binary 50494e470d0a # PING\r\n
8785 tcp-check expect binary 2b504F4e47 # +PONG
8786
8787
8788 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
8789 "tcp-check send", tune.chksize
8790
8791
Willy Tarreaue9656522010-08-17 15:40:09 +02008792tcp-request connection <action> [{if | unless} <condition>]
8793 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +02008794 May be used in sections : defaults | frontend | listen | backend
8795 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +02008796 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02008797 <action> defines the action to perform if the condition applies. See
8798 below.
Willy Tarreau1a687942010-05-23 22:40:30 +02008799
Willy Tarreaue9656522010-08-17 15:40:09 +02008800 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008801
8802 Immediately after acceptance of a new incoming connection, it is possible to
8803 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +02008804 or dropped or have its counters tracked. Those conditions cannot make use of
8805 any data contents because the connection has not been read from yet, and the
8806 buffers are not yet allocated. This is used to selectively and very quickly
8807 accept or drop connections from various sources with a very low overhead. If
8808 some contents need to be inspected in order to take the decision, the
8809 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008810
Willy Tarreaue9656522010-08-17 15:40:09 +02008811 The "tcp-request connection" rules are evaluated in their exact declaration
8812 order. If no rule matches or if there is no rule, the default action is to
8813 accept the incoming connection. There is no specific limit to the number of
8814 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008815
Willy Tarreaua9083d02015-05-08 15:27:59 +02008816 Four types of actions are supported :
Willy Tarreaue9656522010-08-17 15:40:09 +02008817 - accept :
8818 accepts the connection if the condition is true (when used with "if")
8819 or false (when used with "unless"). The first such rule executed ends
8820 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008821
Willy Tarreaue9656522010-08-17 15:40:09 +02008822 - reject :
8823 rejects the connection if the condition is true (when used with "if")
8824 or false (when used with "unless"). The first such rule executed ends
8825 the rules evaluation. Rejected connections do not even become a
8826 session, which is why they are accounted separately for in the stats,
8827 as "denied connections". They are not considered for the session
8828 rate-limit and are not logged either. The reason is that these rules
8829 should only be used to filter extremely high connection rates such as
8830 the ones encountered during a massive DDoS attack. Under these extreme
8831 conditions, the simple action of logging each event would make the
8832 system collapse and would considerably lower the filtering capacity. If
8833 logging is absolutely desired, then "tcp-request content" rules should
Willy Tarreau4f614292016-10-21 17:49:36 +02008834 be used instead, as "tcp-request session" rules will not log either.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008835
Willy Tarreau4f0d9192013-06-11 20:40:55 +02008836 - expect-proxy layer4 :
8837 configures the client-facing connection to receive a PROXY protocol
8838 header before any byte is read from the socket. This is equivalent to
8839 having the "accept-proxy" keyword on the "bind" line, except that using
8840 the TCP rule allows the PROXY protocol to be accepted only for certain
8841 IP address ranges using an ACL. This is convenient when multiple layers
8842 of load balancers are passed through by traffic coming from public
8843 hosts.
8844
Bertrand Jacquin90759682016-06-06 15:35:39 +01008845 - expect-netscaler-cip layer4 :
8846 configures the client-facing connection to receive a NetScaler Client
8847 IP insertion protocol header before any byte is read from the socket.
8848 This is equivalent to having the "accept-netscaler-cip" keyword on the
8849 "bind" line, except that using the TCP rule allows the PROXY protocol
8850 to be accepted only for certain IP address ranges using an ACL. This
8851 is convenient when multiple layers of load balancers are passed
8852 through by traffic coming from public hosts.
8853
Willy Tarreau18bf01e2014-06-13 16:18:52 +02008854 - capture <sample> len <length> :
8855 This only applies to "tcp-request content" rules. It captures sample
8856 expression <sample> from the request buffer, and converts it to a
8857 string of at most <len> characters. The resulting string is stored into
8858 the next request "capture" slot, so it will possibly appear next to
8859 some captured HTTP headers. It will then automatically appear in the
8860 logs, and it will be possible to extract it using sample fetch rules to
8861 feed it into headers or anything. The length should be limited given
8862 that this size will be allocated for each capture during the whole
Willy Tarreaua9083d02015-05-08 15:27:59 +02008863 session life. Please check section 7.3 (Fetching samples) and "capture
8864 request header" for more information.
Willy Tarreau18bf01e2014-06-13 16:18:52 +02008865
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008866 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +02008867 enables tracking of sticky counters from current connection. These
Willy Tarreau09448f72014-06-25 18:12:15 +02008868 rules do not stop evaluation and do not change default action. 3 sets
Willy Tarreaue9656522010-08-17 15:40:09 +02008869 of counters may be simultaneously tracked by the same connection. The
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008870 first "track-sc0" rule executed enables tracking of the counters of the
8871 specified table as the first set. The first "track-sc1" rule executed
Willy Tarreaue9656522010-08-17 15:40:09 +02008872 enables tracking of the counters of the specified table as the second
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008873 set. The first "track-sc2" rule executed enables tracking of the
8874 counters of the specified table as the third set. It is a recommended
8875 practice to use the first set of counters for the per-frontend counters
8876 and the second set for the per-backend ones. But this is just a
8877 guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008878
Willy Tarreaue9656522010-08-17 15:40:09 +02008879 These actions take one or two arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008880 <key> is mandatory, and is a sample expression rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +02008881 in section 7.3. It describes what elements of the incoming
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008882 request or connection will be analysed, extracted, combined,
8883 and used to select which table entry to update the counters.
8884 Note that "tcp-request connection" cannot use content-based
8885 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008886
Willy Tarreaue9656522010-08-17 15:40:09 +02008887 <table> is an optional table to be used instead of the default one,
8888 which is the stick-table declared in the current proxy. All
8889 the counters for the matches and updates for the key will
8890 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008891
Willy Tarreaue9656522010-08-17 15:40:09 +02008892 Once a "track-sc*" rule is executed, the key is looked up in the table
8893 and if it is not found, an entry is allocated for it. Then a pointer to
8894 that entry is kept during all the session's life, and this entry's
8895 counters are updated as often as possible, every time the session's
8896 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008897 Counters are only updated for events that happen after the tracking has
8898 been started. For example, connection counters will not be updated when
8899 tracking layer 7 information, since the connection event happens before
8900 layer7 information is extracted.
8901
Willy Tarreaue9656522010-08-17 15:40:09 +02008902 If the entry tracks concurrent connection counters, one connection is
8903 counted for as long as the entry is tracked, and the entry will not
8904 expire during that time. Tracking counters also provides a performance
8905 advantage over just checking the keys, because only one table lookup is
8906 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008907
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02008908 - sc-inc-gpc0(<sc-id>):
8909 The "sc-inc-gpc0" increments the GPC0 counter according to the sticky
8910 counter designated by <sc-id>. If an error occurs, this action silently
8911 fails and the actions evaluation continues.
8912
Thierry FOURNIER236657b2015-08-19 08:25:14 +02008913 - sc-set-gpt0(<sc-id>) <int>:
8914 This action sets the GPT0 tag according to the sticky counter designated
8915 by <sc-id> and the value of <int>. The expected result is a boolean. If
8916 an error occurs, this action silently fails and the actions evaluation
8917 continues.
8918
William Lallemand2e785f22016-05-25 01:48:42 +02008919 - set-src <expr> :
8920 Is used to set the source IP address to the value of specified
8921 expression. Useful if you want to mask source IP for privacy.
8922 If you want to provide an IP from a HTTP header use "http-request
8923 set-src"
8924
8925 <expr> Is a standard HAProxy expression formed by a sample-fetch
8926 followed by some converters.
8927
8928 Example:
8929
8930 tcp-request connection set-src src,ipmask(24)
8931
Willy Tarreau0c630532016-10-21 17:52:58 +02008932 When possible, set-src preserves the original source port as long as the
8933 address family allows it, otherwise the source port is set to 0.
William Lallemand2e785f22016-05-25 01:48:42 +02008934
William Lallemand44be6402016-05-25 01:51:35 +02008935 - set-src-port <expr> :
8936 Is used to set the source port address to the value of specified
8937 expression.
8938
8939 <expr> Is a standard HAProxy expression formed by a sample-fetch
8940 followed by some converters.
8941
8942 Example:
8943
8944 tcp-request connection set-src-port int(4000)
8945
Willy Tarreau0c630532016-10-21 17:52:58 +02008946 When possible, set-src-port preserves the original source address as long
8947 as the address family supports a port, otherwise it forces the source
8948 address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +02008949
William Lallemand13e9b0c2016-05-25 02:34:07 +02008950 - set-dst <expr> :
8951 Is used to set the destination IP address to the value of specified
8952 expression. Useful if you want to mask IP for privacy in log.
8953 If you want to provide an IP from a HTTP header use "http-request
8954 set-dst". If you want to connect to the new address/port, use
8955 '0.0.0.0:0' as a server address in the backend.
8956
8957 <expr> Is a standard HAProxy expression formed by a sample-fetch
8958 followed by some converters.
8959
8960 Example:
8961
8962 tcp-request connection set-dst dst,ipmask(24)
8963 tcp-request connection set-dst ipv4(10.0.0.1)
8964
Willy Tarreau0c630532016-10-21 17:52:58 +02008965 When possible, set-dst preserves the original destination port as long as
8966 the address family allows it, otherwise the destination port is set to 0.
8967
William Lallemand13e9b0c2016-05-25 02:34:07 +02008968 - set-dst-port <expr> :
8969 Is used to set the destination port address to the value of specified
8970 expression. If you want to connect to the new address/port, use
8971 '0.0.0.0:0' as a server address in the backend.
8972
8973
8974 <expr> Is a standard HAProxy expression formed by a sample-fetch
8975 followed by some converters.
8976
8977 Example:
8978
8979 tcp-request connection set-dst-port int(4000)
8980
Willy Tarreau0c630532016-10-21 17:52:58 +02008981 When possible, set-dst-port preserves the original destination address as
8982 long as the address family supports a port, otherwise it forces the
8983 destination address to IPv4 "0.0.0.0" before rewriting the port.
8984
Willy Tarreau2d392c22015-08-24 01:43:45 +02008985 - "silent-drop" :
8986 This stops the evaluation of the rules and makes the client-facing
8987 connection suddenly disappear using a system-dependant way that tries
8988 to prevent the client from being notified. The effect it then that the
8989 client still sees an established connection while there's none on
8990 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
8991 except that it doesn't use any local resource at all on the machine
8992 running HAProxy. It can resist much higher loads than "tarpit", and
8993 slow down stronger attackers. It is important to undestand the impact
8994 of using this mechanism. All stateful equipments placed between the
8995 client and HAProxy (firewalls, proxies, load balancers) will also keep
8996 the established connection for a long time and may suffer from this
8997 action. On modern Linux systems running with enough privileges, the
8998 TCP_REPAIR socket option is used to block the emission of a TCP
8999 reset. On other systems, the socket's TTL is reduced to 1 so that the
9000 TCP reset doesn't pass the first router, though it's still delivered to
9001 local networks. Do not use it unless you fully understand how it works.
9002
Willy Tarreaue9656522010-08-17 15:40:09 +02009003 Note that the "if/unless" condition is optional. If no condition is set on
9004 the action, it is simply performed unconditionally. That can be useful for
9005 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009006
Willy Tarreaue9656522010-08-17 15:40:09 +02009007 Example: accept all connections from white-listed hosts, reject too fast
9008 connection without counting them, and track accepted connections.
9009 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009010
Willy Tarreaue9656522010-08-17 15:40:09 +02009011 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009012 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009013 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009014
Willy Tarreaue9656522010-08-17 15:40:09 +02009015 Example: accept all connections from white-listed hosts, count all other
9016 connections and reject too fast ones. This results in abusive ones
9017 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009018
Willy Tarreaue9656522010-08-17 15:40:09 +02009019 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009020 tcp-request connection track-sc0 src
9021 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009022
Willy Tarreau4f0d9192013-06-11 20:40:55 +02009023 Example: enable the PROXY protocol for traffic coming from all known proxies.
9024
9025 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
9026
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009027 See section 7 about ACL usage.
9028
Willy Tarreau4f614292016-10-21 17:49:36 +02009029 See also : "tcp-request session", "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009030
9031
Willy Tarreaue9656522010-08-17 15:40:09 +02009032tcp-request content <action> [{if | unless} <condition>]
9033 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +02009034 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02009035 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +02009036 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02009037 <action> defines the action to perform if the condition applies. See
9038 below.
Willy Tarreau62644772008-07-16 18:36:06 +02009039
Willy Tarreaue9656522010-08-17 15:40:09 +02009040 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +02009041
Willy Tarreaue9656522010-08-17 15:40:09 +02009042 A request's contents can be analysed at an early stage of request processing
9043 called "TCP content inspection". During this stage, ACL-based rules are
9044 evaluated every time the request contents are updated, until either an
9045 "accept" or a "reject" rule matches, or the TCP request inspection delay
9046 expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +02009047
Willy Tarreaue9656522010-08-17 15:40:09 +02009048 The first difference between these rules and "tcp-request connection" rules
9049 is that "tcp-request content" rules can make use of contents to take a
9050 decision. Most often, these decisions will consider a protocol recognition or
9051 validity. The second difference is that content-based rules can be used in
Willy Tarreauf3338342014-01-28 21:40:28 +01009052 both frontends and backends. In case of HTTP keep-alive with the client, all
9053 tcp-request content rules are evaluated again, so haproxy keeps a record of
9054 what sticky counters were assigned by a "tcp-request connection" versus a
9055 "tcp-request content" rule, and flushes all the content-related ones after
9056 processing an HTTP request, so that they may be evaluated again by the rules
9057 being evaluated again for the next request. This is of particular importance
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009058 when the rule tracks some L7 information or when it is conditioned by an
Willy Tarreauf3338342014-01-28 21:40:28 +01009059 L7-based ACL, since tracking may change between requests.
Willy Tarreau62644772008-07-16 18:36:06 +02009060
Willy Tarreaue9656522010-08-17 15:40:09 +02009061 Content-based rules are evaluated in their exact declaration order. If no
9062 rule matches or if there is no rule, the default action is to accept the
9063 contents. There is no specific limit to the number of rules which may be
9064 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +02009065
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009066 Several types of actions are supported :
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009067 - accept : the request is accepted
9068 - reject : the request is rejected and the connection is closed
9069 - capture : the specified sample expression is captured
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009070 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02009071 - sc-inc-gpc0(<sc-id>)
Thierry Fournierb9125672016-03-29 19:34:37 +02009072 - sc-set-gpt0(<sc-id>) <int>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009073 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +01009074 - unset-var(<var-name>)
Willy Tarreau2d392c22015-08-24 01:43:45 +02009075 - silent-drop
Willy Tarreau62644772008-07-16 18:36:06 +02009076
Willy Tarreaue9656522010-08-17 15:40:09 +02009077 They have the same meaning as their counter-parts in "tcp-request connection"
9078 so please refer to that section for a complete description.
Willy Tarreau62644772008-07-16 18:36:06 +02009079
Willy Tarreauf3338342014-01-28 21:40:28 +01009080 While there is nothing mandatory about it, it is recommended to use the
9081 track-sc0 in "tcp-request connection" rules, track-sc1 for "tcp-request
9082 content" rules in the frontend, and track-sc2 for "tcp-request content"
9083 rules in the backend, because that makes the configuration more readable
9084 and easier to troubleshoot, but this is just a guideline and all counters
9085 may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +02009086
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009087 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +02009088 the action, it is simply performed unconditionally. That can be useful for
9089 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +02009090
Willy Tarreaue9656522010-08-17 15:40:09 +02009091 It is perfectly possible to match layer 7 contents with "tcp-request content"
Willy Tarreauc0239e02012-04-16 14:42:55 +02009092 rules, since HTTP-specific ACL matches are able to preliminarily parse the
9093 contents of a buffer before extracting the required data. If the buffered
9094 contents do not parse as a valid HTTP message, then the ACL does not match.
9095 The parser which is involved there is exactly the same as for all other HTTP
Willy Tarreauf3338342014-01-28 21:40:28 +01009096 processing, so there is no risk of parsing something differently. In an HTTP
9097 backend connected to from an HTTP frontend, it is guaranteed that HTTP
9098 contents will always be immediately present when the rule is evaluated first.
Willy Tarreau62644772008-07-16 18:36:06 +02009099
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009100 Tracking layer7 information is also possible provided that the information
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02009101 are present when the rule is processed. The rule processing engine is able to
9102 wait until the inspect delay expires when the data to be tracked is not yet
9103 available.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009104
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009105 The "set-var" is used to set the content of a variable. The variable is
Willy Tarreau4f614292016-10-21 17:49:36 +02009106 declared inline. For "tcp-request session" rules, only session-level
9107 variables can be used, without any layer7 contents.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009108
Daniel Schneller0b547052016-03-21 20:46:57 +01009109 <var-name> The name of the variable starts with an indication about
9110 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +01009111 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +01009112 "sess" : the variable is shared with the whole session
9113 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009114 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +01009115 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009116 processing
Daniel Schneller0b547052016-03-21 20:46:57 +01009117 "res" : the variable is shared only during response
9118 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009119 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +01009120 The name may only contain characters 'a-z', 'A-Z', '0-9',
9121 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009122
9123 <expr> Is a standard HAProxy expression formed by a sample-fetch
9124 followed by some converters.
9125
Christopher Faulet85d79c92016-11-09 16:54:56 +01009126 The "unset-var" is used to unset a variable. See above for details about
9127 <var-name>.
9128
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009129 Example:
9130
9131 tcp-request content set-var(sess.my_var) src
Christopher Faulet85d79c92016-11-09 16:54:56 +01009132 tcp-request content unset-var(sess.my_var2)
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009133
Willy Tarreau62644772008-07-16 18:36:06 +02009134 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +02009135 # Accept HTTP requests containing a Host header saying "example.com"
9136 # and reject everything else.
9137 acl is_host_com hdr(Host) -i example.com
9138 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +02009139 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +02009140 tcp-request content reject
9141
9142 Example:
Willy Tarreau62644772008-07-16 18:36:06 +02009143 # reject SMTP connection if client speaks first
9144 tcp-request inspect-delay 30s
9145 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009146 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +02009147
9148 # Forward HTTPS connection only if client speaks
9149 tcp-request inspect-delay 30s
9150 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009151 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +02009152 tcp-request content reject
9153
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009154 Example:
9155 # Track the last IP from X-Forwarded-For
9156 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02009157 tcp-request content track-sc0 hdr(x-forwarded-for,-1)
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009158
9159 Example:
9160 # track request counts per "base" (concatenation of Host+URL)
9161 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02009162 tcp-request content track-sc0 base table req-rate
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009163
Willy Tarreaue9656522010-08-17 15:40:09 +02009164 Example: track per-frontend and per-backend counters, block abusers at the
9165 frontend when the backend detects abuse.
9166
9167 frontend http
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009168 # Use General Purpose Couter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +02009169 # protecting all our sites
9170 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009171 tcp-request connection track-sc0 src
9172 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +02009173 ...
9174 use_backend http_dynamic if { path_end .php }
9175
9176 backend http_dynamic
9177 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009178 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +02009179 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009180 acl click_too_fast sc1_http_req_rate gt 10
9181 acl mark_as_abuser sc0_inc_gpc0 gt 0
9182 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +02009183 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +02009184
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009185 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +02009186
Willy Tarreau4f614292016-10-21 17:49:36 +02009187 See also : "tcp-request connection", "tcp-request session", and
9188 "tcp-request inspect-delay"
Willy Tarreau62644772008-07-16 18:36:06 +02009189
9190
9191tcp-request inspect-delay <timeout>
9192 Set the maximum allowed time to wait for data during content inspection
9193 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02009194 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +02009195 Arguments :
9196 <timeout> is the timeout value specified in milliseconds by default, but
9197 can be in any other unit if the number is suffixed by the unit,
9198 as explained at the top of this document.
9199
9200 People using haproxy primarily as a TCP relay are often worried about the
9201 risk of passing any type of protocol to a server without any analysis. In
9202 order to be able to analyze the request contents, we must first withhold
9203 the data then analyze them. This statement simply enables withholding of
9204 data for at most the specified amount of time.
9205
Willy Tarreaufb356202010-08-03 14:02:05 +02009206 TCP content inspection applies very early when a connection reaches a
9207 frontend, then very early when the connection is forwarded to a backend. This
9208 means that a connection may experience a first delay in the frontend and a
9209 second delay in the backend if both have tcp-request rules.
9210
Willy Tarreau62644772008-07-16 18:36:06 +02009211 Note that when performing content inspection, haproxy will evaluate the whole
9212 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009213 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +02009214 a last check is performed upon expiration, this time considering that the
Willy Tarreaud869b242009-03-15 14:43:58 +01009215 contents are definitive. If no delay is set, haproxy will not wait at all
9216 and will immediately apply a verdict based on the available information.
9217 Obviously this is unlikely to be very useful and might even be racy, so such
9218 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +02009219
9220 As soon as a rule matches, the request is released and continues as usual. If
9221 the timeout is reached and no rule matches, the default policy will be to let
9222 it pass through unaffected.
9223
9224 For most protocols, it is enough to set it to a few seconds, as most clients
9225 send the full request immediately upon connection. Add 3 or more seconds to
9226 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +01009227 to use large values, for instance to ensure that the client never talks
Willy Tarreau62644772008-07-16 18:36:06 +02009228 before the server (eg: SMTP), or to wait for a client to talk before passing
9229 data to the server (eg: SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +02009230 least the inspection delay, otherwise it will expire first. If the client
9231 closes the connection or if the buffer is full, the delay immediately expires
9232 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +02009233
Willy Tarreau55165fe2009-05-10 12:02:55 +02009234 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +02009235 "timeout client".
9236
9237
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009238tcp-response content <action> [{if | unless} <condition>]
9239 Perform an action on a session response depending on a layer 4-7 condition
9240 May be used in sections : defaults | frontend | listen | backend
9241 no | no | yes | yes
9242 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02009243 <action> defines the action to perform if the condition applies. See
9244 below.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009245
9246 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
9247
9248 Response contents can be analysed at an early stage of response processing
9249 called "TCP content inspection". During this stage, ACL-based rules are
9250 evaluated every time the response contents are updated, until either an
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02009251 "accept", "close" or a "reject" rule matches, or a TCP response inspection
9252 delay is set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009253
9254 Most often, these decisions will consider a protocol recognition or validity.
9255
9256 Content-based rules are evaluated in their exact declaration order. If no
9257 rule matches or if there is no rule, the default action is to accept the
9258 contents. There is no specific limit to the number of rules which may be
9259 inserted.
9260
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009261 Several types of actions are supported :
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009262 - accept :
9263 accepts the response if the condition is true (when used with "if")
9264 or false (when used with "unless"). The first such rule executed ends
9265 the rules evaluation.
9266
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02009267 - close :
9268 immediately closes the connection with the server if the condition is
9269 true (when used with "if"), or false (when used with "unless"). The
9270 first such rule executed ends the rules evaluation. The main purpose of
9271 this action is to force a connection to be finished between a client
9272 and a server after an exchange when the application protocol expects
9273 some long time outs to elapse first. The goal is to eliminate idle
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009274 connections which take significant resources on servers with certain
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02009275 protocols.
9276
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009277 - reject :
9278 rejects the response if the condition is true (when used with "if")
9279 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009280 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009281
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009282 - set-var(<var-name>) <expr>
9283 Sets a variable.
9284
Christopher Faulet85d79c92016-11-09 16:54:56 +01009285 - unset-var(<var-name>)
9286 Unsets a variable.
9287
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02009288 - sc-inc-gpc0(<sc-id>):
9289 This action increments the GPC0 counter according to the sticky
9290 counter designated by <sc-id>. If an error occurs, this action fails
9291 silently and the actions evaluation continues.
9292
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009293 - sc-set-gpt0(<sc-id>) <int> :
9294 This action sets the GPT0 tag according to the sticky counter designated
9295 by <sc-id> and the value of <int>. The expected result is a boolean. If
9296 an error occurs, this action silently fails and the actions evaluation
9297 continues.
9298
Willy Tarreau2d392c22015-08-24 01:43:45 +02009299 - "silent-drop" :
9300 This stops the evaluation of the rules and makes the client-facing
9301 connection suddenly disappear using a system-dependant way that tries
9302 to prevent the client from being notified. The effect it then that the
9303 client still sees an established connection while there's none on
9304 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
9305 except that it doesn't use any local resource at all on the machine
9306 running HAProxy. It can resist much higher loads than "tarpit", and
9307 slow down stronger attackers. It is important to undestand the impact
9308 of using this mechanism. All stateful equipments placed between the
9309 client and HAProxy (firewalls, proxies, load balancers) will also keep
9310 the established connection for a long time and may suffer from this
9311 action. On modern Linux systems running with enough privileges, the
9312 TCP_REPAIR socket option is used to block the emission of a TCP
9313 reset. On other systems, the socket's TTL is reduced to 1 so that the
9314 TCP reset doesn't pass the first router, though it's still delivered to
9315 local networks. Do not use it unless you fully understand how it works.
9316
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009317 Note that the "if/unless" condition is optional. If no condition is set on
9318 the action, it is simply performed unconditionally. That can be useful for
9319 for changing the default action to a reject.
9320
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009321 It is perfectly possible to match layer 7 contents with "tcp-response
9322 content" rules, but then it is important to ensure that a full response has
9323 been buffered, otherwise no contents will match. In order to achieve this,
9324 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009325 period.
9326
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009327 The "set-var" is used to set the content of a variable. The variable is
9328 declared inline.
9329
Daniel Schneller0b547052016-03-21 20:46:57 +01009330 <var-name> The name of the variable starts with an indication about
9331 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +01009332 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +01009333 "sess" : the variable is shared with the whole session
9334 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009335 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +01009336 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009337 processing
Daniel Schneller0b547052016-03-21 20:46:57 +01009338 "res" : the variable is shared only during response
9339 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009340 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +01009341 The name may only contain characters 'a-z', 'A-Z', '0-9',
9342 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009343
9344 <expr> Is a standard HAProxy expression formed by a sample-fetch
9345 followed by some converters.
9346
9347 Example:
9348
9349 tcp-request content set-var(sess.my_var) src
9350
Christopher Faulet85d79c92016-11-09 16:54:56 +01009351 The "unset-var" is used to unset a variable. See above for details about
9352 <var-name>.
9353
9354 Example:
9355
9356 tcp-request content unset-var(sess.my_var)
9357
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009358 See section 7 about ACL usage.
9359
9360 See also : "tcp-request content", "tcp-response inspect-delay"
9361
9362
Willy Tarreau4f614292016-10-21 17:49:36 +02009363tcp-request session <action> [{if | unless} <condition>]
9364 Perform an action on a validated session depending on a layer 5 condition
9365 May be used in sections : defaults | frontend | listen | backend
9366 no | yes | yes | no
9367 Arguments :
9368 <action> defines the action to perform if the condition applies. See
9369 below.
9370
9371 <condition> is a standard layer5-only ACL-based condition (see section 7).
9372
9373 Once a session is validated, (ie. after all handshakes have been completed),
9374 it is possible to evaluate some conditions to decide whether this session
9375 must be accepted or dropped or have its counters tracked. Those conditions
9376 cannot make use of any data contents because no buffers are allocated yet and
9377 the processing cannot wait at this stage. The main use case it to copy some
9378 early information into variables (since variables are accessible in the
9379 session), or to keep track of some information collected after the handshake,
9380 such as SSL-level elements (SNI, ciphers, client cert's CN) or information
9381 from the PROXY protocol header (eg: track a source forwarded this way). The
9382 extracted information can thus be copied to a variable or tracked using
9383 "track-sc" rules. Of course it is also possible to decide to accept/reject as
9384 with other rulesets. Most operations performed here could also be performed
9385 in "tcp-request content" rules, except that in HTTP these rules are evaluated
9386 for each new request, and that might not always be acceptable. For example a
9387 rule might increment a counter on each evaluation. It would also be possible
9388 that a country is resolved by geolocation from the source IP address,
9389 assigned to a session-wide variable, then the source address rewritten from
9390 an HTTP header for all requests. If some contents need to be inspected in
9391 order to take the decision, the "tcp-request content" statements must be used
9392 instead.
9393
9394 The "tcp-request session" rules are evaluated in their exact declaration
9395 order. If no rule matches or if there is no rule, the default action is to
9396 accept the incoming session. There is no specific limit to the number of
9397 rules which may be inserted.
9398
9399 Several types of actions are supported :
9400 - accept : the request is accepted
9401 - reject : the request is rejected and the connection is closed
9402 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
9403 - sc-inc-gpc0(<sc-id>)
9404 - sc-set-gpt0(<sc-id>) <int>
9405 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +01009406 - unset-var(<var-name>)
Willy Tarreau4f614292016-10-21 17:49:36 +02009407 - silent-drop
9408
9409 These actions have the same meaning as their respective counter-parts in
9410 "tcp-request connection" and "tcp-request content", so please refer to these
9411 sections for a complete description.
9412
9413 Note that the "if/unless" condition is optional. If no condition is set on
9414 the action, it is simply performed unconditionally. That can be useful for
9415 "track-sc*" actions as well as for changing the default action to a reject.
9416
9417 Example: track the original source address by default, or the one advertised
9418 in the PROXY protocol header for connection coming from the local
9419 proxies. The first connection-level rule enables receipt of the
9420 PROXY protocol for these ones, the second rule tracks whatever
9421 address we decide to keep after optional decoding.
9422
9423 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
9424 tcp-request session track-sc0 src
9425
9426 Example: accept all sessions from white-listed hosts, reject too fast
9427 sessions without counting them, and track accepted sessions.
9428 This results in session rate being capped from abusive sources.
9429
9430 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
9431 tcp-request session reject if { src_sess_rate gt 10 }
9432 tcp-request session track-sc0 src
9433
9434 Example: accept all sessions from white-listed hosts, count all other
9435 sessions and reject too fast ones. This results in abusive ones
9436 being blocked as long as they don't slow down.
9437
9438 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
9439 tcp-request session track-sc0 src
9440 tcp-request session reject if { sc0_sess_rate gt 10 }
9441
9442 See section 7 about ACL usage.
9443
9444 See also : "tcp-request connection", "tcp-request content", "stick-table"
9445
9446
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009447tcp-response inspect-delay <timeout>
9448 Set the maximum allowed time to wait for a response during content inspection
9449 May be used in sections : defaults | frontend | listen | backend
9450 no | no | yes | yes
9451 Arguments :
9452 <timeout> is the timeout value specified in milliseconds by default, but
9453 can be in any other unit if the number is suffixed by the unit,
9454 as explained at the top of this document.
9455
9456 See also : "tcp-response content", "tcp-request inspect-delay".
9457
9458
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009459timeout check <timeout>
9460 Set additional check timeout, but only after a connection has been already
9461 established.
9462
9463 May be used in sections: defaults | frontend | listen | backend
9464 yes | no | yes | yes
9465 Arguments:
9466 <timeout> is the timeout value specified in milliseconds by default, but
9467 can be in any other unit if the number is suffixed by the unit,
9468 as explained at the top of this document.
9469
9470 If set, haproxy uses min("timeout connect", "inter") as a connect timeout
9471 for check and "timeout check" as an additional read timeout. The "min" is
9472 used so that people running with *very* long "timeout connect" (eg. those
9473 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +01009474 (Please also note that there is no valid reason to have such long connect
9475 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
9476 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009477
9478 If "timeout check" is not set haproxy uses "inter" for complete check
9479 timeout (connect + read) exactly like all <1.3.15 version.
9480
9481 In most cases check request is much simpler and faster to handle than normal
9482 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +01009483 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009484
9485 This parameter is specific to backends, but can be specified once for all in
9486 "defaults" sections. This is in fact one of the easiest solutions not to
9487 forget about it.
9488
Willy Tarreau41a340d2008-01-22 12:25:31 +01009489 See also: "timeout connect", "timeout queue", "timeout server",
9490 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009491
9492
Willy Tarreau0ba27502007-12-24 16:55:16 +01009493timeout client <timeout>
9494timeout clitimeout <timeout> (deprecated)
9495 Set the maximum inactivity time on the client side.
9496 May be used in sections : defaults | frontend | listen | backend
9497 yes | yes | yes | no
9498 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01009499 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01009500 can be in any other unit if the number is suffixed by the unit,
9501 as explained at the top of this document.
9502
9503 The inactivity timeout applies when the client is expected to acknowledge or
9504 send data. In HTTP mode, this timeout is particularly important to consider
9505 during the first phase, when the client sends the request, and during the
Baptiste Assmann2e1941e2016-03-06 23:24:12 +01009506 response while it is reading data sent by the server. That said, for the
9507 first phase, it is preferable to set the "timeout http-request" to better
9508 protect HAProxy from Slowloris like attacks. The value is specified in
9509 milliseconds by default, but can be in any other unit if the number is
Willy Tarreau0ba27502007-12-24 16:55:16 +01009510 suffixed by the unit, as specified at the top of this document. In TCP mode
9511 (and to a lesser extent, in HTTP mode), it is highly recommended that the
9512 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01009513 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01009514 losses by specifying timeouts that are slightly above multiples of 3 seconds
Willy Tarreauce887fd2012-05-12 12:50:00 +02009515 (eg: 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
9516 sessions (eg: WebSocket and HTTP), it's worth considering "timeout tunnel",
Willy Tarreau05cdd962014-05-10 14:30:07 +02009517 which overrides "timeout client" and "timeout server" for tunnels, as well as
9518 "timeout client-fin" for half-closed connections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01009519
9520 This parameter is specific to frontends, but can be specified once for all in
9521 "defaults" sections. This is in fact one of the easiest solutions not to
9522 forget about it. An unspecified timeout results in an infinite timeout, which
9523 is not recommended. Such a usage is accepted and works but reports a warning
9524 during startup because it may results in accumulation of expired sessions in
9525 the system if the system's timeouts are not configured either.
9526
9527 This parameter replaces the old, deprecated "clitimeout". It is recommended
9528 to use it to write new configurations. The form "timeout clitimeout" is
9529 provided only by backwards compatibility but its use is strongly discouraged.
9530
Baptiste Assmann2e1941e2016-03-06 23:24:12 +01009531 See also : "clitimeout", "timeout server", "timeout tunnel",
9532 "timeout http-request".
Willy Tarreau0ba27502007-12-24 16:55:16 +01009533
9534
Willy Tarreau05cdd962014-05-10 14:30:07 +02009535timeout client-fin <timeout>
9536 Set the inactivity timeout on the client side for half-closed connections.
9537 May be used in sections : defaults | frontend | listen | backend
9538 yes | yes | yes | no
9539 Arguments :
9540 <timeout> is the timeout value specified in milliseconds by default, but
9541 can be in any other unit if the number is suffixed by the unit,
9542 as explained at the top of this document.
9543
9544 The inactivity timeout applies when the client is expected to acknowledge or
9545 send data while one direction is already shut down. This timeout is different
9546 from "timeout client" in that it only applies to connections which are closed
9547 in one direction. This is particularly useful to avoid keeping connections in
9548 FIN_WAIT state for too long when clients do not disconnect cleanly. This
9549 problem is particularly common long connections such as RDP or WebSocket.
9550 Note that this timeout can override "timeout tunnel" when a connection shuts
9551 down in one direction.
9552
9553 This parameter is specific to frontends, but can be specified once for all in
9554 "defaults" sections. By default it is not set, so half-closed connections
9555 will use the other timeouts (timeout.client or timeout.tunnel).
9556
9557 See also : "timeout client", "timeout server-fin", and "timeout tunnel".
9558
9559
Willy Tarreau0ba27502007-12-24 16:55:16 +01009560timeout connect <timeout>
9561timeout contimeout <timeout> (deprecated)
9562 Set the maximum time to wait for a connection attempt to a server to succeed.
9563 May be used in sections : defaults | frontend | listen | backend
9564 yes | no | yes | yes
9565 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01009566 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01009567 can be in any other unit if the number is suffixed by the unit,
9568 as explained at the top of this document.
9569
9570 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01009571 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01009572 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01009573 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009574 connect timeout also presets both queue and tarpit timeouts to the same value
9575 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +01009576
9577 This parameter is specific to backends, but can be specified once for all in
9578 "defaults" sections. This is in fact one of the easiest solutions not to
9579 forget about it. An unspecified timeout results in an infinite timeout, which
9580 is not recommended. Such a usage is accepted and works but reports a warning
9581 during startup because it may results in accumulation of failed sessions in
9582 the system if the system's timeouts are not configured either.
9583
9584 This parameter replaces the old, deprecated "contimeout". It is recommended
9585 to use it to write new configurations. The form "timeout contimeout" is
9586 provided only by backwards compatibility but its use is strongly discouraged.
9587
Willy Tarreau41a340d2008-01-22 12:25:31 +01009588 See also: "timeout check", "timeout queue", "timeout server", "contimeout",
9589 "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +01009590
9591
Willy Tarreaub16a5742010-01-10 14:46:16 +01009592timeout http-keep-alive <timeout>
9593 Set the maximum allowed time to wait for a new HTTP request to appear
9594 May be used in sections : defaults | frontend | listen | backend
9595 yes | yes | yes | yes
9596 Arguments :
9597 <timeout> is the timeout value specified in milliseconds by default, but
9598 can be in any other unit if the number is suffixed by the unit,
9599 as explained at the top of this document.
9600
9601 By default, the time to wait for a new request in case of keep-alive is set
9602 by "timeout http-request". However this is not always convenient because some
9603 people want very short keep-alive timeouts in order to release connections
9604 faster, and others prefer to have larger ones but still have short timeouts
9605 once the request has started to present itself.
9606
9607 The "http-keep-alive" timeout covers these needs. It will define how long to
9608 wait for a new HTTP request to start coming after a response was sent. Once
9609 the first byte of request has been seen, the "http-request" timeout is used
9610 to wait for the complete request to come. Note that empty lines prior to a
9611 new request do not refresh the timeout and are not counted as a new request.
9612
9613 There is also another difference between the two timeouts : when a connection
9614 expires during timeout http-keep-alive, no error is returned, the connection
9615 just closes. If the connection expires in "http-request" while waiting for a
9616 connection to complete, a HTTP 408 error is returned.
9617
9618 In general it is optimal to set this value to a few tens to hundreds of
9619 milliseconds, to allow users to fetch all objects of a page at once but
9620 without waiting for further clicks. Also, if set to a very small value (eg:
9621 1 millisecond) it will probably only accept pipelined requests but not the
9622 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +02009623 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +01009624
9625 If this parameter is not set, the "http-request" timeout applies, and if both
9626 are not set, "timeout client" still applies at the lower level. It should be
9627 set in the frontend to take effect, unless the frontend is in TCP mode, in
9628 which case the HTTP backend's timeout will be used.
9629
9630 See also : "timeout http-request", "timeout client".
9631
9632
Willy Tarreau036fae02008-01-06 13:24:40 +01009633timeout http-request <timeout>
9634 Set the maximum allowed time to wait for a complete HTTP request
9635 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +02009636 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +01009637 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01009638 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +01009639 can be in any other unit if the number is suffixed by the unit,
9640 as explained at the top of this document.
9641
9642 In order to offer DoS protection, it may be required to lower the maximum
9643 accepted time to receive a complete HTTP request without affecting the client
9644 timeout. This helps protecting against established connections on which
9645 nothing is sent. The client timeout cannot offer a good protection against
9646 this abuse because it is an inactivity timeout, which means that if the
9647 attacker sends one character every now and then, the timeout will not
9648 trigger. With the HTTP request timeout, no matter what speed the client
Willy Tarreau2705a612014-05-23 17:38:34 +02009649 types, the request will be aborted if it does not complete in time. When the
9650 timeout expires, an HTTP 408 response is sent to the client to inform it
9651 about the problem, and the connection is closed. The logs will report
9652 termination codes "cR". Some recent browsers are having problems with this
9653 standard, well-documented behaviour, so it might be needed to hide the 408
Willy Tarreau0f228a02015-05-01 15:37:53 +02009654 code using "option http-ignore-probes" or "errorfile 408 /dev/null". See
9655 more details in the explanations of the "cR" termination code in section 8.5.
Willy Tarreau036fae02008-01-06 13:24:40 +01009656
Baptiste Assmanneccdf432015-10-28 13:49:01 +01009657 By default, this timeout only applies to the header part of the request,
9658 and not to any data. As soon as the empty line is received, this timeout is
9659 not used anymore. When combined with "option http-buffer-request", this
9660 timeout also applies to the body of the request..
9661 It is used again on keep-alive connections to wait for a second
Willy Tarreaub16a5742010-01-10 14:46:16 +01009662 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +01009663
9664 Generally it is enough to set it to a few seconds, as most clients send the
9665 full request immediately upon connection. Add 3 or more seconds to cover TCP
9666 retransmits but that's all. Setting it to very low values (eg: 50 ms) will
9667 generally work on local networks as long as there are no packet losses. This
9668 will prevent people from sending bare HTTP requests using telnet.
9669
9670 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +02009671 chunk of the incoming request. It should be set in the frontend to take
9672 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
9673 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +01009674
Willy Tarreau0f228a02015-05-01 15:37:53 +02009675 See also : "errorfile", "http-ignore-probes", "timeout http-keep-alive", and
Baptiste Assmanneccdf432015-10-28 13:49:01 +01009676 "timeout client", "option http-buffer-request".
Willy Tarreau036fae02008-01-06 13:24:40 +01009677
Willy Tarreau844e3c52008-01-11 16:28:18 +01009678
9679timeout queue <timeout>
9680 Set the maximum time to wait in the queue for a connection slot to be free
9681 May be used in sections : defaults | frontend | listen | backend
9682 yes | no | yes | yes
9683 Arguments :
9684 <timeout> is the timeout value specified in milliseconds by default, but
9685 can be in any other unit if the number is suffixed by the unit,
9686 as explained at the top of this document.
9687
9688 When a server's maxconn is reached, connections are left pending in a queue
9689 which may be server-specific or global to the backend. In order not to wait
9690 indefinitely, a timeout is applied to requests pending in the queue. If the
9691 timeout is reached, it is considered that the request will almost never be
9692 served, so it is dropped and a 503 error is returned to the client.
9693
9694 The "timeout queue" statement allows to fix the maximum time for a request to
9695 be left pending in a queue. If unspecified, the same value as the backend's
9696 connection timeout ("timeout connect") is used, for backwards compatibility
9697 with older versions with no "timeout queue" parameter.
9698
9699 See also : "timeout connect", "contimeout".
9700
9701
9702timeout server <timeout>
9703timeout srvtimeout <timeout> (deprecated)
9704 Set the maximum inactivity time on the server side.
9705 May be used in sections : defaults | frontend | listen | backend
9706 yes | no | yes | yes
9707 Arguments :
9708 <timeout> is the timeout value specified in milliseconds by default, but
9709 can be in any other unit if the number is suffixed by the unit,
9710 as explained at the top of this document.
9711
9712 The inactivity timeout applies when the server is expected to acknowledge or
9713 send data. In HTTP mode, this timeout is particularly important to consider
9714 during the first phase of the server's response, when it has to send the
9715 headers, as it directly represents the server's processing time for the
9716 request. To find out what value to put there, it's often good to start with
9717 what would be considered as unacceptable response times, then check the logs
9718 to observe the response time distribution, and adjust the value accordingly.
9719
9720 The value is specified in milliseconds by default, but can be in any other
9721 unit if the number is suffixed by the unit, as specified at the top of this
9722 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
9723 recommended that the client timeout remains equal to the server timeout in
9724 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01009725 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01009726 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreauce887fd2012-05-12 12:50:00 +02009727 seconds (eg: 4 or 5 seconds minimum). If some long-lived sessions are mixed
9728 with short-lived sessions (eg: WebSocket and HTTP), it's worth considering
9729 "timeout tunnel", which overrides "timeout client" and "timeout server" for
9730 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009731
9732 This parameter is specific to backends, but can be specified once for all in
9733 "defaults" sections. This is in fact one of the easiest solutions not to
9734 forget about it. An unspecified timeout results in an infinite timeout, which
9735 is not recommended. Such a usage is accepted and works but reports a warning
9736 during startup because it may results in accumulation of expired sessions in
9737 the system if the system's timeouts are not configured either.
9738
9739 This parameter replaces the old, deprecated "srvtimeout". It is recommended
9740 to use it to write new configurations. The form "timeout srvtimeout" is
9741 provided only by backwards compatibility but its use is strongly discouraged.
9742
Willy Tarreauce887fd2012-05-12 12:50:00 +02009743 See also : "srvtimeout", "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +01009744
Willy Tarreau05cdd962014-05-10 14:30:07 +02009745
9746timeout server-fin <timeout>
9747 Set the inactivity timeout on the server side for half-closed connections.
9748 May be used in sections : defaults | frontend | listen | backend
9749 yes | no | yes | yes
9750 Arguments :
9751 <timeout> is the timeout value specified in milliseconds by default, but
9752 can be in any other unit if the number is suffixed by the unit,
9753 as explained at the top of this document.
9754
9755 The inactivity timeout applies when the server is expected to acknowledge or
9756 send data while one direction is already shut down. This timeout is different
9757 from "timeout server" in that it only applies to connections which are closed
9758 in one direction. This is particularly useful to avoid keeping connections in
9759 FIN_WAIT state for too long when a remote server does not disconnect cleanly.
9760 This problem is particularly common long connections such as RDP or WebSocket.
9761 Note that this timeout can override "timeout tunnel" when a connection shuts
9762 down in one direction. This setting was provided for completeness, but in most
9763 situations, it should not be needed.
9764
9765 This parameter is specific to backends, but can be specified once for all in
9766 "defaults" sections. By default it is not set, so half-closed connections
9767 will use the other timeouts (timeout.server or timeout.tunnel).
9768
9769 See also : "timeout client-fin", "timeout server", and "timeout tunnel".
9770
Willy Tarreau844e3c52008-01-11 16:28:18 +01009771
9772timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +01009773 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +01009774 May be used in sections : defaults | frontend | listen | backend
9775 yes | yes | yes | yes
9776 Arguments :
9777 <timeout> is the tarpit duration specified in milliseconds by default, but
9778 can be in any other unit if the number is suffixed by the unit,
9779 as explained at the top of this document.
9780
9781 When a connection is tarpitted using "reqtarpit", it is maintained open with
9782 no activity for a certain amount of time, then closed. "timeout tarpit"
9783 defines how long it will be maintained open.
9784
9785 The value is specified in milliseconds by default, but can be in any other
9786 unit if the number is suffixed by the unit, as specified at the top of this
9787 document. If unspecified, the same value as the backend's connection timeout
9788 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +01009789 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009790
9791 See also : "timeout connect", "contimeout".
9792
9793
Willy Tarreauce887fd2012-05-12 12:50:00 +02009794timeout tunnel <timeout>
9795 Set the maximum inactivity time on the client and server side for tunnels.
9796 May be used in sections : defaults | frontend | listen | backend
9797 yes | no | yes | yes
9798 Arguments :
9799 <timeout> is the timeout value specified in milliseconds by default, but
9800 can be in any other unit if the number is suffixed by the unit,
9801 as explained at the top of this document.
9802
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009803 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +02009804 between a client and a server, and the connection remains inactive in both
9805 directions. This timeout supersedes both the client and server timeouts once
9806 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
9807 analyser remains attached to either connection (eg: tcp content rules are
9808 accepted). In HTTP, this timeout is used when a connection is upgraded (eg:
9809 when switching to the WebSocket protocol, or forwarding a CONNECT request
9810 to a proxy), or after the first response when no keepalive/close option is
9811 specified.
9812
Willy Tarreau05cdd962014-05-10 14:30:07 +02009813 Since this timeout is usually used in conjunction with long-lived connections,
9814 it usually is a good idea to also set "timeout client-fin" to handle the
9815 situation where a client suddenly disappears from the net and does not
9816 acknowledge a close, or sends a shutdown and does not acknowledge pending
9817 data anymore. This can happen in lossy networks where firewalls are present,
9818 and is detected by the presence of large amounts of sessions in a FIN_WAIT
9819 state.
9820
Willy Tarreauce887fd2012-05-12 12:50:00 +02009821 The value is specified in milliseconds by default, but can be in any other
9822 unit if the number is suffixed by the unit, as specified at the top of this
9823 document. Whatever the expected normal idle time, it is a good practice to
9824 cover at least one or several TCP packet losses by specifying timeouts that
9825 are slightly above multiples of 3 seconds (eg: 4 or 5 seconds minimum).
9826
9827 This parameter is specific to backends, but can be specified once for all in
9828 "defaults" sections. This is in fact one of the easiest solutions not to
9829 forget about it.
9830
9831 Example :
9832 defaults http
9833 option http-server-close
9834 timeout connect 5s
9835 timeout client 30s
Willy Tarreau05cdd962014-05-10 14:30:07 +02009836 timeout client-fin 30s
Willy Tarreauce887fd2012-05-12 12:50:00 +02009837 timeout server 30s
9838 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
9839
Willy Tarreau05cdd962014-05-10 14:30:07 +02009840 See also : "timeout client", "timeout client-fin", "timeout server".
Willy Tarreauce887fd2012-05-12 12:50:00 +02009841
9842
Willy Tarreau844e3c52008-01-11 16:28:18 +01009843transparent (deprecated)
9844 Enable client-side transparent proxying
9845 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01009846 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01009847 Arguments : none
9848
9849 This keyword was introduced in order to provide layer 7 persistence to layer
9850 3 load balancers. The idea is to use the OS's ability to redirect an incoming
9851 connection for a remote address to a local process (here HAProxy), and let
9852 this process know what address was initially requested. When this option is
9853 used, sessions without cookies will be forwarded to the original destination
9854 IP address of the incoming request (which should match that of another
9855 equipment), while requests with cookies will still be forwarded to the
9856 appropriate server.
9857
9858 The "transparent" keyword is deprecated, use "option transparent" instead.
9859
9860 Note that contrary to a common belief, this option does NOT make HAProxy
9861 present the client's IP to the server when establishing the connection.
9862
Willy Tarreau844e3c52008-01-11 16:28:18 +01009863 See also: "option transparent"
9864
William Lallemanda73203e2012-03-12 12:48:57 +01009865unique-id-format <string>
9866 Generate a unique ID for each request.
9867 May be used in sections : defaults | frontend | listen | backend
9868 yes | yes | yes | no
9869 Arguments :
9870 <string> is a log-format string.
9871
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009872 This keyword creates a ID for each request using the custom log format. A
9873 unique ID is useful to trace a request passing through many components of
9874 a complex infrastructure. The newly created ID may also be logged using the
9875 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +01009876
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009877 The format should be composed from elements that are guaranteed to be
9878 unique when combined together. For instance, if multiple haproxy instances
9879 are involved, it might be important to include the node name. It is often
9880 needed to log the incoming connection's source and destination addresses
9881 and ports. Note that since multiple requests may be performed over the same
9882 connection, including a request counter may help differentiate them.
9883 Similarly, a timestamp may protect against a rollover of the counter.
9884 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +01009885
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009886 It is recommended to use hexadecimal notation for many fields since it
9887 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +01009888
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009889 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01009890
Julien Vehentf21be322014-03-07 08:27:34 -05009891 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +01009892
9893 will generate:
9894
9895 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
9896
9897 See also: "unique-id-header"
9898
9899unique-id-header <name>
9900 Add a unique ID header in the HTTP request.
9901 May be used in sections : defaults | frontend | listen | backend
9902 yes | yes | yes | no
9903 Arguments :
9904 <name> is the name of the header.
9905
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009906 Add a unique-id header in the HTTP request sent to the server, using the
9907 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +01009908
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009909 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01009910
Julien Vehentf21be322014-03-07 08:27:34 -05009911 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +01009912 unique-id-header X-Unique-ID
9913
9914 will generate:
9915
9916 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
9917
9918 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +01009919
Willy Tarreauf51658d2014-04-23 01:21:56 +02009920use_backend <backend> [{if | unless} <condition>]
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02009921 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009922 May be used in sections : defaults | frontend | listen | backend
9923 no | yes | yes | no
9924 Arguments :
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01009925 <backend> is the name of a valid backend or "listen" section, or a
9926 "log-format" string resolving to a backend name.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009927
Willy Tarreauf51658d2014-04-23 01:21:56 +02009928 <condition> is a condition composed of ACLs, as described in section 7. If
9929 it is omitted, the rule is unconditionally applied.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009930
9931 When doing content-switching, connections arrive on a frontend and are then
9932 dispatched to various backends depending on a number of conditions. The
9933 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02009934 "use_backend" keyword. While it is normally used with HTTP processing, it can
9935 also be used in pure TCP, either without content using stateless ACLs (eg:
9936 source address validation) or combined with a "tcp-request" rule to wait for
9937 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009938
9939 There may be as many "use_backend" rules as desired. All of these rules are
9940 evaluated in their declaration order, and the first one which matches will
9941 assign the backend.
9942
9943 In the first form, the backend will be used if the condition is met. In the
9944 second form, the backend will be used if the condition is not met. If no
9945 condition is valid, the backend defined with "default_backend" will be used.
9946 If no default backend is defined, either the servers in the same section are
9947 used (in case of a "listen" section) or, in case of a frontend, no server is
9948 used and a 503 service unavailable response is returned.
9949
Willy Tarreau51aecc72009-07-12 09:47:04 +02009950 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009951 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +02009952 and backend processing will immediately follow, or the backend will wait for
9953 a complete HTTP request to get in. This feature is useful when a frontend
9954 must decode several protocols on a unique port, one of them being HTTP.
9955
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01009956 When <backend> is a simple name, it is resolved at configuration time, and an
9957 error is reported if the specified backend does not exist. If <backend> is
9958 a log-format string instead, no check may be done at configuration time, so
9959 the backend name is resolved dynamically at run time. If the resulting
9960 backend name does not correspond to any valid backend, no other rule is
9961 evaluated, and the default_backend directive is applied instead. Note that
9962 when using dynamic backend names, it is highly recommended to use a prefix
9963 that no other backend uses in order to ensure that an unauthorized backend
9964 cannot be forced from the request.
9965
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009966 It is worth mentioning that "use_backend" rules with an explicit name are
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01009967 used to detect the association between frontends and backends to compute the
9968 backend's "fullconn" setting. This cannot be done for dynamic names.
9969
9970 See also: "default_backend", "tcp-request", "fullconn", "log-format", and
9971 section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +01009972
Willy Tarreau036fae02008-01-06 13:24:40 +01009973
Willy Tarreau4a5cade2012-04-05 21:09:48 +02009974use-server <server> if <condition>
9975use-server <server> unless <condition>
9976 Only use a specific server if/unless an ACL-based condition is matched.
9977 May be used in sections : defaults | frontend | listen | backend
9978 no | no | yes | yes
9979 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009980 <server> is the name of a valid server in the same backend section.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02009981
9982 <condition> is a condition composed of ACLs, as described in section 7.
9983
9984 By default, connections which arrive to a backend are load-balanced across
9985 the available servers according to the configured algorithm, unless a
9986 persistence mechanism such as a cookie is used and found in the request.
9987
9988 Sometimes it is desirable to forward a particular request to a specific
9989 server without having to declare a dedicated backend for this server. This
9990 can be achieved using the "use-server" rules. These rules are evaluated after
9991 the "redirect" rules and before evaluating cookies, and they have precedence
9992 on them. There may be as many "use-server" rules as desired. All of these
9993 rules are evaluated in their declaration order, and the first one which
9994 matches will assign the server.
9995
9996 If a rule designates a server which is down, and "option persist" is not used
9997 and no force-persist rule was validated, it is ignored and evaluation goes on
9998 with the next rules until one matches.
9999
10000 In the first form, the server will be used if the condition is met. In the
10001 second form, the server will be used if the condition is not met. If no
10002 condition is valid, the processing continues and the server will be assigned
10003 according to other persistence mechanisms.
10004
10005 Note that even if a rule is matched, cookie processing is still performed but
10006 does not assign the server. This allows prefixed cookies to have their prefix
10007 stripped.
10008
10009 The "use-server" statement works both in HTTP and TCP mode. This makes it
10010 suitable for use with content-based inspection. For instance, a server could
10011 be selected in a farm according to the TLS SNI field. And if these servers
10012 have their weight set to zero, they will not be used for other traffic.
10013
10014 Example :
10015 # intercept incoming TLS requests based on the SNI field
10016 use-server www if { req_ssl_sni -i www.example.com }
10017 server www 192.168.0.1:443 weight 0
10018 use-server mail if { req_ssl_sni -i mail.example.com }
10019 server mail 192.168.0.1:587 weight 0
10020 use-server imap if { req_ssl_sni -i imap.example.com }
10021 server mail 192.168.0.1:993 weight 0
10022 # all the rest is forwarded to this server
10023 server default 192.168.0.2:443 check
10024
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010025 See also: "use_backend", section 5 about server and section 7 about ACLs.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020010026
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010027
100285. Bind and Server options
10029--------------------------
10030
10031The "bind", "server" and "default-server" keywords support a number of settings
10032depending on some build options and on the system HAProxy was built on. These
10033settings generally each consist in one word sometimes followed by a value,
10034written on the same line as the "bind" or "server" line. All these options are
10035described in this section.
10036
10037
100385.1. Bind options
10039-----------------
10040
10041The "bind" keyword supports a certain number of settings which are all passed
10042as arguments on the same line. The order in which those arguments appear makes
10043no importance, provided that they appear after the bind address. All of these
10044parameters are optional. Some of them consist in a single words (booleans),
10045while other ones expect a value after them. In this case, the value must be
10046provided immediately after the setting name.
10047
10048The currently supported settings are the following ones.
10049
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010010050accept-netscaler-cip <magic number>
10051 Enforces the use of the NetScaler Client IP insertion protocol over any
10052 connection accepted by any of the TCP sockets declared on the same line. The
10053 NetScaler Client IP insertion protocol dictates the layer 3/4 addresses of
10054 the incoming connection to be used everywhere an address is used, with the
10055 only exception of "tcp-request connection" rules which will only see the
10056 real connection address. Logs will reflect the addresses indicated in the
10057 protocol, unless it is violated, in which case the real address will still
10058 be used. This keyword combined with support from external components can be
10059 used as an efficient and reliable alternative to the X-Forwarded-For
Bertrand Jacquin90759682016-06-06 15:35:39 +010010060 mechanism which is not always reliable and not even always usable. See also
10061 "tcp-request connection expect-netscaler-cip" for a finer-grained setting of
10062 which client is allowed to use the protocol.
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010010063
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010064accept-proxy
10065 Enforces the use of the PROXY protocol over any connection accepted by any of
Willy Tarreau77992672014-06-14 11:06:17 +020010066 the sockets declared on the same line. Versions 1 and 2 of the PROXY protocol
10067 are supported and correctly detected. The PROXY protocol dictates the layer
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010068 3/4 addresses of the incoming connection to be used everywhere an address is
10069 used, with the only exception of "tcp-request connection" rules which will
10070 only see the real connection address. Logs will reflect the addresses
10071 indicated in the protocol, unless it is violated, in which case the real
10072 address will still be used. This keyword combined with support from external
10073 components can be used as an efficient and reliable alternative to the
10074 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +020010075 usable. See also "tcp-request connection expect-proxy" for a finer-grained
10076 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010077
Willy Tarreauab861d32013-04-02 02:30:41 +020010078alpn <protocols>
10079 This enables the TLS ALPN extension and advertises the specified protocol
10080 list as supported on top of ALPN. The protocol list consists in a comma-
10081 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
10082 quotes). This requires that the SSL library is build with support for TLS
10083 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
10084 initial NPN extension.
10085
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010086backlog <backlog>
10087 Sets the socket's backlog to this value. If unspecified, the frontend's
10088 backlog is used instead, which generally defaults to the maxconn value.
10089
Emeric Brun7fb34422012-09-28 15:26:15 +020010090ecdhe <named curve>
10091 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +010010092 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
10093 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +020010094
Emeric Brunfd33a262012-10-11 16:28:27 +020010095ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +020010096 This setting is only available when support for OpenSSL was built in. It
10097 designates a PEM file from which to load CA certificates used to verify
10098 client's certificate.
10099
Emeric Brunb6dc9342012-09-28 17:55:37 +020010100ca-ignore-err [all|<errorID>,...]
10101 This setting is only available when support for OpenSSL was built in.
10102 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
10103 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
10104 error is ignored.
10105
Christopher Faulet31af49d2015-06-09 17:29:50 +020010106ca-sign-file <cafile>
10107 This setting is only available when support for OpenSSL was built in. It
10108 designates a PEM file containing both the CA certificate and the CA private
10109 key used to create and sign server's certificates. This is a mandatory
10110 setting when the dynamic generation of certificates is enabled. See
10111 'generate-certificates' for details.
10112
10113ca-sign-passphrase <passphrase>
10114 This setting is only available when support for OpenSSL was built in. It is
10115 the CA private key passphrase. This setting is optional and used only when
10116 the dynamic generation of certificates is enabled. See
10117 'generate-certificates' for details.
10118
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010119ciphers <ciphers>
10120 This setting is only available when support for OpenSSL was built in. It sets
10121 the string describing the list of cipher algorithms ("cipher suite") that are
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010122 negotiated during the SSL/TLS handshake. The format of the string is defined
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010123 in "man 1 ciphers" from OpenSSL man pages, and can be for instance a string
10124 such as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes).
10125
Emeric Brunfd33a262012-10-11 16:28:27 +020010126crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +020010127 This setting is only available when support for OpenSSL was built in. It
10128 designates a PEM file from which to load certificate revocation list used
10129 to verify client's certificate.
10130
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010131crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +000010132 This setting is only available when support for OpenSSL was built in. It
10133 designates a PEM file containing both the required certificates and any
10134 associated private keys. This file can be built by concatenating multiple
10135 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
10136 requires an intermediate certificate, this can also be concatenated into this
10137 file.
10138
10139 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
10140 are loaded.
10141
10142 If a directory name is used instead of a PEM file, then all files found in
Cyril Bonté3180f7b2015-01-25 00:16:08 +010010143 that directory will be loaded in alphabetic order unless their name ends with
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010010144 '.issuer', '.ocsp' or '.sctl' (reserved extensions). This directive may be
10145 specified multiple times in order to load certificates from multiple files or
10146 directories. The certificates will be presented to clients who provide a
10147 valid TLS Server Name Indication field matching one of their CN or alt
10148 subjects. Wildcards are supported, where a wildcard character '*' is used
10149 instead of the first hostname component (eg: *.example.org matches
10150 www.example.org but not www.sub.example.org).
Alex Davies0fbf0162013-03-02 16:04:50 +000010151
10152 If no SNI is provided by the client or if the SSL library does not support
10153 TLS extensions, or if the client provides an SNI hostname which does not
10154 match any certificate, then the first loaded certificate will be presented.
10155 This means that when loading certificates from a directory, it is highly
Cyril Bonté3180f7b2015-01-25 00:16:08 +010010156 recommended to load the default one first as a file or to ensure that it will
10157 always be the first one in the directory.
Alex Davies0fbf0162013-03-02 16:04:50 +000010158
Emeric Brune032bfa2012-09-28 13:01:45 +020010159 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010160
Alex Davies0fbf0162013-03-02 16:04:50 +000010161 Some CAs (such as Godaddy) offer a drop down list of server types that do not
10162 include HAProxy when obtaining a certificate. If this happens be sure to
Godbach8bf60a12014-04-21 21:42:41 +080010163 choose a webserver that the CA believes requires an intermediate CA (for
Alex Davies0fbf0162013-03-02 16:04:50 +000010164 Godaddy, selection Apache Tomcat will get the correct bundle, but many
10165 others, e.g. nginx, result in a wrong bundle that will not work for some
10166 clients).
10167
Emeric Brun4147b2e2014-06-16 18:36:30 +020010168 For each PEM file, haproxy checks for the presence of file at the same path
10169 suffixed by ".ocsp". If such file is found, support for the TLS Certificate
10170 Status Request extension (also known as "OCSP stapling") is automatically
10171 enabled. The content of this file is optional. If not empty, it must contain
10172 a valid OCSP Response in DER format. In order to be valid an OCSP Response
10173 must comply with the following rules: it has to indicate a good status,
10174 it has to be a single response for the certificate of the PEM file, and it
10175 has to be valid at the moment of addition. If these rules are not respected
10176 the OCSP Response is ignored and a warning is emitted. In order to identify
10177 which certificate an OCSP Response applies to, the issuer's certificate is
10178 necessary. If the issuer's certificate is not found in the PEM file, it will
10179 be loaded from a file at the same path as the PEM file suffixed by ".issuer"
10180 if it exists otherwise it will fail with an error.
10181
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010010182 For each PEM file, haproxy also checks for the presence of file at the same
10183 path suffixed by ".sctl". If such file is found, support for Certificate
10184 Transparency (RFC6962) TLS extension is enabled. The file must contain a
10185 valid Signed Certificate Timestamp List, as described in RFC. File is parsed
10186 to check basic syntax, but no signatures are verified.
10187
yanbzhu6c25e9e2016-01-05 12:52:02 -050010188 There are cases where it is desirable to support multiple key types, e.g. RSA
10189 and ECDSA in the cipher suites offered to the clients. This allows clients
10190 that support EC certificates to be able to use EC ciphers, while
10191 simultaneously supporting older, RSA only clients.
yanbzhud19630c2015-12-14 15:10:25 -050010192
10193 In order to provide this functionality, multiple PEM files, each with a
10194 different key type, are required. To associate these PEM files into a
10195 "cert bundle" that is recognized by haproxy, they must be named in the
10196 following way: All PEM files that are to be bundled must have the same base
10197 name, with a suffix indicating the key type. Currently, three suffixes are
10198 supported: rsa, dsa and ecdsa. For example, if www.example.com has two PEM
10199 files, an RSA file and an ECDSA file, they must be named: "example.pem.rsa"
10200 and "example.pem.ecdsa". The first part of the filename is arbitrary; only the
10201 suffix matters. To load this bundle into haproxy, specify the base name only:
10202
10203 Example : bind :8443 ssl crt example.pem
10204
yanbzhu6c25e9e2016-01-05 12:52:02 -050010205 Note that the suffix is not given to haproxy; this tells haproxy to look for
yanbzhud19630c2015-12-14 15:10:25 -050010206 a cert bundle.
10207
10208 Haproxy will load all PEM files in the bundle at the same time to try to
10209 support multiple key types. PEM files are combined based on Common Name
10210 (CN) and Subject Alternative Name (SAN) to support SNI lookups. This means
10211 that even if you give haproxy a cert bundle, if there are no shared CN/SAN
10212 entries in the certificates in that bundle, haproxy will not be able to
10213 provide multi-cert support.
10214
10215 Assuming bundle in the example above contained the following:
10216
10217 Filename | CN | SAN
10218 -------------------+-----------------+-------------------
10219 example.pem.rsa | www.example.com | rsa.example.com
yanbzhu6c25e9e2016-01-05 12:52:02 -050010220 -------------------+-----------------+-------------------
yanbzhud19630c2015-12-14 15:10:25 -050010221 example.pem.ecdsa | www.example.com | ecdsa.example.com
10222 -------------------+-----------------+-------------------
10223
10224 Users connecting with an SNI of "www.example.com" will be able
10225 to use both RSA and ECDSA cipher suites. Users connecting with an SNI of
10226 "rsa.example.com" will only be able to use RSA cipher suites, and users
10227 connecting with "ecdsa.example.com" will only be able to use ECDSA cipher
10228 suites.
10229
10230 If a directory name is given as the <cert> argument, haproxy will
10231 automatically search and load bundled files in that directory.
10232
10233 OSCP files (.ocsp) and issuer files (.issuer) are supported with multi-cert
10234 bundling. Each certificate can have its own .ocsp and .issuer file. At this
10235 time, sctl is not supported in multi-certificate bundling.
10236
Emeric Brunb6dc9342012-09-28 17:55:37 +020010237crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +000010238 This setting is only available when support for OpenSSL was built in. Sets a
10239 comma separated list of errorIDs to ignore during verify at depth == 0. If
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010240 set to 'all', all errors are ignored. SSL handshake is not aborted if an error
Alex Davies0fbf0162013-03-02 16:04:50 +000010241 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +020010242
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010010243crt-list <file>
10244 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +020010245 designates a list of PEM file with an optional list of SNI filter per
10246 certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010010247
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +020010248 <crtfile> [[!]<snifilter> ...]
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010010249
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +020010250 Wildcards are supported in the SNI filter. Negative filter are also supported,
10251 only useful in combination with a wildcard filter to exclude a particular SNI.
10252 The certificates will be presented to clients who provide a valid TLS Server
10253 Name Indication field matching one of the SNI filters. If no SNI filter is
10254 specified, the CN and alt subjects are used. This directive may be specified
10255 multiple times. See the "crt" option for more information. The default
10256 certificate is still needed to meet OpenSSL expectations. If it is not used,
10257 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010010258
yanbzhu6c25e9e2016-01-05 12:52:02 -050010259 Multi-cert bundling (see "crt") is supported with crt-list, as long as only
Emmanuel Hocdetd294aea2016-05-13 11:14:06 +020010260 the base name is given in the crt-list. SNI filter will do the same work on
10261 all bundled certificates.
yanbzhud19630c2015-12-14 15:10:25 -050010262
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010263defer-accept
10264 Is an optional keyword which is supported only on certain Linux kernels. It
10265 states that a connection will only be accepted once some data arrive on it,
10266 or at worst after the first retransmit. This should be used only on protocols
10267 for which the client talks first (eg: HTTP). It can slightly improve
10268 performance by ensuring that most of the request is already available when
10269 the connection is accepted. On the other hand, it will not be able to detect
10270 connections which don't talk. It is important to note that this option is
10271 broken in all kernels up to 2.6.31, as the connection is never accepted until
10272 the client talks. This can cause issues with front firewalls which would see
10273 an established connection while the proxy will only see it in SYN_RECV. This
10274 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
10275
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010276force-sslv3
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010277 This option enforces use of SSLv3 only on SSL connections instantiated from
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010278 this listener. SSLv3 is generally less expensive than the TLS counterparts
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010279 for high connection rates. This option is also available on global statement
10280 "ssl-default-bind-options". See also "no-tlsv*" and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010281
10282force-tlsv10
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010283 This option enforces use of TLSv1.0 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010284 this listener. This option is also available on global statement
10285 "ssl-default-bind-options". See also "no-tlsv*" and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010286
10287force-tlsv11
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010288 This option enforces use of TLSv1.1 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010289 this listener. This option is also available on global statement
10290 "ssl-default-bind-options". See also "no-tlsv*", and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010291
10292force-tlsv12
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010293 This option enforces use of TLSv1.2 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010294 this listener. This option is also available on global statement
10295 "ssl-default-bind-options". See also "no-tlsv*", and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010296
Christopher Faulet31af49d2015-06-09 17:29:50 +020010297generate-certificates
10298 This setting is only available when support for OpenSSL was built in. It
10299 enables the dynamic SSL certificates generation. A CA certificate and its
10300 private key are necessary (see 'ca-sign-file'). When HAProxy is configured as
10301 a transparent forward proxy, SSL requests generate errors because of a common
10302 name mismatch on the certificate presented to the client. With this option
10303 enabled, HAProxy will try to forge a certificate using the SNI hostname
10304 indicated by the client. This is done only if no certificate matches the SNI
10305 hostname (see 'crt-list'). If an error occurs, the default certificate is
10306 used, else the 'strict-sni' option is set.
10307 It can also be used when HAProxy is configured as a reverse proxy to ease the
10308 deployment of an architecture with many backends.
10309
10310 Creating a SSL certificate is an expensive operation, so a LRU cache is used
10311 to store forged certificates (see 'tune.ssl.ssl-ctx-cache-size'). It
10312 increases the HAProxy's memroy footprint to reduce latency when the same
10313 certificate is used many times.
10314
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010315gid <gid>
10316 Sets the group of the UNIX sockets to the designated system gid. It can also
10317 be set by default in the global section's "unix-bind" statement. Note that
10318 some platforms simply ignore this. This setting is equivalent to the "group"
10319 setting except that the group ID is used instead of its name. This setting is
10320 ignored by non UNIX sockets.
10321
10322group <group>
10323 Sets the group of the UNIX sockets to the designated system group. It can
10324 also be set by default in the global section's "unix-bind" statement. Note
10325 that some platforms simply ignore this. This setting is equivalent to the
10326 "gid" setting except that the group name is used instead of its gid. This
10327 setting is ignored by non UNIX sockets.
10328
10329id <id>
10330 Fixes the socket ID. By default, socket IDs are automatically assigned, but
10331 sometimes it is more convenient to fix them to ease monitoring. This value
10332 must be strictly positive and unique within the listener/frontend. This
10333 option can only be used when defining only a single socket.
10334
10335interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +010010336 Restricts the socket to a specific interface. When specified, only packets
10337 received from that particular interface are processed by the socket. This is
10338 currently only supported on Linux. The interface must be a primary system
10339 interface, not an aliased interface. It is also possible to bind multiple
10340 frontends to the same address if they are bound to different interfaces. Note
10341 that binding to a network interface requires root privileges. This parameter
10342 is only compatible with TCPv4/TCPv6 sockets.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010343
Willy Tarreauabb175f2012-09-24 12:43:26 +020010344level <level>
10345 This setting is used with the stats sockets only to restrict the nature of
10346 the commands that can be issued on the socket. It is ignored by other
10347 sockets. <level> can be one of :
10348 - "user" is the least privileged level ; only non-sensitive stats can be
10349 read, and no change is allowed. It would make sense on systems where it
10350 is not easy to restrict access to the socket.
10351 - "operator" is the default level and fits most common uses. All data can
10352 be read, and only non-sensitive changes are permitted (eg: clear max
10353 counters).
10354 - "admin" should be used with care, as everything is permitted (eg: clear
10355 all counters).
10356
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010357maxconn <maxconn>
10358 Limits the sockets to this number of concurrent connections. Extraneous
10359 connections will remain in the system's backlog until a connection is
10360 released. If unspecified, the limit will be the same as the frontend's
10361 maxconn. Note that in case of port ranges or multiple addresses, the same
10362 value will be applied to each socket. This setting enables different
10363 limitations on expensive sockets, for instance SSL entries which may easily
10364 eat all memory.
10365
10366mode <mode>
10367 Sets the octal mode used to define access permissions on the UNIX socket. It
10368 can also be set by default in the global section's "unix-bind" statement.
10369 Note that some platforms simply ignore this. This setting is ignored by non
10370 UNIX sockets.
10371
10372mss <maxseg>
10373 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
10374 connections. This can be used to force a lower MSS for certain specific
10375 ports, for instance for connections passing through a VPN. Note that this
10376 relies on a kernel feature which is theoretically supported under Linux but
10377 was buggy in all versions prior to 2.6.28. It may or may not work on other
10378 operating systems. It may also not change the advertised value but change the
10379 effective size of outgoing segments. The commonly advertised value for TCPv4
10380 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
10381 positive, it will be used as the advertised MSS. If it is negative, it will
10382 indicate by how much to reduce the incoming connection's advertised MSS for
10383 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
10384
10385name <name>
10386 Sets an optional name for these sockets, which will be reported on the stats
10387 page.
10388
Willy Tarreaud72f0f32015-10-13 14:50:22 +020010389namespace <name>
10390 On Linux, it is possible to specify which network namespace a socket will
10391 belong to. This directive makes it possible to explicitly bind a listener to
10392 a namespace different from the default one. Please refer to your operating
10393 system's documentation to find more details about network namespaces.
10394
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010395nice <nice>
10396 Sets the 'niceness' of connections initiated from the socket. Value must be
10397 in the range -1024..1024 inclusive, and defaults to zero. Positive values
10398 means that such connections are more friendly to others and easily offer
10399 their place in the scheduler. On the opposite, negative values mean that
10400 connections want to run with a higher priority than others. The difference
10401 only happens under high loads when the system is close to saturation.
10402 Negative values are appropriate for low-latency or administration services,
10403 and high values are generally recommended for CPU intensive tasks such as SSL
10404 processing or bulk transfers which are less sensible to latency. For example,
10405 it may make sense to use a positive value for an SMTP socket and a negative
10406 one for an RDP socket.
10407
Emeric Brun9b3009b2012-10-05 11:55:06 +020010408no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010409 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010410 disables support for SSLv3 on any sockets instantiated from the listener when
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010411 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010412 be enabled using any configuration option. This option is also available on
10413 global statement "ssl-default-bind-options". See also "force-tls*",
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010414 and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010415
Emeric Brun90ad8722012-10-02 14:00:59 +020010416no-tls-tickets
10417 This setting is only available when support for OpenSSL was built in. It
10418 disables the stateless session resumption (RFC 5077 TLS Ticket
10419 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010420 session resumption is more expensive in CPU usage. This option is also
10421 available on global statement "ssl-default-bind-options".
Emeric Brun90ad8722012-10-02 14:00:59 +020010422
Emeric Brun9b3009b2012-10-05 11:55:06 +020010423no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010424 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010425 disables support for TLSv1.0 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010426 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010427 cannot be enabled using any configuration option. This option is also
10428 available on global statement "ssl-default-bind-options". See also
10429 "force-tlsv*", and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010430
Emeric Brun9b3009b2012-10-05 11:55:06 +020010431no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +020010432 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010433 disables support for TLSv1.1 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010434 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010435 cannot be enabled using any configuration option. This option is also
10436 available on global statement "ssl-default-bind-options". See also
10437 "force-tlsv*", and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +020010438
Emeric Brun9b3009b2012-10-05 11:55:06 +020010439no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +020010440 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010441 disables support for TLSv1.2 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010442 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010443 cannot be enabled using any configuration option. This option is also
10444 available on global statement "ssl-default-bind-options". See also
10445 "force-tlsv*", and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +020010446
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020010447npn <protocols>
10448 This enables the NPN TLS extension and advertises the specified protocol list
10449 as supported on top of NPN. The protocol list consists in a comma-delimited
10450 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
10451 This requires that the SSL library is build with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +020010452 enabled (check with haproxy -vv). Note that the NPN extension has been
10453 replaced with the ALPN extension (see the "alpn" keyword).
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020010454
Willy Tarreau6ae1ba62014-05-07 19:01:58 +020010455process [ all | odd | even | <number 1-64>[-<number 1-64>] ]
10456 This restricts the list of processes on which this listener is allowed to
10457 run. It does not enforce any process but eliminates those which do not match.
10458 If the frontend uses a "bind-process" setting, the intersection between the
10459 two is applied. If in the end the listener is not allowed to run on any
10460 remaining process, a warning is emitted, and the listener will either run on
10461 the first process of the listener if a single process was specified, or on
10462 all of its processes if multiple processes were specified. For the unlikely
Willy Tarreauae302532014-05-07 19:22:24 +020010463 case where several ranges are needed, this directive may be repeated. The
10464 main purpose of this directive is to be used with the stats sockets and have
10465 one different socket per process. The second purpose is to have multiple bind
10466 lines sharing the same IP:port but not the same process in a listener, so
10467 that the system can distribute the incoming connections into multiple queues
10468 and allow a smoother inter-process load balancing. Currently Linux 3.9 and
10469 above is known for supporting this. See also "bind-process" and "nbproc".
Willy Tarreau6ae1ba62014-05-07 19:01:58 +020010470
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010471ssl
10472 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010473 enables SSL deciphering on connections instantiated from this listener. A
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010474 certificate is necessary (see "crt" above). All contents in the buffers will
10475 appear in clear text, so that ACLs and HTTP processing will only have access
10476 to deciphered contents.
10477
Emmanuel Hocdet65623372013-01-24 17:17:15 +010010478strict-sni
10479 This setting is only available when support for OpenSSL was built in. The
10480 SSL/TLS negotiation is allow only if the client provided an SNI which match
10481 a certificate. The default certificate is not used.
10482 See the "crt" option for more information.
10483
Willy Tarreau2af207a2015-02-04 00:45:58 +010010484tcp-ut <delay>
10485 Sets the TCP User Timeout for all incoming connections instanciated from this
10486 listening socket. This option is available on Linux since version 2.6.37. It
10487 allows haproxy to configure a timeout for sockets which contain data not
10488 receiving an acknoledgement for the configured delay. This is especially
10489 useful on long-lived connections experiencing long idle periods such as
10490 remote terminals or database connection pools, where the client and server
10491 timeouts must remain high to allow a long period of idle, but where it is
10492 important to detect that the client has disappeared in order to release all
10493 resources associated with its connection (and the server's session). The
10494 argument is a delay expressed in milliseconds by default. This only works
10495 for regular TCP connections, and is ignored for other protocols.
10496
Willy Tarreau1c862c52012-10-05 16:21:00 +020010497tfo
Lukas Tribus0defb902013-02-13 23:35:39 +010010498 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +020010499 enables TCP Fast Open on the listening socket, which means that clients which
10500 support this feature will be able to send a request and receive a response
10501 during the 3-way handshake starting from second connection, thus saving one
10502 round-trip after the first connection. This only makes sense with protocols
10503 that use high connection rates and where each round trip matters. This can
10504 possibly cause issues with many firewalls which do not accept data on SYN
10505 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +020010506 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
10507 need to build HAProxy with USE_TFO=1 if your libc doesn't define
10508 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +020010509
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010010510tls-ticket-keys <keyfile>
10511 Sets the TLS ticket keys file to load the keys from. The keys need to be 48
10512 bytes long, encoded with base64 (ex. openssl rand -base64 48). Number of keys
10513 is specified by the TLS_TICKETS_NO build option (default 3) and at least as
10514 many keys need to be present in the file. Last TLS_TICKETS_NO keys will be
10515 used for decryption and the penultimate one for encryption. This enables easy
10516 key rotation by just appending new key to the file and reloading the process.
10517 Keys must be periodically rotated (ex. every 12h) or Perfect Forward Secrecy
10518 is compromised. It is also a good idea to keep the keys off any permanent
10519 storage such as hard drives (hint: use tmpfs and don't swap those files).
10520 Lifetime hint can be changed using tune.ssl.timeout.
10521
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010522transparent
10523 Is an optional keyword which is supported only on certain Linux kernels. It
10524 indicates that the addresses will be bound even if they do not belong to the
10525 local machine, and that packets targeting any of these addresses will be
10526 intercepted just as if the addresses were locally configured. This normally
10527 requires that IP forwarding is enabled. Caution! do not use this with the
10528 default address '*', as it would redirect any traffic for the specified port.
10529 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
10530 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
10531 kernel version. Some distribution kernels include backports of the feature,
10532 so check for support with your vendor.
10533
Willy Tarreau77e3af92012-11-24 15:07:23 +010010534v4v6
10535 Is an optional keyword which is supported only on most recent systems
10536 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
10537 and IPv6 when it uses the default address. Doing so is sometimes necessary
10538 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010539 sockets, and is overridden by the "v6only" option.
Willy Tarreau77e3af92012-11-24 15:07:23 +010010540
Willy Tarreau9b6700f2012-11-24 11:55:28 +010010541v6only
10542 Is an optional keyword which is supported only on most recent systems
10543 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
10544 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +010010545 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
10546 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +010010547
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010548uid <uid>
10549 Sets the owner of the UNIX sockets to the designated system uid. It can also
10550 be set by default in the global section's "unix-bind" statement. Note that
10551 some platforms simply ignore this. This setting is equivalent to the "user"
10552 setting except that the user numeric ID is used instead of its name. This
10553 setting is ignored by non UNIX sockets.
10554
10555user <user>
10556 Sets the owner of the UNIX sockets to the designated system user. It can also
10557 be set by default in the global section's "unix-bind" statement. Note that
10558 some platforms simply ignore this. This setting is equivalent to the "uid"
10559 setting except that the user name is used instead of its uid. This setting is
10560 ignored by non UNIX sockets.
10561
Emeric Brun1a073b42012-09-28 17:07:34 +020010562verify [none|optional|required]
10563 This setting is only available when support for OpenSSL was built in. If set
10564 to 'none', client certificate is not requested. This is the default. In other
10565 cases, a client certificate is requested. If the client does not provide a
10566 certificate after the request and if 'verify' is set to 'required', then the
10567 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +020010568 certificate provided by the client is always verified using CAs from
10569 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
10570 is aborted, regardless of the 'verify' option, unless the error code exactly
10571 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020010572
Willy Tarreaub6205fd2012-09-24 12:27:33 +0200105735.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +010010574------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020010575
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010010576The "server" and "default-server" keywords support a certain number of settings
10577which are all passed as arguments on the server line. The order in which those
10578arguments appear does not count, and they are all optional. Some of those
10579settings are single words (booleans) while others expect one or several values
10580after them. In this case, the values must immediately follow the setting name.
10581Except default-server, all those settings must be specified after the server's
10582address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +020010583
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010584 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010010585 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +020010586
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010587The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010588
Willy Tarreauceb4ac92012-04-28 00:41:46 +020010589addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010590 Using the "addr" parameter, it becomes possible to use a different IP address
Baptiste Assmann13f83532016-03-06 23:14:36 +010010591 to send health-checks or to probe the agent-check. On some servers, it may be
10592 desirable to dedicate an IP address to specific component able to perform
10593 complex tests which are more suitable to health-checks than the application.
10594 This parameter is ignored if the "check" parameter is not set. See also the
10595 "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010596
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010597 Supported in default-server: No
10598
Simon Hormand60d6912013-11-25 10:46:36 +090010599agent-check
10600 Enable an auxiliary agent check which is run independently of a regular
Willy Tarreau81f5d942013-12-09 20:51:51 +010010601 health check. An agent health check is performed by making a TCP connection
10602 to the port set by the "agent-port" parameter and reading an ASCII string.
10603 The string is made of a series of words delimited by spaces, tabs or commas
10604 in any order, optionally terminated by '\r' and/or '\n', each consisting of :
Simon Hormand60d6912013-11-25 10:46:36 +090010605
Willy Tarreau81f5d942013-12-09 20:51:51 +010010606 - An ASCII representation of a positive integer percentage, e.g. "75%".
Simon Hormand60d6912013-11-25 10:46:36 +090010607 Values in this format will set the weight proportional to the initial
Willy Tarreauc5af3a62014-10-07 15:27:33 +020010608 weight of a server as configured when haproxy starts. Note that a zero
10609 weight is reported on the stats page as "DRAIN" since it has the same
10610 effect on the server (it's removed from the LB farm).
Simon Hormand60d6912013-11-25 10:46:36 +090010611
Nenad Merdanovic174dd372016-04-24 23:10:06 +020010612 - The string "maxconn:" followed by an integer (no space between). Values in
10613 this format will set the maxconn of a server. The maximum number of
10614 connections advertised needs to be multipled by the number of load balancers
10615 and different backends that use this health check to get the total number
10616 of connections the server might receive. Example: maxconn:30
10617
Willy Tarreau81f5d942013-12-09 20:51:51 +010010618 - The word "ready". This will turn the server's administrative state to the
10619 READY mode, thus cancelling any DRAIN or MAINT state
Simon Hormand60d6912013-11-25 10:46:36 +090010620
Willy Tarreau81f5d942013-12-09 20:51:51 +010010621 - The word "drain". This will turn the server's administrative state to the
10622 DRAIN mode, thus it will not accept any new connections other than those
10623 that are accepted via persistence.
Simon Hormand60d6912013-11-25 10:46:36 +090010624
Willy Tarreau81f5d942013-12-09 20:51:51 +010010625 - The word "maint". This will turn the server's administrative state to the
10626 MAINT mode, thus it will not accept any new connections at all, and health
10627 checks will be stopped.
Simon Hormand60d6912013-11-25 10:46:36 +090010628
Willy Tarreau81f5d942013-12-09 20:51:51 +010010629 - The words "down", "failed", or "stopped", optionally followed by a
10630 description string after a sharp ('#'). All of these mark the server's
10631 operating state as DOWN, but since the word itself is reported on the stats
10632 page, the difference allows an administrator to know if the situation was
10633 expected or not : the service may intentionally be stopped, may appear up
10634 but fail some validity tests, or may be seen as down (eg: missing process,
10635 or port not responding).
Simon Hormand60d6912013-11-25 10:46:36 +090010636
Willy Tarreau81f5d942013-12-09 20:51:51 +010010637 - The word "up" sets back the server's operating state as UP if health checks
10638 also report that the service is accessible.
Simon Hormand60d6912013-11-25 10:46:36 +090010639
Willy Tarreau81f5d942013-12-09 20:51:51 +010010640 Parameters which are not advertised by the agent are not changed. For
10641 example, an agent might be designed to monitor CPU usage and only report a
10642 relative weight and never interact with the operating status. Similarly, an
10643 agent could be designed as an end-user interface with 3 radio buttons
10644 allowing an administrator to change only the administrative state. However,
10645 it is important to consider that only the agent may revert its own actions,
10646 so if a server is set to DRAIN mode or to DOWN state using the agent, the
10647 agent must implement the other equivalent actions to bring the service into
10648 operations again.
Simon Hormand60d6912013-11-25 10:46:36 +090010649
Simon Horman2f1f9552013-11-25 10:46:37 +090010650 Failure to connect to the agent is not considered an error as connectivity
10651 is tested by the regular health check which is enabled by the "check"
Willy Tarreau81f5d942013-12-09 20:51:51 +010010652 parameter. Warning though, it is not a good idea to stop an agent after it
10653 reports "down", since only an agent reporting "up" will be able to turn the
10654 server up again. Note that the CLI on the Unix stats socket is also able to
Willy Tarreau989222a2016-01-15 10:26:26 +010010655 force an agent's result in order to work around a bogus agent if needed.
Simon Horman2f1f9552013-11-25 10:46:37 +090010656
Willy Tarreau81f5d942013-12-09 20:51:51 +010010657 Requires the "agent-port" parameter to be set. See also the "agent-inter"
10658 parameter.
Simon Hormand60d6912013-11-25 10:46:36 +090010659
10660 Supported in default-server: No
10661
James Brown55f9ff12015-10-21 18:19:05 -070010662agent-send <string>
10663 If this option is specified, haproxy will send the given string (verbatim)
10664 to the agent server upon connection. You could, for example, encode
10665 the backend name into this string, which would enable your agent to send
10666 different responses based on the backend. Make sure to include a '\n' if
10667 you want to terminate your request with a newline.
10668
Simon Hormand60d6912013-11-25 10:46:36 +090010669agent-inter <delay>
10670 The "agent-inter" parameter sets the interval between two agent checks
10671 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
10672
10673 Just as with every other time-based parameter, it may be entered in any
10674 other explicit unit among { us, ms, s, m, h, d }. The "agent-inter"
10675 parameter also serves as a timeout for agent checks "timeout check" is
10676 not set. In order to reduce "resonance" effects when multiple servers are
10677 hosted on the same hardware, the agent and health checks of all servers
10678 are started with a small time offset between them. It is also possible to
10679 add some random noise in the agent and health checks interval using the
10680 global "spread-checks" keyword. This makes sense for instance when a lot
10681 of backends use the same servers.
10682
10683 See also the "agent-check" and "agent-port" parameters.
10684
10685 Supported in default-server: Yes
10686
10687agent-port <port>
10688 The "agent-port" parameter sets the TCP port used for agent checks.
10689
10690 See also the "agent-check" and "agent-inter" parameters.
10691
10692 Supported in default-server: Yes
10693
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010694backup
10695 When "backup" is present on a server line, the server is only used in load
10696 balancing when all other non-backup servers are unavailable. Requests coming
10697 with a persistence cookie referencing the server will always be served
10698 though. By default, only the first operational backup server is used, unless
10699 the "allbackups" option is set in the backend. See also the "allbackups"
10700 option.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010701
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010702 Supported in default-server: No
10703
Emeric Brunef42d922012-10-11 16:11:36 +020010704ca-file <cafile>
10705 This setting is only available when support for OpenSSL was built in. It
10706 designates a PEM file from which to load CA certificates used to verify
10707 server's certificate.
10708
10709 Supported in default-server: No
10710
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010711check
10712 This option enables health checks on the server. By default, a server is
Patrick Mézardb7aeec62012-01-22 16:01:22 +010010713 always considered available. If "check" is set, the server is available when
10714 accepting periodic TCP connections, to ensure that it is really able to serve
10715 requests. The default address and port to send the tests to are those of the
10716 server, and the default source is the same as the one defined in the
10717 backend. It is possible to change the address using the "addr" parameter, the
10718 port using the "port" parameter, the source address using the "source"
10719 address, and the interval and timers using the "inter", "rise" and "fall"
Simon Hormanafc47ee2013-11-25 10:46:35 +090010720 parameters. The request method is define in the backend using the "httpchk",
10721 "smtpchk", "mysql-check", "pgsql-check" and "ssl-hello-chk" options. Please
10722 refer to those options and parameters for more information.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010723
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010724 Supported in default-server: No
10725
Willy Tarreau6c16adc2012-10-05 00:04:16 +020010726check-send-proxy
10727 This option forces emission of a PROXY protocol line with outgoing health
10728 checks, regardless of whether the server uses send-proxy or not for the
10729 normal traffic. By default, the PROXY protocol is enabled for health checks
10730 if it is already enabled for normal traffic and if no "port" nor "addr"
10731 directive is present. However, if such a directive is present, the
10732 "check-send-proxy" option needs to be used to force the use of the
10733 protocol. See also the "send-proxy" option for more information.
10734
10735 Supported in default-server: No
10736
Willy Tarreau763a95b2012-10-04 23:15:39 +020010737check-ssl
10738 This option forces encryption of all health checks over SSL, regardless of
10739 whether the server uses SSL or not for the normal traffic. This is generally
10740 used when an explicit "port" or "addr" directive is specified and SSL health
10741 checks are not inherited. It is important to understand that this option
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010742 inserts an SSL transport layer below the checks, so that a simple TCP connect
Willy Tarreau763a95b2012-10-04 23:15:39 +020010743 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
10744 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
10745 All SSL settings are common to health checks and traffic (eg: ciphers).
10746 See the "ssl" option for more information.
10747
10748 Supported in default-server: No
10749
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010750ciphers <ciphers>
10751 This option sets the string describing the list of cipher algorithms that is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010752 is negotiated during the SSL/TLS handshake with the server. The format of the
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010753 string is defined in "man 1 ciphers". When SSL is used to communicate with
10754 servers on the local network, it is common to see a weaker set of algorithms
10755 than what is used over the internet. Doing so reduces CPU usage on both the
10756 server and haproxy while still keeping it compatible with deployed software.
10757 Some algorithms such as RC4-SHA1 are reasonably cheap. If no security at all
10758 is needed and just connectivity, using DES can be appropriate.
10759
Willy Tarreau763a95b2012-10-04 23:15:39 +020010760 Supported in default-server: No
10761
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010762cookie <value>
10763 The "cookie" parameter sets the cookie value assigned to the server to
10764 <value>. This value will be checked in incoming requests, and the first
10765 operational server possessing the same value will be selected. In return, in
10766 cookie insertion or rewrite modes, this value will be assigned to the cookie
10767 sent to the client. There is nothing wrong in having several servers sharing
10768 the same cookie value, and it is in fact somewhat common between normal and
10769 backup servers. See also the "cookie" keyword in backend section.
10770
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010771 Supported in default-server: No
10772
Emeric Brunef42d922012-10-11 16:11:36 +020010773crl-file <crlfile>
10774 This setting is only available when support for OpenSSL was built in. It
10775 designates a PEM file from which to load certificate revocation list used
10776 to verify server's certificate.
10777
10778 Supported in default-server: No
10779
Emeric Bruna7aa3092012-10-26 12:58:00 +020010780crt <cert>
10781 This setting is only available when support for OpenSSL was built in.
10782 It designates a PEM file from which to load both a certificate and the
10783 associated private key. This file can be built by concatenating both PEM
10784 files into one. This certificate will be sent if the server send a client
10785 certificate request.
10786
10787 Supported in default-server: No
10788
Willy Tarreau96839092010-03-29 10:02:24 +020010789disabled
10790 The "disabled" keyword starts the server in the "disabled" state. That means
10791 that it is marked down in maintenance mode, and no connection other than the
10792 ones allowed by persist mode will reach it. It is very well suited to setup
10793 new servers, because normal traffic will never reach them, while it is still
10794 possible to test the service by making use of the force-persist mechanism.
10795
10796 Supported in default-server: No
10797
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010798error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +010010799 If health observing is enabled, the "error-limit" parameter specifies the
10800 number of consecutive errors that triggers event selected by the "on-error"
10801 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010802
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010803 Supported in default-server: Yes
10804
10805 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010806
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010807fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010808 The "fall" parameter states that a server will be considered as dead after
10809 <count> consecutive unsuccessful health checks. This value defaults to 3 if
10810 unspecified. See also the "check", "inter" and "rise" parameters.
10811
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010812 Supported in default-server: Yes
10813
Emeric Brun8694b9a2012-10-05 14:39:07 +020010814force-sslv3
10815 This option enforces use of SSLv3 only when SSL is used to communicate with
10816 the server. SSLv3 is generally less expensive than the TLS counterparts for
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010817 high connection rates. This option is also available on global statement
10818 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010819
10820 Supported in default-server: No
10821
10822force-tlsv10
10823 This option enforces use of TLSv1.0 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010824 the server. This option is also available on global statement
10825 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010826
10827 Supported in default-server: No
10828
10829force-tlsv11
10830 This option enforces use of TLSv1.1 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010831 the server. This option is also available on global statement
10832 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010833
10834 Supported in default-server: No
10835
10836force-tlsv12
10837 This option enforces use of TLSv1.2 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010838 the server. This option is also available on global statement
10839 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010840
10841 Supported in default-server: No
10842
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010843id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +020010844 Set a persistent ID for the server. This ID must be positive and unique for
10845 the proxy. An unused ID will automatically be assigned if unset. The first
10846 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010847
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010848 Supported in default-server: No
10849
Willy Tarreau6a031d12016-11-07 19:42:35 +010010850init-addr {last | libc | none | <ip>},[...]*
10851 Indicate in what order the server's address should be resolved upon startup
10852 if it uses an FQDN. Attempts are made to resolve the address by applying in
10853 turn each of the methods mentionned in the comma-delimited list. The first
10854 method which succeeds is used. If the end of the list is reached without
10855 finding a working method, an error is thrown. Method "last" suggests to pick
10856 the address which appears in the state file (see "server-state-file"). Method
10857 "libc" uses the libc's internal resolver (gethostbyname() or getaddrinfo()
10858 depending on the operating system and build options). Method "none"
10859 specifically indicates that the server should start without any valid IP
10860 address in a down state. It can be useful to ignore some DNS issues upon
10861 startup, waiting for the situation to get fixed later. Finally, an IP address
10862 (IPv4 or IPv6) may be provided. It can be the currently known address of the
10863 server (eg: filled by a configuration generator), or the address of a dummy
10864 server used to catch old sessions and present them with a decent error
10865 message for example. When the "first" load balancing algorithm is used, this
10866 IP address could point to a fake server used to trigger the creation of new
10867 instances on the fly. This option defaults to "last,libc" indicating that the
10868 previous address found in the state file (if any) is used first, otherwise
10869 the libc's resolver is used. This ensures continued compatibility with the
10870 historic behaviour.
10871
10872 Example:
10873 defaults
10874 # never fail on address resolution
10875 default-server init-addr last,libc,none
10876
10877 Supported in default-server: Yes
10878
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010879inter <delay>
10880fastinter <delay>
10881downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010882 The "inter" parameter sets the interval between two consecutive health checks
10883 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
10884 It is also possible to use "fastinter" and "downinter" to optimize delays
10885 between checks depending on the server state :
10886
Pieter Baauw44fc9df2015-09-17 21:30:46 +020010887 Server state | Interval used
10888 ----------------------------------------+----------------------------------
10889 UP 100% (non-transitional) | "inter"
10890 ----------------------------------------+----------------------------------
10891 Transitionally UP (going down "fall"), | "fastinter" if set,
10892 Transitionally DOWN (going up "rise"), | "inter" otherwise.
10893 or yet unchecked. |
10894 ----------------------------------------+----------------------------------
10895 DOWN 100% (non-transitional) | "downinter" if set,
10896 | "inter" otherwise.
10897 ----------------------------------------+----------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +010010898
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010899 Just as with every other time-based parameter, they can be entered in any
10900 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
10901 serves as a timeout for health checks sent to servers if "timeout check" is
10902 not set. In order to reduce "resonance" effects when multiple servers are
Simon Hormand60d6912013-11-25 10:46:36 +090010903 hosted on the same hardware, the agent and health checks of all servers
10904 are started with a small time offset between them. It is also possible to
10905 add some random noise in the agent and health checks interval using the
10906 global "spread-checks" keyword. This makes sense for instance when a lot
10907 of backends use the same servers.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010908
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010909 Supported in default-server: Yes
10910
10911maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010912 The "maxconn" parameter specifies the maximal number of concurrent
10913 connections that will be sent to this server. If the number of incoming
10914 concurrent requests goes higher than this value, they will be queued, waiting
10915 for a connection to be released. This parameter is very important as it can
10916 save fragile servers from going down under extreme loads. If a "minconn"
10917 parameter is specified, the limit becomes dynamic. The default value is "0"
10918 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
10919 the backend's "fullconn" keyword.
10920
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010921 Supported in default-server: Yes
10922
10923maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010924 The "maxqueue" parameter specifies the maximal number of connections which
10925 will wait in the queue for this server. If this limit is reached, next
10926 requests will be redispatched to other servers instead of indefinitely
10927 waiting to be served. This will break persistence but may allow people to
10928 quickly re-log in when the server they try to connect to is dying. The
10929 default value is "0" which means the queue is unlimited. See also the
10930 "maxconn" and "minconn" parameters.
10931
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010932 Supported in default-server: Yes
10933
10934minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010935 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
10936 limit following the backend's load. The server will always accept at least
10937 <minconn> connections, never more than <maxconn>, and the limit will be on
10938 the ramp between both values when the backend has less than <fullconn>
10939 concurrent connections. This makes it possible to limit the load on the
10940 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010941 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010942 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010943
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010944 Supported in default-server: Yes
10945
Willy Tarreaud72f0f32015-10-13 14:50:22 +020010946namespace <name>
10947 On Linux, it is possible to specify which network namespace a socket will
10948 belong to. This directive makes it possible to explicitly bind a server to
10949 a namespace different from the default one. Please refer to your operating
10950 system's documentation to find more details about network namespaces.
10951
Willy Tarreau2a3fb1c2015-02-05 16:47:07 +010010952no-ssl-reuse
10953 This option disables SSL session reuse when SSL is used to communicate with
10954 the server. It will force the server to perform a full handshake for every
10955 new connection. It's probably only useful for benchmarking, troubleshooting,
10956 and for paranoid users.
10957
10958 Supported in default-server: No
10959
Emeric Brun9b3009b2012-10-05 11:55:06 +020010960no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010961 This option disables support for SSLv3 when SSL is used to communicate with
10962 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emeric Brun8694b9a2012-10-05 14:39:07 +020010963 using any configuration option. See also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010964
Willy Tarreau763a95b2012-10-04 23:15:39 +020010965 Supported in default-server: No
10966
Emeric Brunf9c5c472012-10-11 15:28:34 +020010967no-tls-tickets
10968 This setting is only available when support for OpenSSL was built in. It
10969 disables the stateless session resumption (RFC 5077 TLS Ticket
10970 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010971 session resumption is more expensive in CPU usage for servers. This option
10972 is also available on global statement "ssl-default-server-options".
Emeric Brunf9c5c472012-10-11 15:28:34 +020010973
10974 Supported in default-server: No
10975
Emeric Brun9b3009b2012-10-05 11:55:06 +020010976no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +020010977 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020010978 the server. Note that SSLv2 is disabled in the code and cannot be enabled
10979 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010980 often makes sense to disable it when communicating with local servers. This
10981 option is also available on global statement "ssl-default-server-options".
10982 See also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +020010983
Willy Tarreau763a95b2012-10-04 23:15:39 +020010984 Supported in default-server: No
10985
Emeric Brun9b3009b2012-10-05 11:55:06 +020010986no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +020010987 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020010988 the server. Note that SSLv2 is disabled in the code and cannot be enabled
10989 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010990 often makes sense to disable it when communicating with local servers. This
10991 option is also available on global statement "ssl-default-server-options".
10992 See also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +020010993
Willy Tarreau763a95b2012-10-04 23:15:39 +020010994 Supported in default-server: No
10995
Emeric Brun9b3009b2012-10-05 11:55:06 +020010996no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +020010997 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010998 the server. Note that SSLv2 is disabled in the code and cannot be enabled
10999 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011000 often makes sense to disable it when communicating with local servers. This
11001 option is also available on global statement "ssl-default-server-options".
11002 See also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011003
Willy Tarreau763a95b2012-10-04 23:15:39 +020011004 Supported in default-server: No
11005
Simon Hormanfa461682011-06-25 09:39:49 +090011006non-stick
11007 Never add connections allocated to this sever to a stick-table.
11008 This may be used in conjunction with backup to ensure that
11009 stick-table persistence is disabled for backup servers.
11010
Willy Tarreau763a95b2012-10-04 23:15:39 +020011011 Supported in default-server: No
11012
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011013observe <mode>
11014 This option enables health adjusting based on observing communication with
11015 the server. By default this functionality is disabled and enabling it also
11016 requires to enable health checks. There are two supported modes: "layer4" and
11017 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
11018 significant. In layer7, which is only allowed for http proxies, responses
11019 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +010011020 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011021
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011022 Supported in default-server: No
11023
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011024 See also the "check", "on-error" and "error-limit".
11025
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011026on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011027 Select what should happen when enough consecutive errors are detected.
11028 Currently, four modes are available:
11029 - fastinter: force fastinter
11030 - fail-check: simulate a failed check, also forces fastinter (default)
11031 - sudden-death: simulate a pre-fatal failed health check, one more failed
11032 check will mark a server down, forces fastinter
11033 - mark-down: mark the server immediately down and force fastinter
11034
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011035 Supported in default-server: Yes
11036
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011037 See also the "check", "observe" and "error-limit".
11038
Simon Hormane0d1bfb2011-06-21 14:34:58 +090011039on-marked-down <action>
11040 Modify what occurs when a server is marked down.
11041 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070011042 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
11043 all connections to the server are immediately terminated when the server
11044 goes down. It might be used if the health check detects more complex cases
11045 than a simple connection status, and long timeouts would cause the service
11046 to remain unresponsive for too long a time. For instance, a health check
11047 might detect that a database is stuck and that there's no chance to reuse
11048 existing connections anymore. Connections killed this way are logged with
11049 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +090011050
11051 Actions are disabled by default
11052
11053 Supported in default-server: Yes
11054
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070011055on-marked-up <action>
11056 Modify what occurs when a server is marked up.
11057 Currently one action is available:
11058 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
11059 done only if the server is not in backup state and if it is not disabled
11060 (it must have an effective weight > 0). This can be used sometimes to force
11061 an active server to take all the traffic back after recovery when dealing
11062 with long sessions (eg: LDAP, SQL, ...). Doing this can cause more trouble
11063 than it tries to solve (eg: incomplete transactions), so use this feature
11064 with extreme care. Sessions killed because a server comes up are logged
11065 with an 'U' termination code (for "Up").
11066
11067 Actions are disabled by default
11068
11069 Supported in default-server: Yes
11070
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011071port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011072 Using the "port" parameter, it becomes possible to use a different port to
11073 send health-checks. On some servers, it may be desirable to dedicate a port
11074 to a specific component able to perform complex tests which are more suitable
11075 to health-checks than the application. It is common to run a simple script in
11076 inetd for instance. This parameter is ignored if the "check" parameter is not
11077 set. See also the "addr" parameter.
11078
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011079 Supported in default-server: Yes
11080
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011081redir <prefix>
11082 The "redir" parameter enables the redirection mode for all GET and HEAD
11083 requests addressing this server. This means that instead of having HAProxy
11084 forward the request to the server, it will send an "HTTP 302" response with
11085 the "Location" header composed of this prefix immediately followed by the
11086 requested URI beginning at the leading '/' of the path component. That means
11087 that no trailing slash should be used after <prefix>. All invalid requests
11088 will be rejected, and all non-GET or HEAD requests will be normally served by
11089 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011090 mangling nor cookie insertion is possible in the response. However, cookies in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011091 requests are still analysed, making this solution completely usable to direct
11092 users to a remote location in case of local disaster. Main use consists in
11093 increasing bandwidth for static servers by having the clients directly
11094 connect to them. Note: never use a relative location here, it would cause a
11095 loop between the client and HAProxy!
11096
11097 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
11098
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011099 Supported in default-server: No
11100
11101rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011102 The "rise" parameter states that a server will be considered as operational
11103 after <count> consecutive successful health checks. This value defaults to 2
11104 if unspecified. See also the "check", "inter" and "fall" parameters.
11105
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011106 Supported in default-server: Yes
11107
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011108resolve-prefer <family>
11109 When DNS resolution is enabled for a server and multiple IP addresses from
11110 different families are returned, HAProxy will prefer using an IP address
11111 from the family mentioned in the "resolve-prefer" parameter.
11112 Available families: "ipv4" and "ipv6"
11113
Baptiste Assmannc4aabae2015-08-04 22:43:06 +020011114 Default value: ipv6
11115
11116 Supported in default-server: Yes
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011117
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020011118 Example:
11119
11120 server s1 app1.domain.com:80 resolvers mydns resolve-prefer ipv6
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011121
Thierry Fournierac88cfe2016-02-17 22:05:30 +010011122resolve-net <network>[,<network[,...]]
11123 This options prioritize th choice of an ip address matching a network. This is
11124 useful with clouds to prefer a local ip. In some cases, a cloud high
11125 avalailibility service can be announced with many ip addresses on many
11126 differents datacenters. The latency between datacenter is not negligible, so
11127 this patch permitsto prefers a local datacenter. If none address matchs the
11128 configured network, another address is selected.
11129
11130 Supported in default-server: Yes
11131
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020011132 Example:
11133
11134 server s1 app1.domain.com:80 resolvers mydns resolve-net 10.0.0.0/8
Thierry Fournierac88cfe2016-02-17 22:05:30 +010011135
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011136resolvers <id>
11137 Points to an existing "resolvers" section to resolve current server's
11138 hostname.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011139 In order to be operational, DNS resolution requires that health check is
11140 enabled on the server. Actually, health checks triggers the DNS resolution.
11141 You must precise one 'resolvers' parameter on each server line where DNS
11142 resolution is required.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011143
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011144 Supported in default-server: No
11145
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020011146 Example:
11147
11148 server s1 app1.domain.com:80 check resolvers mydns
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011149
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020011150 See also section 5.3
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011151
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010011152send-proxy
11153 The "send-proxy" parameter enforces use of the PROXY protocol over any
11154 connection established to this server. The PROXY protocol informs the other
11155 end about the layer 3/4 addresses of the incoming connection, so that it can
11156 know the client's address or the public address it accessed to, whatever the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010011157 upper layer protocol. For connections accepted by an "accept-proxy" or
11158 "accept-netscaler-cip" listener, the advertised address will be used. Only
11159 TCPv4 and TCPv6 address families are supported. Other families such as
11160 Unix sockets, will report an UNKNOWN family. Servers using this option can
11161 fully be chained to another instance of haproxy listening with an
11162 "accept-proxy" setting. This setting must not be used if the server isn't
11163 aware of the protocol. When health checks are sent to the server, the PROXY
11164 protocol is automatically used when this option is set, unless there is an
11165 explicit "port" or "addr" directive, in which case an explicit
11166 "check-send-proxy" directive would also be needed to use the PROXY protocol.
11167 See also the "accept-proxy" and "accept-netscaler-cip" option of the "bind"
11168 keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010011169
11170 Supported in default-server: No
11171
David Safb76832014-05-08 23:42:08 -040011172send-proxy-v2
11173 The "send-proxy-v2" parameter enforces use of the PROXY protocol version 2
11174 over any connection established to this server. The PROXY protocol informs
11175 the other end about the layer 3/4 addresses of the incoming connection, so
11176 that it can know the client's address or the public address it accessed to,
11177 whatever the upper layer protocol. This setting must not be used if the
11178 server isn't aware of this version of the protocol. See also the "send-proxy"
11179 option of the "bind" keyword.
11180
11181 Supported in default-server: No
11182
11183send-proxy-v2-ssl
11184 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
11185 2 over any connection established to this server. The PROXY protocol informs
11186 the other end about the layer 3/4 addresses of the incoming connection, so
11187 that it can know the client's address or the public address it accessed to,
11188 whatever the upper layer protocol. In addition, the SSL information extension
11189 of the PROXY protocol is added to the PROXY protocol header. This setting
11190 must not be used if the server isn't aware of this version of the protocol.
11191 See also the "send-proxy-v2" option of the "bind" keyword.
11192
11193 Supported in default-server: No
11194
11195send-proxy-v2-ssl-cn
11196 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
11197 2 over any connection established to this server. The PROXY protocol informs
11198 the other end about the layer 3/4 addresses of the incoming connection, so
11199 that it can know the client's address or the public address it accessed to,
11200 whatever the upper layer protocol. In addition, the SSL information extension
11201 of the PROXY protocol, along along with the Common Name from the subject of
11202 the client certificate (if any), is added to the PROXY protocol header. This
11203 setting must not be used if the server isn't aware of this version of the
11204 protocol. See also the "send-proxy-v2" option of the "bind" keyword.
11205
11206 Supported in default-server: No
11207
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011208slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011209 The "slowstart" parameter for a server accepts a value in milliseconds which
11210 indicates after how long a server which has just come back up will run at
11211 full speed. Just as with every other time-based parameter, it can be entered
11212 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
11213 linearly from 0 to 100% during this time. The limitation applies to two
11214 parameters :
11215
11216 - maxconn: the number of connections accepted by the server will grow from 1
11217 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
11218
11219 - weight: when the backend uses a dynamic weighted algorithm, the weight
11220 grows linearly from 1 to 100%. In this case, the weight is updated at every
11221 health-check. For this reason, it is important that the "inter" parameter
11222 is smaller than the "slowstart", in order to maximize the number of steps.
11223
11224 The slowstart never applies when haproxy starts, otherwise it would cause
11225 trouble to running servers. It only applies when a server has been previously
11226 seen as failed.
11227
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011228 Supported in default-server: Yes
11229
Willy Tarreau732eac42015-07-09 11:40:25 +020011230sni <expression>
11231 The "sni" parameter evaluates the sample fetch expression, converts it to a
11232 string and uses the result as the host name sent in the SNI TLS extension to
11233 the server. A typical use case is to send the SNI received from the client in
11234 a bridged HTTPS scenario, using the "ssl_fc_sni" sample fetch for the
11235 expression, though alternatives such as req.hdr(host) can also make sense.
11236
11237 Supported in default-server: no
11238
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020011239source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +020011240source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020011241source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011242 The "source" parameter sets the source address which will be used when
11243 connecting to the server. It follows the exact same parameters and principle
11244 as the backend "source" keyword, except that it only applies to the server
11245 referencing it. Please consult the "source" keyword for details.
11246
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020011247 Additionally, the "source" statement on a server line allows one to specify a
11248 source port range by indicating the lower and higher bounds delimited by a
11249 dash ('-'). Some operating systems might require a valid IP address when a
11250 source port range is specified. It is permitted to have the same IP/range for
11251 several servers. Doing so makes it possible to bypass the maximum of 64k
11252 total concurrent connections. The limit will then reach 64k connections per
11253 server.
11254
Lukas Tribus7d56c6d2016-09-13 09:51:15 +000011255 Since Linux 4.2/libc 2.23 IP_BIND_ADDRESS_NO_PORT is set for connections
11256 specifying the source address without port(s).
11257
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011258 Supported in default-server: No
11259
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011260ssl
Willy Tarreau44f65392013-06-25 07:56:20 +020011261 This option enables SSL ciphering on outgoing connections to the server. It
11262 is critical to verify server certificates using "verify" when using SSL to
11263 connect to servers, otherwise the communication is prone to trivial man in
11264 the-middle attacks rendering SSL useless. When this option is used, health
11265 checks are automatically sent in SSL too unless there is a "port" or an
11266 "addr" directive indicating the check should be sent to a different location.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011267 See the "check-ssl" option to force SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +020011268
11269 Supported in default-server: No
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011270
Willy Tarreau163d4622015-10-13 16:16:41 +020011271tcp-ut <delay>
11272 Sets the TCP User Timeout for all outgoing connections to this server. This
11273 option is available on Linux since version 2.6.37. It allows haproxy to
11274 configure a timeout for sockets which contain data not receiving an
11275 acknoledgement for the configured delay. This is especially useful on
11276 long-lived connections experiencing long idle periods such as remote
11277 terminals or database connection pools, where the client and server timeouts
11278 must remain high to allow a long period of idle, but where it is important to
11279 detect that the server has disappeared in order to release all resources
11280 associated with its connection (and the client's session). One typical use
11281 case is also to force dead server connections to die when health checks are
11282 too slow or during a soft reload since health checks are then disabled. The
11283 argument is a delay expressed in milliseconds by default. This only works for
11284 regular TCP connections, and is ignored for other protocols.
11285
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011286track [<proxy>/]<server>
Willy Tarreau32091232014-05-16 13:52:00 +020011287 This option enables ability to set the current state of the server by tracking
11288 another one. It is possible to track a server which itself tracks another
11289 server, provided that at the end of the chain, a server has health checks
11290 enabled. If <proxy> is omitted the current one is used. If disable-on-404 is
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011291 used, it has to be enabled on both proxies.
11292
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011293 Supported in default-server: No
11294
Emeric Brunef42d922012-10-11 16:11:36 +020011295verify [none|required]
11296 This setting is only available when support for OpenSSL was built in. If set
Emeric Brun850efd52014-01-29 12:24:34 +010011297 to 'none', server certificate is not verified. In the other case, The
11298 certificate provided by the server is verified using CAs from 'ca-file'
11299 and optional CRLs from 'crl-file'. If 'ssl_server_verify' is not specified
11300 in global section, this is the default. On verify failure the handshake
Willy Tarreau44f65392013-06-25 07:56:20 +020011301 is aborted. It is critically important to verify server certificates when
11302 using SSL to connect to servers, otherwise the communication is prone to
11303 trivial man-in-the-middle attacks rendering SSL totally useless.
Emeric Brunef42d922012-10-11 16:11:36 +020011304
11305 Supported in default-server: No
11306
Evan Broderbe554312013-06-27 00:05:25 -070011307verifyhost <hostname>
11308 This setting is only available when support for OpenSSL was built in, and
11309 only takes effect if 'verify required' is also specified. When set, the
11310 hostnames in the subject and subjectAlternateNames of the certificate
11311 provided by the server are checked. If none of the hostnames in the
11312 certificate match the specified hostname, the handshake is aborted. The
11313 hostnames in the server-provided certificate may include wildcards.
11314
11315 Supported in default-server: No
11316
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011317weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011318 The "weight" parameter is used to adjust the server's weight relative to
11319 other servers. All servers will receive a load proportional to their weight
11320 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +020011321 load. The default weight is 1, and the maximal value is 256. A value of 0
11322 means the server will not participate in load-balancing but will still accept
11323 persistent connections. If this parameter is used to distribute the load
11324 according to server's capacity, it is recommended to start with values which
11325 can both grow and shrink, for instance between 10 and 100 to leave enough
11326 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011327
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011328 Supported in default-server: Yes
11329
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011330
Cyril Bonté46175dd2015-07-02 22:45:32 +0200113315.3. Server IP address resolution using DNS
11332-------------------------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011333
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011334HAProxy allows using a host name on the server line to retrieve its IP address
11335using name servers. By default, HAProxy resolves the name when parsing the
11336configuration file, at startup and cache the result for the process' life.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011337This is not sufficient in some cases, such as in Amazon where a server's IP
11338can change after a reboot or an ELB Virtual IP can change based on current
11339workload.
11340This chapter describes how HAProxy can be configured to process server's name
11341resolution at run time.
11342Whether run time server name resolution has been enable or not, HAProxy will
11343carry on doing the first resolution when parsing the configuration.
11344
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011345Bear in mind that DNS resolution is triggered by health checks. This makes
11346health checks mandatory to allow DNS resolution.
11347
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011348
Cyril Bonté46175dd2015-07-02 22:45:32 +0200113495.3.1. Global overview
11350----------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011351
11352As we've seen in introduction, name resolution in HAProxy occurs at two
11353different steps of the process life:
11354
11355 1. when starting up, HAProxy parses the server line definition and matches a
11356 host name. It uses libc functions to get the host name resolved. This
11357 resolution relies on /etc/resolv.conf file.
11358
11359 2. at run time, when HAProxy gets prepared to run a health check on a server,
11360 it verifies if the current name resolution is still considered as valid.
11361 If not, it processes a new resolution, in parallel of the health check.
11362
11363A few other events can trigger a name resolution at run time:
11364 - when a server's health check ends up in a connection timeout: this may be
11365 because the server has a new IP address. So we need to trigger a name
11366 resolution to know this new IP.
11367
11368A few things important to notice:
11369 - all the name servers are queried in the mean time. HAProxy will process the
11370 first valid response.
11371
11372 - a resolution is considered as invalid (NX, timeout, refused), when all the
11373 servers return an error.
11374
11375
Cyril Bonté46175dd2015-07-02 22:45:32 +0200113765.3.2. The resolvers section
11377----------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011378
11379This section is dedicated to host information related to name resolution in
11380HAProxy.
11381There can be as many as resolvers section as needed. Each section can contain
11382many name servers.
11383
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011384When multiple name servers are configured in a resolvers section, then HAProxy
11385uses the first valid response. In case of invalid responses, only the last one
11386is treated. Purpose is to give the chance to a slow server to deliver a valid
11387answer after a fast faulty or outdated server.
11388
11389When each server returns a different error type, then only the last error is
11390used by HAProxy to decide what type of behavior to apply.
11391
11392Two types of behavior can be applied:
11393 1. stop DNS resolution
11394 2. replay the DNS query with a new query type
11395 In such case, the following types are applied in this exact order:
11396 1. ANY query type
11397 2. query type corresponding to family pointed by resolve-prefer
11398 server's parameter
11399 3. remaining family type
11400
11401HAProxy stops DNS resolution when the following errors occur:
11402 - invalid DNS response packet
11403 - wrong name in the query section of the response
11404 - NX domain
11405 - Query refused by server
11406 - CNAME not pointing to an IP address
11407
11408HAProxy tries a new query type when the following errors occur:
11409 - no Answer records in the response
11410 - DNS response truncated
11411 - Error in DNS response
11412 - No expected DNS records found in the response
11413 - name server timeout
11414
11415For example, with 2 name servers configured in a resolvers section:
11416 - first response is valid and is applied directly, second response is ignored
11417 - first response is invalid and second one is valid, then second response is
11418 applied;
11419 - first response is a NX domain and second one a truncated response, then
11420 HAProxy replays the query with a new type;
11421 - first response is truncated and second one is a NX Domain, then HAProxy
11422 stops resolution.
11423
11424
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011425resolvers <resolvers id>
11426 Creates a new name server list labelled <resolvers id>
11427
11428A resolvers section accept the following parameters:
11429
11430nameserver <id> <ip>:<port>
11431 DNS server description:
11432 <id> : label of the server, should be unique
11433 <ip> : IP address of the server
11434 <port> : port where the DNS service actually runs
11435
11436hold <status> <period>
11437 Defines <period> during which the last name resolution should be kept based
11438 on last resolution <status>
Baptiste Assmann987e16d2016-11-02 22:23:31 +010011439 <status> : last name resolution status. Acceptable values are "nx",
11440 "other", "refused", "timeout", "valid".
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011441 <period> : interval between two successive name resolution when the last
11442 answer was in <status>. It follows the HAProxy time format.
11443 <period> is in milliseconds by default.
11444
Baptiste Assmann987e16d2016-11-02 22:23:31 +010011445 Default value is 10s for "valid" and 30s for others.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011446
11447 Note: since the name resolution is triggered by the health checks, a new
11448 resolution is triggered after <period> modulo the <inter> parameter of
11449 the healch check.
11450
11451resolve_retries <nb>
11452 Defines the number <nb> of queries to send to resolve a server name before
11453 giving up.
11454 Default value: 3
11455
Baptiste Assmann62b75b42015-09-09 01:11:36 +020011456 A retry occurs on name server timeout or when the full sequence of DNS query
11457 type failover is over and we need to start up from the default ANY query
11458 type.
11459
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011460timeout <event> <time>
11461 Defines timeouts related to name resolution
11462 <event> : the event on which the <time> timeout period applies to.
11463 events available are:
11464 - retry: time between two DNS queries, when no response have
11465 been received.
11466 Default value: 1s
11467 <time> : time related to the event. It follows the HAProxy time format.
11468 <time> is expressed in milliseconds.
11469
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020011470 Example:
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011471
11472 resolvers mydns
11473 nameserver dns1 10.0.0.1:53
11474 nameserver dns2 10.0.0.2:53
11475 resolve_retries 3
11476 timeout retry 1s
Baptiste Assmann987e16d2016-11-02 22:23:31 +010011477 hold other 30s
11478 hold refused 30s
11479 hold nx 30s
11480 hold timeout 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020011481 hold valid 10s
11482
11483
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200114846. HTTP header manipulation
11485---------------------------
11486
11487In HTTP mode, it is possible to rewrite, add or delete some of the request and
11488response headers based on regular expressions. It is also possible to block a
11489request or a response if a particular header matches a regular expression,
11490which is enough to stop most elementary protocol attacks, and to protect
Willy Tarreau70dffda2014-01-30 03:07:23 +010011491against information leak from the internal network.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011492
Willy Tarreau70dffda2014-01-30 03:07:23 +010011493If HAProxy encounters an "Informational Response" (status code 1xx), it is able
11494to process all rsp* rules which can allow, deny, rewrite or delete a header,
11495but it will refuse to add a header to any such messages as this is not
11496HTTP-compliant. The reason for still processing headers in such responses is to
11497stop and/or fix any possible information leak which may happen, for instance
11498because another downstream equipment would unconditionally add a header, or if
11499a server name appears there. When such messages are seen, normal processing
11500still occurs on the next non-informational messages.
Willy Tarreau816b9792009-09-15 21:25:21 +020011501
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011502This section covers common usage of the following keywords, described in detail
11503in section 4.2 :
11504
11505 - reqadd <string>
11506 - reqallow <search>
11507 - reqiallow <search>
11508 - reqdel <search>
11509 - reqidel <search>
11510 - reqdeny <search>
11511 - reqideny <search>
11512 - reqpass <search>
11513 - reqipass <search>
11514 - reqrep <search> <replace>
11515 - reqirep <search> <replace>
11516 - reqtarpit <search>
11517 - reqitarpit <search>
11518 - rspadd <string>
11519 - rspdel <search>
11520 - rspidel <search>
11521 - rspdeny <search>
11522 - rspideny <search>
11523 - rsprep <search> <replace>
11524 - rspirep <search> <replace>
11525
11526With all these keywords, the same conventions are used. The <search> parameter
11527is a POSIX extended regular expression (regex) which supports grouping through
11528parenthesis (without the backslash). Spaces and other delimiters must be
11529prefixed with a backslash ('\') to avoid confusion with a field delimiter.
11530Other characters may be prefixed with a backslash to change their meaning :
11531
11532 \t for a tab
11533 \r for a carriage return (CR)
11534 \n for a new line (LF)
11535 \ to mark a space and differentiate it from a delimiter
11536 \# to mark a sharp and differentiate it from a comment
11537 \\ to use a backslash in a regex
11538 \\\\ to use a backslash in the text (*2 for regex, *2 for haproxy)
11539 \xXX to write the ASCII hex code XX as in the C language
11540
11541The <replace> parameter contains the string to be used to replace the largest
11542portion of text matching the regex. It can make use of the special characters
11543above, and can reference a substring which is delimited by parenthesis in the
11544regex, by writing a backslash ('\') immediately followed by one digit from 0 to
115459 indicating the group position (0 designating the entire line). This practice
11546is very common to users of the "sed" program.
11547
11548The <string> parameter represents the string which will systematically be added
11549after the last header line. It can also use special character sequences above.
11550
11551Notes related to these keywords :
11552---------------------------------
11553 - these keywords are not always convenient to allow/deny based on header
11554 contents. It is strongly recommended to use ACLs with the "block" keyword
11555 instead, resulting in far more flexible and manageable rules.
11556
11557 - lines are always considered as a whole. It is not possible to reference
11558 a header name only or a value only. This is important because of the way
11559 headers are written (notably the number of spaces after the colon).
11560
11561 - the first line is always considered as a header, which makes it possible to
11562 rewrite or filter HTTP requests URIs or response codes, but in turn makes
11563 it harder to distinguish between headers and request line. The regex prefix
11564 ^[^\ \t]*[\ \t] matches any HTTP method followed by a space, and the prefix
11565 ^[^ \t:]*: matches any header name followed by a colon.
11566
11567 - for performances reasons, the number of characters added to a request or to
11568 a response is limited at build time to values between 1 and 4 kB. This
11569 should normally be far more than enough for most usages. If it is too short
11570 on occasional usages, it is possible to gain some space by removing some
11571 useless headers before adding new ones.
11572
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011573 - keywords beginning with "reqi" and "rspi" are the same as their counterpart
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011574 without the 'i' letter except that they ignore case when matching patterns.
11575
11576 - when a request passes through a frontend then a backend, all req* rules
11577 from the frontend will be evaluated, then all req* rules from the backend
11578 will be evaluated. The reverse path is applied to responses.
11579
11580 - req* statements are applied after "block" statements, so that "block" is
11581 always the first one, but before "use_backend" in order to permit rewriting
Willy Tarreaud72758d2010-01-12 10:42:19 +010011582 before switching.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011583
11584
Willy Tarreau74ca5042013-06-11 23:12:07 +0200115857. Using ACLs and fetching samples
11586----------------------------------
11587
11588Haproxy is capable of extracting data from request or response streams, from
11589client or server information, from tables, environmental information etc...
11590The action of extracting such data is called fetching a sample. Once retrieved,
11591these samples may be used for various purposes such as a key to a stick-table,
11592but most common usages consist in matching them against predefined constant
11593data called patterns.
11594
11595
115967.1. ACL basics
11597---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011598
11599The use of Access Control Lists (ACL) provides a flexible solution to perform
11600content switching and generally to take decisions based on content extracted
11601from the request, the response or any environmental status. The principle is
11602simple :
11603
Willy Tarreau74ca5042013-06-11 23:12:07 +020011604 - extract a data sample from a stream, table or the environment
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011605 - optionally apply some format conversion to the extracted sample
Willy Tarreau74ca5042013-06-11 23:12:07 +020011606 - apply one or multiple pattern matching methods on this sample
11607 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011608
Willy Tarreau74ca5042013-06-11 23:12:07 +020011609The actions generally consist in blocking a request, selecting a backend, or
11610adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011611
11612In order to define a test, the "acl" keyword is used. The syntax is :
11613
Willy Tarreau74ca5042013-06-11 23:12:07 +020011614 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011615
11616This creates a new ACL <aclname> or completes an existing one with new tests.
11617Those tests apply to the portion of request/response specified in <criterion>
11618and may be adjusted with optional flags [flags]. Some criteria also support
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011619an operator which may be specified before the set of values. Optionally some
11620conversion operators may be applied to the sample, and they will be specified
11621as a comma-delimited list of keywords just after the first keyword. The values
11622are of the type supported by the criterion, and are separated by spaces.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011623
11624ACL names must be formed from upper and lower case letters, digits, '-' (dash),
11625'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
11626which means that "my_acl" and "My_Acl" are two different ACLs.
11627
11628There is no enforced limit to the number of ACLs. The unused ones do not affect
11629performance, they just consume a small amount of memory.
11630
Willy Tarreau74ca5042013-06-11 23:12:07 +020011631The criterion generally is the name of a sample fetch method, or one of its ACL
11632specific declinations. The default test method is implied by the output type of
11633this sample fetch method. The ACL declinations can describe alternate matching
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011634methods of a same sample fetch method. The sample fetch methods are the only
11635ones supporting a conversion.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011636
11637Sample fetch methods return data which can be of the following types :
11638 - boolean
11639 - integer (signed or unsigned)
11640 - IPv4 or IPv6 address
11641 - string
11642 - data block
11643
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011644Converters transform any of these data into any of these. For example, some
11645converters might convert a string to a lower-case string while other ones
11646would turn a string to an IPv4 address, or apply a netmask to an IP address.
11647The resulting sample is of the type of the last converter applied to the list,
11648which defaults to the type of the sample fetch method.
11649
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011650Each sample or converter returns data of a specific type, specified with its
11651keyword in this documentation. When an ACL is declared using a standard sample
11652fetch method, certain types automatically involved a default matching method
11653which are summarized in the table below :
11654
11655 +---------------------+-----------------+
11656 | Sample or converter | Default |
11657 | output type | matching method |
11658 +---------------------+-----------------+
11659 | boolean | bool |
11660 +---------------------+-----------------+
11661 | integer | int |
11662 +---------------------+-----------------+
11663 | ip | ip |
11664 +---------------------+-----------------+
11665 | string | str |
11666 +---------------------+-----------------+
11667 | binary | none, use "-m" |
11668 +---------------------+-----------------+
11669
11670Note that in order to match a binary samples, it is mandatory to specify a
11671matching method, see below.
11672
Willy Tarreau74ca5042013-06-11 23:12:07 +020011673The ACL engine can match these types against patterns of the following types :
11674 - boolean
11675 - integer or integer range
11676 - IP address / network
11677 - string (exact, substring, suffix, prefix, subdir, domain)
11678 - regular expression
11679 - hex block
11680
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011681The following ACL flags are currently supported :
11682
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011683 -i : ignore case during matching of all subsequent patterns.
11684 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011685 -m : use a specific pattern matching method
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010011686 -n : forbid the DNS resolutions
Thierry FOURNIER9860c412014-01-29 14:23:29 +010011687 -M : load the file pointed by -f like a map file.
Thierry FOURNIER3534d882014-01-20 17:01:44 +010011688 -u : force the unique id of the ACL
Willy Tarreau6a06a402007-07-15 20:15:28 +020011689 -- : force end of flags. Useful when a string looks like one of the flags.
11690
Willy Tarreau74ca5042013-06-11 23:12:07 +020011691The "-f" flag is followed by the name of a file from which all lines will be
11692read as individual values. It is even possible to pass multiple "-f" arguments
11693if the patterns are to be loaded from multiple files. Empty lines as well as
11694lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
11695will be stripped. If it is absolutely necessary to insert a valid pattern
11696beginning with a sharp, just prefix it with a space so that it is not taken for
11697a comment. Depending on the data type and match method, haproxy may load the
11698lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
11699exact string matching. In this case, duplicates will automatically be removed.
11700
Thierry FOURNIER9860c412014-01-29 14:23:29 +010011701The "-M" flag allows an ACL to use a map file. If this flag is set, the file is
11702parsed as two column file. The first column contains the patterns used by the
11703ACL, and the second column contain the samples. The sample can be used later by
11704a map. This can be useful in some rare cases where an ACL would just be used to
11705check for the existence of a pattern in a map before a mapping is applied.
11706
Thierry FOURNIER3534d882014-01-20 17:01:44 +010011707The "-u" flag forces the unique id of the ACL. This unique id is used with the
11708socket interface to identify ACL and dynamically change its values. Note that a
11709file is always identified by its name even if an id is set.
11710
Willy Tarreau74ca5042013-06-11 23:12:07 +020011711Also, note that the "-i" flag applies to subsequent entries and not to entries
11712loaded from files preceding it. For instance :
11713
11714 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
11715
11716In this example, each line of "exact-ua.lst" will be exactly matched against
11717the "user-agent" header of the request. Then each line of "generic-ua" will be
11718case-insensitively matched. Then the word "test" will be insensitively matched
11719as well.
11720
11721The "-m" flag is used to select a specific pattern matching method on the input
11722sample. All ACL-specific criteria imply a pattern matching method and generally
11723do not need this flag. However, this flag is useful with generic sample fetch
11724methods to describe how they're going to be matched against the patterns. This
11725is required for sample fetches which return data type for which there is no
11726obvious matching method (eg: string or binary). When "-m" is specified and
11727followed by a pattern matching method name, this method is used instead of the
11728default one for the criterion. This makes it possible to match contents in ways
11729that were not initially planned, or with sample fetch methods which return a
11730string. The matching method also affects the way the patterns are parsed.
11731
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010011732The "-n" flag forbids the dns resolutions. It is used with the load of ip files.
11733By default, if the parser cannot parse ip address it considers that the parsed
11734string is maybe a domain name and try dns resolution. The flag "-n" disable this
11735resolution. It is useful for detecting malformed ip lists. Note that if the DNS
11736server is not reachable, the haproxy configuration parsing may last many minutes
11737waiting fir the timeout. During this time no error messages are displayed. The
11738flag "-n" disable this behavior. Note also that during the runtime, this
11739function is disabled for the dynamic acl modifications.
11740
Willy Tarreau74ca5042013-06-11 23:12:07 +020011741There are some restrictions however. Not all methods can be used with all
11742sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
11743be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +020011744
11745 - "found" : only check if the requested sample could be found in the stream,
11746 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +020011747 to pass any pattern to avoid confusion. This matching method is
11748 particularly useful to detect presence of certain contents such
11749 as headers, cookies, etc... even if they are empty and without
11750 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011751
11752 - "bool" : check the value as a boolean. It can only be applied to fetches
11753 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011754 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011755
11756 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +020011757 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011758
11759 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +020011760 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011761
11762 - "bin" : match the contents against an hexadecimal string representing a
11763 binary sequence. This may be used with binary or string samples.
11764
11765 - "len" : match the sample's length as an integer. This may be used with
11766 binary or string samples.
11767
Willy Tarreau74ca5042013-06-11 23:12:07 +020011768 - "str" : exact match : match the contents against a string. This may be
11769 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011770
Willy Tarreau74ca5042013-06-11 23:12:07 +020011771 - "sub" : substring match : check that the contents contain at least one of
11772 the provided string patterns. This may be used with binary or
11773 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011774
Willy Tarreau74ca5042013-06-11 23:12:07 +020011775 - "reg" : regex match : match the contents against a list of regular
11776 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011777
Willy Tarreau74ca5042013-06-11 23:12:07 +020011778 - "beg" : prefix match : check that the contents begin like the provided
11779 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011780
Willy Tarreau74ca5042013-06-11 23:12:07 +020011781 - "end" : suffix match : check that the contents end like the provided
11782 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011783
Willy Tarreau74ca5042013-06-11 23:12:07 +020011784 - "dir" : subdir match : check that a slash-delimited portion of the
11785 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011786 This may be used with binary or string samples.
11787
Willy Tarreau74ca5042013-06-11 23:12:07 +020011788 - "dom" : domain match : check that a dot-delimited portion of the contents
11789 exactly match one of the provided string patterns. This may be
11790 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011791
11792For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
11793request, it is possible to do :
11794
11795 acl jsess_present cook(JSESSIONID) -m found
11796
11797In order to apply a regular expression on the 500 first bytes of data in the
11798buffer, one would use the following acl :
11799
11800 acl script_tag payload(0,500) -m reg -i <script>
11801
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011802On systems where the regex library is much slower when using "-i", it is
11803possible to convert the sample to lowercase before matching, like this :
11804
11805 acl script_tag payload(0,500),lower -m reg <script>
11806
Willy Tarreau74ca5042013-06-11 23:12:07 +020011807All ACL-specific criteria imply a default matching method. Most often, these
11808criteria are composed by concatenating the name of the original sample fetch
11809method and the matching method. For example, "hdr_beg" applies the "beg" match
11810to samples retrieved using the "hdr" fetch method. Since all ACL-specific
11811criteria rely on a sample fetch method, it is always possible instead to use
11812the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011813
Willy Tarreau74ca5042013-06-11 23:12:07 +020011814If an alternate match is specified using "-m" on an ACL-specific criterion,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011815the matching method is simply applied to the underlying sample fetch method.
11816For example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011817
Willy Tarreau74ca5042013-06-11 23:12:07 +020011818 acl short_form hdr_beg(host) www.
11819 acl alternate1 hdr_beg(host) -m beg www.
11820 acl alternate2 hdr_dom(host) -m beg www.
11821 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011822
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011823
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011824The table below summarizes the compatibility matrix between sample or converter
11825types and the pattern types to fetch against. It indicates for each compatible
11826combination the name of the matching method to be used, surrounded with angle
11827brackets ">" and "<" when the method is the default one and will work by
11828default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +010011829
Willy Tarreau74ca5042013-06-11 23:12:07 +020011830 +-------------------------------------------------+
11831 | Input sample type |
11832 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011833 | pattern type | boolean | integer | ip | string | binary |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011834 +----------------------+---------+---------+---------+---------+---------+
11835 | none (presence only) | found | found | found | found | found |
11836 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011837 | none (boolean value) |> bool <| bool | | bool | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011838 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011839 | integer (value) | int |> int <| int | int | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011840 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011841 | integer (length) | len | len | len | len | len |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011842 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011843 | IP address | | |> ip <| ip | ip |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011844 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011845 | exact string | str | str | str |> str <| str |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011846 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011847 | prefix | beg | beg | beg | beg | beg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011848 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011849 | suffix | end | end | end | end | end |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011850 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011851 | substring | sub | sub | sub | sub | sub |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011852 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011853 | subdir | dir | dir | dir | dir | dir |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011854 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011855 | domain | dom | dom | dom | dom | dom |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011856 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011857 | regex | reg | reg | reg | reg | reg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011858 +----------------------+---------+---------+---------+---------+---------+
11859 | hex block | | | | bin | bin |
11860 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +020011861
11862
Willy Tarreau74ca5042013-06-11 23:12:07 +0200118637.1.1. Matching booleans
11864------------------------
11865
11866In order to match a boolean, no value is needed and all values are ignored.
11867Boolean matching is used by default for all fetch methods of type "boolean".
11868When boolean matching is used, the fetched value is returned as-is, which means
11869that a boolean "true" will always match and a boolean "false" will never match.
11870
11871Boolean matching may also be enforced using "-m bool" on fetch methods which
11872return an integer value. Then, integer value 0 is converted to the boolean
11873"false" and all other values are converted to "true".
11874
Willy Tarreau6a06a402007-07-15 20:15:28 +020011875
Willy Tarreau74ca5042013-06-11 23:12:07 +0200118767.1.2. Matching integers
11877------------------------
11878
11879Integer matching applies by default to integer fetch methods. It can also be
11880enforced on boolean fetches using "-m int". In this case, "false" is converted
11881to the integer 0, and "true" is converted to the integer 1.
11882
11883Integer matching also supports integer ranges and operators. Note that integer
11884matching only applies to positive values. A range is a value expressed with a
11885lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011886
11887For instance, "1024:65535" is a valid range to represent a range of
11888unprivileged ports, and "1024:" would also work. "0:1023" is a valid
11889representation of privileged ports, and ":1023" would also work.
11890
Willy Tarreau62644772008-07-16 18:36:06 +020011891As a special case, some ACL functions support decimal numbers which are in fact
11892two integers separated by a dot. This is used with some version checks for
11893instance. All integer properties apply to those decimal numbers, including
11894ranges and operators.
11895
Willy Tarreau6a06a402007-07-15 20:15:28 +020011896For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +010011897operators with ranges does not make much sense and is strongly discouraged.
11898Similarly, it does not make much sense to perform order comparisons with a set
11899of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011900
Willy Tarreau0ba27502007-12-24 16:55:16 +010011901Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +020011902
11903 eq : true if the tested value equals at least one value
11904 ge : true if the tested value is greater than or equal to at least one value
11905 gt : true if the tested value is greater than at least one value
11906 le : true if the tested value is less than or equal to at least one value
11907 lt : true if the tested value is less than at least one value
11908
Willy Tarreau0ba27502007-12-24 16:55:16 +010011909For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +020011910
11911 acl negative-length hdr_val(content-length) lt 0
11912
Willy Tarreau62644772008-07-16 18:36:06 +020011913This one matches SSL versions between 3.0 and 3.1 (inclusive) :
11914
11915 acl sslv3 req_ssl_ver 3:3.1
11916
Willy Tarreau6a06a402007-07-15 20:15:28 +020011917
Willy Tarreau74ca5042013-06-11 23:12:07 +0200119187.1.3. Matching strings
11919-----------------------
11920
11921String matching applies to string or binary fetch methods, and exists in 6
11922different forms :
11923
11924 - exact match (-m str) : the extracted string must exactly match the
11925 patterns ;
11926
11927 - substring match (-m sub) : the patterns are looked up inside the
11928 extracted string, and the ACL matches if any of them is found inside ;
11929
11930 - prefix match (-m beg) : the patterns are compared with the beginning of
11931 the extracted string, and the ACL matches if any of them matches.
11932
11933 - suffix match (-m end) : the patterns are compared with the end of the
11934 extracted string, and the ACL matches if any of them matches.
11935
Baptiste Assmann33db6002016-03-06 23:32:10 +010011936 - subdir match (-m dir) : the patterns are looked up inside the extracted
Willy Tarreau74ca5042013-06-11 23:12:07 +020011937 string, delimited with slashes ("/"), and the ACL matches if any of them
11938 matches.
11939
11940 - domain match (-m dom) : the patterns are looked up inside the extracted
11941 string, delimited with dots ("."), and the ACL matches if any of them
11942 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011943
11944String matching applies to verbatim strings as they are passed, with the
11945exception of the backslash ("\") which makes it possible to escape some
11946characters such as the space. If the "-i" flag is passed before the first
11947string, then the matching will be performed ignoring the case. In order
11948to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +010011949before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020011950
11951
Willy Tarreau74ca5042013-06-11 23:12:07 +0200119527.1.4. Matching regular expressions (regexes)
11953---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020011954
11955Just like with string matching, regex matching applies to verbatim strings as
11956they are passed, with the exception of the backslash ("\") which makes it
11957possible to escape some characters such as the space. If the "-i" flag is
11958passed before the first regex, then the matching will be performed ignoring
11959the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +010011960the "--" flag before the first string. Same principle applies of course to
11961match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020011962
11963
Willy Tarreau74ca5042013-06-11 23:12:07 +0200119647.1.5. Matching arbitrary data blocks
11965-------------------------------------
11966
11967It is possible to match some extracted samples against a binary block which may
11968not safely be represented as a string. For this, the patterns must be passed as
11969a series of hexadecimal digits in an even number, when the match method is set
11970to binary. Each sequence of two digits will represent a byte. The hexadecimal
11971digits may be used upper or lower case.
11972
11973Example :
11974 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
11975 acl hello payload(0,6) -m bin 48656c6c6f0a
11976
11977
119787.1.6. Matching IPv4 and IPv6 addresses
11979---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020011980
11981IPv4 addresses values can be specified either as plain addresses or with a
11982netmask appended, in which case the IPv4 address matches whenever it is
11983within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010011984host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +010011985difficult to read and debug configurations. If hostnames are used, you should
11986at least ensure that they are present in /etc/hosts so that the configuration
11987does not depend on any random DNS match at the moment the configuration is
11988parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011989
Daniel Schnellereba56342016-04-13 00:26:52 +020011990The dotted IPv4 address notation is supported in both regular as well as the
11991abbreviated form with all-0-octets omitted:
11992
11993 +------------------+------------------+------------------+
11994 | Example 1 | Example 2 | Example 3 |
11995 +------------------+------------------+------------------+
11996 | 192.168.0.1 | 10.0.0.12 | 127.0.0.1 |
11997 | 192.168.1 | 10.12 | 127.1 |
11998 | 192.168.0.1/22 | 10.0.0.12/8 | 127.0.0.1/8 |
11999 | 192.168.1/22 | 10.12/8 | 127.1/8 |
12000 +------------------+------------------+------------------+
12001
12002Notice that this is different from RFC 4632 CIDR address notation in which
12003192.168.42/24 would be equivalent to 192.168.42.0/24.
12004
Willy Tarreauceb4ac92012-04-28 00:41:46 +020012005IPv6 may be entered in their usual form, with or without a netmask appended.
12006Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
12007trouble with randomly resolved IP addresses, host names are never allowed in
12008IPv6 patterns.
12009
12010HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
12011following situations :
12012 - tested address is IPv4, pattern address is IPv4, the match applies
12013 in IPv4 using the supplied mask if any.
12014 - tested address is IPv6, pattern address is IPv6, the match applies
12015 in IPv6 using the supplied mask if any.
12016 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
12017 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
12018 ::IPV4 or ::ffff:IPV4, otherwise it fails.
12019 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
12020 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
12021 applied in IPv6 using the supplied IPv6 mask.
12022
Willy Tarreau74ca5042013-06-11 23:12:07 +020012023
120247.2. Using ACLs to form conditions
12025----------------------------------
12026
12027Some actions are only performed upon a valid condition. A condition is a
12028combination of ACLs with operators. 3 operators are supported :
12029
12030 - AND (implicit)
12031 - OR (explicit with the "or" keyword or the "||" operator)
12032 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +020012033
Willy Tarreau74ca5042013-06-11 23:12:07 +020012034A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +020012035
Willy Tarreau74ca5042013-06-11 23:12:07 +020012036 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +020012037
Willy Tarreau74ca5042013-06-11 23:12:07 +020012038Such conditions are generally used after an "if" or "unless" statement,
12039indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +020012040
Willy Tarreau74ca5042013-06-11 23:12:07 +020012041For instance, to block HTTP requests to the "*" URL with methods other than
12042"OPTIONS", as well as POST requests without content-length, and GET or HEAD
12043requests with a content-length greater than 0, and finally every request which
12044is not either GET/HEAD/POST/OPTIONS !
12045
12046 acl missing_cl hdr_cnt(Content-length) eq 0
12047 block if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
12048 block if METH_GET HTTP_CONTENT
12049 block unless METH_GET or METH_POST or METH_OPTIONS
12050
12051To select a different backend for requests to static contents on the "www" site
12052and to every request on the "img", "video", "download" and "ftp" hosts :
12053
12054 acl url_static path_beg /static /images /img /css
12055 acl url_static path_end .gif .png .jpg .css .js
12056 acl host_www hdr_beg(host) -i www
12057 acl host_static hdr_beg(host) -i img. video. download. ftp.
12058
12059 # now use backend "static" for all static-only hosts, and for static urls
12060 # of host "www". Use backend "www" for the rest.
12061 use_backend static if host_static or host_www url_static
12062 use_backend www if host_www
12063
12064It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
12065expressions that are built on the fly without needing to be declared. They must
12066be enclosed between braces, with a space before and after each brace (because
12067the braces must be seen as independent words). Example :
12068
12069 The following rule :
12070
12071 acl missing_cl hdr_cnt(Content-length) eq 0
12072 block if METH_POST missing_cl
12073
12074 Can also be written that way :
12075
12076 block if METH_POST { hdr_cnt(Content-length) eq 0 }
12077
12078It is generally not recommended to use this construct because it's a lot easier
12079to leave errors in the configuration when written that way. However, for very
12080simple rules matching only one source IP address for instance, it can make more
12081sense to use them than to declare ACLs with random names. Another example of
12082good use is the following :
12083
12084 With named ACLs :
12085
12086 acl site_dead nbsrv(dynamic) lt 2
12087 acl site_dead nbsrv(static) lt 2
12088 monitor fail if site_dead
12089
12090 With anonymous ACLs :
12091
12092 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
12093
12094See section 4.2 for detailed help on the "block" and "use_backend" keywords.
12095
12096
120977.3. Fetching samples
12098---------------------
12099
12100Historically, sample fetch methods were only used to retrieve data to match
12101against patterns using ACLs. With the arrival of stick-tables, a new class of
12102sample fetch methods was created, most often sharing the same syntax as their
12103ACL counterpart. These sample fetch methods are also known as "fetches". As
12104of now, ACLs and fetches have converged. All ACL fetch methods have been made
12105available as fetch methods, and ACLs may use any sample fetch method as well.
12106
12107This section details all available sample fetch methods and their output type.
12108Some sample fetch methods have deprecated aliases that are used to maintain
12109compatibility with existing configurations. They are then explicitly marked as
12110deprecated and should not be used in new setups.
12111
12112The ACL derivatives are also indicated when available, with their respective
12113matching methods. These ones all have a well defined default pattern matching
12114method, so it is never necessary (though allowed) to pass the "-m" option to
12115indicate how the sample will be matched using ACLs.
12116
12117As indicated in the sample type versus matching compatibility matrix above,
12118when using a generic sample fetch method in an ACL, the "-m" option is
12119mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
12120the same keyword exists as an ACL keyword and as a standard fetch method, the
12121ACL engine will automatically pick the ACL-only one by default.
12122
12123Some of these keywords support one or multiple mandatory arguments, and one or
12124multiple optional arguments. These arguments are strongly typed and are checked
12125when the configuration is parsed so that there is no risk of running with an
12126incorrect argument (eg: an unresolved backend name). Fetch function arguments
12127are passed between parenthesis and are delimited by commas. When an argument
12128is optional, it will be indicated below between square brackets ('[ ]'). When
12129all arguments are optional, the parenthesis may be omitted.
12130
12131Thus, the syntax of a standard sample fetch method is one of the following :
12132 - name
12133 - name(arg1)
12134 - name(arg1,arg2)
12135
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012136
121377.3.1. Converters
12138-----------------
12139
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012140Sample fetch methods may be combined with transformations to be applied on top
12141of the fetched sample (also called "converters"). These combinations form what
12142is called "sample expressions" and the result is a "sample". Initially this
12143was only supported by "stick on" and "stick store-request" directives but this
12144has now be extended to all places where samples may be used (acls, log-format,
12145unique-id-format, add-header, ...).
12146
12147These transformations are enumerated as a series of specific keywords after the
12148sample fetch method. These keywords may equally be appended immediately after
12149the fetch keyword's argument, delimited by a comma. These keywords can also
12150support some arguments (eg: a netmask) which must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +010012151
Willy Tarreau97707872015-01-27 15:12:13 +010012152A certain category of converters are bitwise and arithmetic operators which
12153support performing basic operations on integers. Some bitwise operations are
12154supported (and, or, xor, cpl) and some arithmetic operations are supported
12155(add, sub, mul, div, mod, neg). Some comparators are provided (odd, even, not,
12156bool) which make it possible to report a match without having to write an ACL.
12157
Willy Tarreau74ca5042013-06-11 23:12:07 +020012158The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +010012159
Willy Tarreau97707872015-01-27 15:12:13 +010012160add(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012161 Adds <value> to the input value of type signed integer, and returns the
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012162 result as a signed integer. <value> can be a numeric value or a variable
Daniel Schneller0b547052016-03-21 20:46:57 +010012163 name. The name of the variable starts with an indication about its scope. The
12164 scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010012165 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010012166 "sess" : the variable is shared with the whole session
12167 "txn" : the variable is shared with the transaction (request and response)
12168 "req" : the variable is shared only during request processing
12169 "res" : the variable is shared only during response processing
12170 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010012171 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012172
12173and(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012174 Performs a bitwise "AND" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012175 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010012176 numeric value or a variable name. The name of the variable starts with an
12177 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010012178 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010012179 "sess" : the variable is shared with the whole session
12180 "txn" : the variable is shared with the transaction (request and response)
12181 "req" : the variable is shared only during request processing
12182 "res" : the variable is shared only during response processing
12183 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010012184 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012185
Emeric Brun53d1a982014-04-30 18:21:37 +020012186base64
12187 Converts a binary input sample to a base64 string. It is used to log or
12188 transfer binary content in a way that can be reliably transferred (eg:
12189 an SSL ID can be copied in a header).
12190
Willy Tarreau97707872015-01-27 15:12:13 +010012191bool
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012192 Returns a boolean TRUE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010012193 non-null, otherwise returns FALSE. Used in conjunction with and(), it can be
12194 used to report true/false for bit testing on input values (eg: verify the
12195 presence of a flag).
12196
Emeric Brun54c4ac82014-11-03 15:32:43 +010012197bytes(<offset>[,<length>])
12198 Extracts some bytes from an input binary sample. The result is a binary
12199 sample starting at an offset (in bytes) of the original sample and
12200 optionnaly truncated at the given length.
12201
Willy Tarreau97707872015-01-27 15:12:13 +010012202cpl
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012203 Takes the input value of type signed integer, applies a ones-complement
12204 (flips all bits) and returns the result as an signed integer.
Willy Tarreau97707872015-01-27 15:12:13 +010012205
Willy Tarreau80599772015-01-20 19:35:24 +010012206crc32([<avalanche>])
12207 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32
12208 hash function. Optionally, it is possible to apply a full avalanche hash
12209 function to the output if the optional <avalanche> argument equals 1. This
12210 converter uses the same functions as used by the various hash-based load
12211 balancing algorithms, so it will provide exactly the same results. It is
12212 provided for compatibility with other software which want a CRC32 to be
12213 computed on some input keys, so it follows the most common implementation as
12214 found in Ethernet, Gzip, PNG, etc... It is slower than the other algorithms
12215 but may provide a better or at least less predictable distribution. It must
12216 not be used for security purposes as a 32-bit hash is trivial to break. See
12217 also "djb2", "sdbm", "wt6" and the "hash-type" directive.
12218
David Carlier29b3ca32015-09-25 14:09:21 +010012219da-csv-conv(<prop>[,<prop>*])
David Carlier4542b102015-06-01 13:54:29 +020012220 Asks the DeviceAtlas converter to identify the User Agent string passed on
12221 input, and to emit a string made of the concatenation of the properties
12222 enumerated in argument, delimited by the separator defined by the global
12223 keyword "deviceatlas-property-separator", or by default the pipe character
David Carlier840b0242016-03-16 10:09:55 +000012224 ('|'). There's a limit of 12 different properties imposed by the haproxy
David Carlier4542b102015-06-01 13:54:29 +020012225 configuration language.
12226
12227 Example:
12228 frontend www
Cyril Bonté307ee1e2015-09-28 23:16:06 +020012229 bind *:8881
12230 default_backend servers
David Carlier840b0242016-03-16 10:09:55 +000012231 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 +020012232
Thierry FOURNIER9687c772015-05-07 15:46:29 +020012233debug
12234 This converter is used as debug tool. It dumps on screen the content and the
12235 type of the input sample. The sample is returned as is on its output. This
12236 converter only exists when haproxy was built with debugging enabled.
12237
Willy Tarreau97707872015-01-27 15:12:13 +010012238div(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012239 Divides the input value of type signed integer by <value>, and returns the
12240 result as an signed integer. If <value> is null, the largest unsigned
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012241 integer is returned (typically 2^63-1). <value> can be a numeric value or a
Daniel Schneller0b547052016-03-21 20:46:57 +010012242 variable name. The name of the variable starts with an indication about its
12243 scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010012244 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010012245 "sess" : the variable is shared with the whole session
12246 "txn" : the variable is shared with the transaction (request and response)
12247 "req" : the variable is shared only during request processing
12248 "res" : the variable is shared only during response processing
12249 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010012250 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012251
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012252djb2([<avalanche>])
12253 Hashes a binary input sample into an unsigned 32-bit quantity using the DJB2
12254 hash function. Optionally, it is possible to apply a full avalanche hash
12255 function to the output if the optional <avalanche> argument equals 1. This
12256 converter uses the same functions as used by the various hash-based load
12257 balancing algorithms, so it will provide exactly the same results. It is
12258 mostly intended for debugging, but can be used as a stick-table entry to
12259 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010012260 32-bit hash is trivial to break. See also "crc32", "sdbm", "wt6" and the
12261 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012262
Willy Tarreau97707872015-01-27 15:12:13 +010012263even
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012264 Returns a boolean TRUE if the input value of type signed integer is even
Willy Tarreau97707872015-01-27 15:12:13 +010012265 otherwise returns FALSE. It is functionally equivalent to "not,and(1),bool".
12266
Emeric Brunf399b0d2014-11-03 17:07:03 +010012267field(<index>,<delimiters>)
12268 Extracts the substring at the given index considering given delimiters from
12269 an input string. Indexes start at 1 and delimiters are a string formatted
12270 list of chars.
12271
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012272hex
12273 Converts a binary input sample to an hex string containing two hex digits per
12274 input byte. It is used to log or transfer hex dumps of some binary input data
12275 in a way that can be reliably transferred (eg: an SSL ID can be copied in a
12276 header).
Thierry FOURNIER2f49d6d2014-03-12 15:01:52 +010012277
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012278http_date([<offset>])
12279 Converts an integer supposed to contain a date since epoch to a string
12280 representing this date in a format suitable for use in HTTP header fields. If
12281 an offset value is specified, then it is a number of seconds that is added to
12282 the date before the conversion is operated. This is particularly useful to
12283 emit Date header fields, Expires values in responses when combined with a
12284 positive offset, or Last-Modified values when the offset is negative.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012285
Willy Tarreaud9f316a2014-07-10 14:03:38 +020012286in_table(<table>)
12287 Uses the string representation of the input sample to perform a look up in
12288 the specified table. If the key is not found in the table, a boolean false
12289 is returned. Otherwise a boolean true is returned. This can be used to verify
12290 the presence of a certain key in a table tracking some elements (eg: whether
12291 or not a source IP address or an Authorization header was already seen).
12292
Willy Tarreauffcb2e42014-07-10 16:29:08 +020012293ipmask(<mask>)
12294 Apply a mask to an IPv4 address, and use the result for lookups and storage.
12295 This can be used to make all hosts within a certain mask to share the same
12296 table entries and as such use the same server. The mask can be passed in
12297 dotted form (eg: 255.255.255.0) or in CIDR form (eg: 24).
12298
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020012299json([<input-code>])
12300 Escapes the input string and produces an ASCII ouput string ready to use as a
12301 JSON string. The converter tries to decode the input string according to the
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020012302 <input-code> parameter. It can be "ascii", "utf8", "utf8s", "utf8p" or
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020012303 "utf8ps". The "ascii" decoder never fails. The "utf8" decoder detects 3 types
12304 of errors:
12305 - bad UTF-8 sequence (lone continuation byte, bad number of continuation
12306 bytes, ...)
12307 - invalid range (the decoded value is within a UTF-8 prohibited range),
12308 - code overlong (the value is encoded with more bytes than necessary).
12309
12310 The UTF-8 JSON encoding can produce a "too long value" error when the UTF-8
12311 character is greater than 0xffff because the JSON string escape specification
12312 only authorizes 4 hex digits for the value encoding. The UTF-8 decoder exists
12313 in 4 variants designated by a combination of two suffix letters : "p" for
12314 "permissive" and "s" for "silently ignore". The behaviors of the decoders
12315 are :
12316 - "ascii" : never fails ;
12317 - "utf8" : fails on any detected errors ;
12318 - "utf8s" : never fails, but removes characters corresponding to errors ;
12319 - "utf8p" : accepts and fixes the overlong errors, but fails on any other
12320 error ;
12321 - "utf8ps" : never fails, accepts and fixes the overlong errors, but removes
12322 characters corresponding to the other errors.
12323
12324 This converter is particularly useful for building properly escaped JSON for
12325 logging to servers which consume JSON-formated traffic logs.
12326
12327 Example:
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020012328 capture request header Host len 15
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020012329 capture request header user-agent len 150
12330 log-format '{"ip":"%[src]","user-agent":"%[capture.req.hdr(1),json(utf8s)]"}'
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020012331
12332 Input request from client 127.0.0.1:
12333 GET / HTTP/1.0
12334 User-Agent: Very "Ugly" UA 1/2
12335
12336 Output log:
12337 {"ip":"127.0.0.1","user-agent":"Very \"Ugly\" UA 1\/2"}
12338
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012339language(<value>[,<default>])
12340 Returns the value with the highest q-factor from a list as extracted from the
12341 "accept-language" header using "req.fhdr". Values with no q-factor have a
12342 q-factor of 1. Values with a q-factor of 0 are dropped. Only values which
12343 belong to the list of semi-colon delimited <values> will be considered. The
12344 argument <value> syntax is "lang[;lang[;lang[;...]]]". If no value matches the
12345 given list and a default value is provided, it is returned. Note that language
12346 names may have a variant after a dash ('-'). If this variant is present in the
12347 list, it will be matched, but if it is not, only the base language is checked.
12348 The match is case-sensitive, and the output string is always one of those
12349 provided in arguments. The ordering of arguments is meaningless, only the
12350 ordering of the values in the request counts, as the first value among
12351 multiple sharing the same q-factor is used.
Thierry FOURNIERad903512014-04-11 17:51:01 +020012352
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012353 Example :
Thierry FOURNIERad903512014-04-11 17:51:01 +020012354
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012355 # this configuration switches to the backend matching a
12356 # given language based on the request :
Thierry FOURNIERad903512014-04-11 17:51:01 +020012357
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012358 acl es req.fhdr(accept-language),language(es;fr;en) -m str es
12359 acl fr req.fhdr(accept-language),language(es;fr;en) -m str fr
12360 acl en req.fhdr(accept-language),language(es;fr;en) -m str en
12361 use_backend spanish if es
12362 use_backend french if fr
12363 use_backend english if en
12364 default_backend choose_your_language
Thierry FOURNIERad903512014-04-11 17:51:01 +020012365
Willy Tarreauffcb2e42014-07-10 16:29:08 +020012366lower
12367 Convert a string sample to lower case. This can only be placed after a string
12368 sample fetch function or after a transformation keyword returning a string
12369 type. The result is of type string.
12370
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020012371ltime(<format>[,<offset>])
12372 Converts an integer supposed to contain a date since epoch to a string
12373 representing this date in local time using a format defined by the <format>
12374 string using strftime(3). The purpose is to allow any date format to be used
12375 in logs. An optional <offset> in seconds may be applied to the input date
12376 (positive or negative). See the strftime() man page for the format supported
12377 by your operating system. See also the utime converter.
12378
12379 Example :
12380
12381 # Emit two colons, one with the local time and another with ip:port
12382 # Eg: 20140710162350 127.0.0.1:57325
12383 log-format %[date,ltime(%Y%m%d%H%M%S)]\ %ci:%cp
12384
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012385map(<map_file>[,<default_value>])
12386map_<match_type>(<map_file>[,<default_value>])
12387map_<match_type>_<output_type>(<map_file>[,<default_value>])
12388 Search the input value from <map_file> using the <match_type> matching method,
12389 and return the associated value converted to the type <output_type>. If the
12390 input value cannot be found in the <map_file>, the converter returns the
12391 <default_value>. If the <default_value> is not set, the converter fails and
12392 acts as if no input value could be fetched. If the <match_type> is not set, it
12393 defaults to "str". Likewise, if the <output_type> is not set, it defaults to
12394 "str". For convenience, the "map" keyword is an alias for "map_str" and maps a
12395 string to another string.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012396
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012397 It is important to avoid overlapping between the keys : IP addresses and
12398 strings are stored in trees, so the first of the finest match will be used.
12399 Other keys are stored in lists, so the first matching occurrence will be used.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012400
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012401 The following array contains the list of all map functions avalaible sorted by
12402 input type, match type and output type.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012403
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012404 input type | match method | output type str | output type int | output type ip
12405 -----------+--------------+-----------------+-----------------+---------------
12406 str | str | map_str | map_str_int | map_str_ip
12407 -----------+--------------+-----------------+-----------------+---------------
Willy Tarreau787a4c02014-05-10 07:55:30 +020012408 str | beg | map_beg | map_beg_int | map_end_ip
12409 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012410 str | sub | map_sub | map_sub_int | map_sub_ip
12411 -----------+--------------+-----------------+-----------------+---------------
12412 str | dir | map_dir | map_dir_int | map_dir_ip
12413 -----------+--------------+-----------------+-----------------+---------------
12414 str | dom | map_dom | map_dom_int | map_dom_ip
12415 -----------+--------------+-----------------+-----------------+---------------
12416 str | end | map_end | map_end_int | map_end_ip
12417 -----------+--------------+-----------------+-----------------+---------------
Thierry Fournier8feaa662016-02-10 22:55:20 +010012418 | | map_reg | |
12419 str | reg +-----------------+ map_reg_int | map_reg_ip
12420 | | map_regm | |
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012421 -----------+--------------+-----------------+-----------------+---------------
12422 int | int | map_int | map_int_int | map_int_ip
12423 -----------+--------------+-----------------+-----------------+---------------
12424 ip | ip | map_ip | map_ip_int | map_ip_ip
12425 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012426
Thierry Fournier8feaa662016-02-10 22:55:20 +010012427 The special map called "map_regm" expect matching zone in the regular
12428 expression and modify the output replacing back reference (like "\1") by
12429 the corresponding match text.
12430
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012431 The file contains one key + value per line. Lines which start with '#' are
12432 ignored, just like empty lines. Leading tabs and spaces are stripped. The key
12433 is then the first "word" (series of non-space/tabs characters), and the value
12434 is what follows this series of space/tab till the end of the line excluding
12435 trailing spaces/tabs.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012436
Thierry FOURNIER060762e2014-04-23 13:29:15 +020012437 Example :
12438
12439 # this is a comment and is ignored
12440 2.22.246.0/23 United Kingdom \n
12441 <-><-----------><--><------------><---->
12442 | | | | `- trailing spaces ignored
12443 | | | `---------- value
12444 | | `-------------------- middle spaces ignored
12445 | `---------------------------- key
12446 `------------------------------------ leading spaces ignored
12447
Willy Tarreau97707872015-01-27 15:12:13 +010012448mod(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012449 Divides the input value of type signed integer by <value>, and returns the
12450 remainder as an signed integer. If <value> is null, then zero is returned.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012451 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010012452 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010012453 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010012454 "sess" : the variable is shared with the whole session
12455 "txn" : the variable is shared with the transaction (request and response)
12456 "req" : the variable is shared only during request processing
12457 "res" : the variable is shared only during response processing
12458 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010012459 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012460
12461mul(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012462 Multiplies the input value of type signed integer by <value>, and returns
Thierry FOURNIER00c005c2015-07-08 01:10:21 +020012463 the product as an signed integer. In case of overflow, the largest possible
12464 value for the sign is returned so that the operation doesn't wrap around.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012465 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010012466 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010012467 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010012468 "sess" : the variable is shared with the whole session
12469 "txn" : the variable is shared with the transaction (request and response)
12470 "req" : the variable is shared only during request processing
12471 "res" : the variable is shared only during response processing
12472 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010012473 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012474
12475neg
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012476 Takes the input value of type signed integer, computes the opposite value,
12477 and returns the remainder as an signed integer. 0 is identity. This operator
12478 is provided for reversed subtracts : in order to subtract the input from a
12479 constant, simply perform a "neg,add(value)".
Willy Tarreau97707872015-01-27 15:12:13 +010012480
12481not
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012482 Returns a boolean FALSE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010012483 non-null, otherwise returns TRUE. Used in conjunction with and(), it can be
12484 used to report true/false for bit testing on input values (eg: verify the
12485 absence of a flag).
12486
12487odd
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012488 Returns a boolean TRUE if the input value of type signed integer is odd
Willy Tarreau97707872015-01-27 15:12:13 +010012489 otherwise returns FALSE. It is functionally equivalent to "and(1),bool".
12490
12491or(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012492 Performs a bitwise "OR" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012493 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010012494 numeric value or a variable name. The name of the variable starts with an
12495 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010012496 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010012497 "sess" : the variable is shared with the whole session
12498 "txn" : the variable is shared with the transaction (request and response)
12499 "req" : the variable is shared only during request processing
12500 "res" : the variable is shared only during response processing
12501 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010012502 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012503
Willy Tarreauc4dc3502015-01-23 20:39:28 +010012504regsub(<regex>,<subst>[,<flags>])
Willy Tarreau7eda8492015-01-20 19:47:06 +010012505 Applies a regex-based substitution to the input string. It does the same
12506 operation as the well-known "sed" utility with "s/<regex>/<subst>/". By
12507 default it will replace in the input string the first occurrence of the
12508 largest part matching the regular expression <regex> with the substitution
12509 string <subst>. It is possible to replace all occurrences instead by adding
12510 the flag "g" in the third argument <flags>. It is also possible to make the
12511 regex case insensitive by adding the flag "i" in <flags>. Since <flags> is a
12512 string, it is made up from the concatenation of all desired flags. Thus if
12513 both "i" and "g" are desired, using "gi" or "ig" will have the same effect.
12514 It is important to note that due to the current limitations of the
Baptiste Assmann66025d82016-03-06 23:36:48 +010012515 configuration parser, some characters such as closing parenthesis, closing
12516 square brackets or comma are not possible to use in the arguments. The first
12517 use of this converter is to replace certain characters or sequence of
12518 characters with other ones.
Willy Tarreau7eda8492015-01-20 19:47:06 +010012519
12520 Example :
12521
12522 # de-duplicate "/" in header "x-path".
12523 # input: x-path: /////a///b/c/xzxyz/
12524 # output: x-path: /a/b/c/xzxyz/
12525 http-request set-header x-path %[hdr(x-path),regsub(/+,/,g)]
12526
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020012527capture-req(<id>)
12528 Capture the string entry in the request slot <id> and returns the entry as
12529 is. If the slot doesn't exist, the capture fails silently.
12530
12531 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020012532 "http-response capture", "capture.req.hdr" and
12533 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020012534
12535capture-res(<id>)
12536 Capture the string entry in the response slot <id> and returns the entry as
12537 is. If the slot doesn't exist, the capture fails silently.
12538
12539 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020012540 "http-response capture", "capture.req.hdr" and
12541 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020012542
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012543sdbm([<avalanche>])
12544 Hashes a binary input sample into an unsigned 32-bit quantity using the SDBM
12545 hash function. Optionally, it is possible to apply a full avalanche hash
12546 function to the output if the optional <avalanche> argument equals 1. This
12547 converter uses the same functions as used by the various hash-based load
12548 balancing algorithms, so it will provide exactly the same results. It is
12549 mostly intended for debugging, but can be used as a stick-table entry to
12550 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010012551 32-bit hash is trivial to break. See also "crc32", "djb2", "wt6" and the
12552 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012553
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012554set-var(<var name>)
Daniel Schneller0b547052016-03-21 20:46:57 +010012555 Sets a variable with the input content and returns the content on the output as
12556 is. The variable keeps the value and the associated input type. The name of the
12557 variable starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010012558 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010012559 "sess" : the variable is shared with the whole session
12560 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012561 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010012562 "req" : the variable is shared only during request processing,
12563 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012564 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010012565 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012566
Willy Tarreau97707872015-01-27 15:12:13 +010012567sub(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012568 Subtracts <value> from the input value of type signed integer, and returns
12569 the result as an signed integer. Note: in order to subtract the input from
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012570 a constant, simply perform a "neg,add(value)". <value> can be a numeric value
Daniel Schneller0b547052016-03-21 20:46:57 +010012571 or a variable name. The name of the variable starts with an indication about
12572 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010012573 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010012574 "sess" : the variable is shared with the whole session
12575 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012576 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010012577 "req" : the variable is shared only during request processing,
12578 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012579 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010012580 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012581
Willy Tarreaud9f316a2014-07-10 14:03:38 +020012582table_bytes_in_rate(<table>)
12583 Uses the string representation of the input sample to perform a look up in
12584 the specified table. If the key is not found in the table, integer value zero
12585 is returned. Otherwise the converter returns the average client-to-server
12586 bytes rate associated with the input sample in the designated table, measured
12587 in amount of bytes over the period configured in the table. See also the
12588 sc_bytes_in_rate sample fetch keyword.
12589
12590
12591table_bytes_out_rate(<table>)
12592 Uses the string representation of the input sample to perform a look up in
12593 the specified table. If the key is not found in the table, integer value zero
12594 is returned. Otherwise the converter returns the average server-to-client
12595 bytes rate associated with the input sample in the designated table, measured
12596 in amount of bytes over the period configured in the table. See also the
12597 sc_bytes_out_rate sample fetch keyword.
12598
12599table_conn_cnt(<table>)
12600 Uses the string representation of the input sample to perform a look up in
12601 the specified table. If the key is not found in the table, integer value zero
12602 is returned. Otherwise the converter returns the cumulated amount of incoming
12603 connections associated with the input sample in the designated table. See
12604 also the sc_conn_cnt sample fetch keyword.
12605
12606table_conn_cur(<table>)
12607 Uses the string representation of the input sample to perform a look up in
12608 the specified table. If the key is not found in the table, integer value zero
12609 is returned. Otherwise the converter returns the current amount of concurrent
12610 tracked connections associated with the input sample in the designated table.
12611 See also the sc_conn_cur sample fetch keyword.
12612
12613table_conn_rate(<table>)
12614 Uses the string representation of the input sample to perform a look up in
12615 the specified table. If the key is not found in the table, integer value zero
12616 is returned. Otherwise the converter returns the average incoming connection
12617 rate associated with the input sample in the designated table. See also the
12618 sc_conn_rate sample fetch keyword.
12619
Thierry FOURNIER236657b2015-08-19 08:25:14 +020012620table_gpt0(<table>)
12621 Uses the string representation of the input sample to perform a look up in
12622 the specified table. If the key is not found in the table, boolean value zero
12623 is returned. Otherwise the converter returns the current value of the first
12624 general purpose tag associated with the input sample in the designated table.
12625 See also the sc_get_gpt0 sample fetch keyword.
12626
Willy Tarreaud9f316a2014-07-10 14:03:38 +020012627table_gpc0(<table>)
12628 Uses the string representation of the input sample to perform a look up in
12629 the specified table. If the key is not found in the table, integer value zero
12630 is returned. Otherwise the converter returns the current value of the first
12631 general purpose counter associated with the input sample in the designated
12632 table. See also the sc_get_gpc0 sample fetch keyword.
12633
12634table_gpc0_rate(<table>)
12635 Uses the string representation of the input sample to perform a look up in
12636 the specified table. If the key is not found in the table, integer value zero
12637 is returned. Otherwise the converter returns the frequency which the gpc0
12638 counter was incremented over the configured period in the table, associated
12639 with the input sample in the designated table. See also the sc_get_gpc0_rate
12640 sample fetch keyword.
12641
12642table_http_err_cnt(<table>)
12643 Uses the string representation of the input sample to perform a look up in
12644 the specified table. If the key is not found in the table, integer value zero
12645 is returned. Otherwise the converter returns the cumulated amount of HTTP
12646 errors associated with the input sample in the designated table. See also the
12647 sc_http_err_cnt sample fetch keyword.
12648
12649table_http_err_rate(<table>)
12650 Uses the string representation of the input sample to perform a look up in
12651 the specified table. If the key is not found in the table, integer value zero
12652 is returned. Otherwise the average rate of HTTP errors associated with the
12653 input sample in the designated table, measured in amount of errors over the
12654 period configured in the table. See also the sc_http_err_rate sample fetch
12655 keyword.
12656
12657table_http_req_cnt(<table>)
12658 Uses the string representation of the input sample to perform a look up in
12659 the specified table. If the key is not found in the table, integer value zero
12660 is returned. Otherwise the converter returns the cumulated amount of HTTP
12661 requests associated with the input sample in the designated table. See also
12662 the sc_http_req_cnt sample fetch keyword.
12663
12664table_http_req_rate(<table>)
12665 Uses the string representation of the input sample to perform a look up in
12666 the specified table. If the key is not found in the table, integer value zero
12667 is returned. Otherwise the average rate of HTTP requests associated with the
12668 input sample in the designated table, measured in amount of requests over the
12669 period configured in the table. See also the sc_http_req_rate sample fetch
12670 keyword.
12671
12672table_kbytes_in(<table>)
12673 Uses the string representation of the input sample to perform a look up in
12674 the specified table. If the key is not found in the table, integer value zero
12675 is returned. Otherwise the converter returns the cumulated amount of client-
12676 to-server data associated with the input sample in the designated table,
12677 measured in kilobytes. The test is currently performed on 32-bit integers,
12678 which limits values to 4 terabytes. See also the sc_kbytes_in sample fetch
12679 keyword.
12680
12681table_kbytes_out(<table>)
12682 Uses the string representation of the input sample to perform a look up in
12683 the specified table. If the key is not found in the table, integer value zero
12684 is returned. Otherwise the converter returns the cumulated amount of server-
12685 to-client data associated with the input sample in the designated table,
12686 measured in kilobytes. The test is currently performed on 32-bit integers,
12687 which limits values to 4 terabytes. See also the sc_kbytes_out sample fetch
12688 keyword.
12689
12690table_server_id(<table>)
12691 Uses the string representation of the input sample to perform a look up in
12692 the specified table. If the key is not found in the table, integer value zero
12693 is returned. Otherwise the converter returns the server ID associated with
12694 the input sample in the designated table. A server ID is associated to a
12695 sample by a "stick" rule when a connection to a server succeeds. A server ID
12696 zero means that no server is associated with this key.
12697
12698table_sess_cnt(<table>)
12699 Uses the string representation of the input sample to perform a look up in
12700 the specified table. If the key is not found in the table, integer value zero
12701 is returned. Otherwise the converter returns the cumulated amount of incoming
12702 sessions associated with the input sample in the designated table. Note that
12703 a session here refers to an incoming connection being accepted by the
12704 "tcp-request connection" rulesets. See also the sc_sess_cnt sample fetch
12705 keyword.
12706
12707table_sess_rate(<table>)
12708 Uses the string representation of the input sample to perform a look up in
12709 the specified table. If the key is not found in the table, integer value zero
12710 is returned. Otherwise the converter returns the average incoming session
12711 rate associated with the input sample in the designated table. Note that a
12712 session here refers to an incoming connection being accepted by the
12713 "tcp-request connection" rulesets. See also the sc_sess_rate sample fetch
12714 keyword.
12715
12716table_trackers(<table>)
12717 Uses the string representation of the input sample to perform a look up in
12718 the specified table. If the key is not found in the table, integer value zero
12719 is returned. Otherwise the converter returns the current amount of concurrent
12720 connections tracking the same key as the input sample in the designated
12721 table. It differs from table_conn_cur in that it does not rely on any stored
12722 information but on the table's reference count (the "use" value which is
12723 returned by "show table" on the CLI). This may sometimes be more suited for
12724 layer7 tracking. It can be used to tell a server how many concurrent
12725 connections there are from a given address for example. See also the
12726 sc_trackers sample fetch keyword.
12727
Willy Tarreauffcb2e42014-07-10 16:29:08 +020012728upper
12729 Convert a string sample to upper case. This can only be placed after a string
12730 sample fetch function or after a transformation keyword returning a string
12731 type. The result is of type string.
12732
Thierry FOURNIER82ff3c92015-05-07 15:46:20 +020012733url_dec
12734 Takes an url-encoded string provided as input and returns the decoded
12735 version as output. The input and the output are of type string.
12736
Christopher Faulet85d79c92016-11-09 16:54:56 +010012737unset-var(<var name>)
12738 Unsets a variable if the input content is defined. The name of the variable
12739 starts with an indication about its scope. The scopes allowed are:
12740 "proc" : the variable is shared with the whole process
12741 "sess" : the variable is shared with the whole session
12742 "txn" : the variable is shared with the transaction (request and
12743 response),
12744 "req" : the variable is shared only during request processing,
12745 "res" : the variable is shared only during response processing.
12746 This prefix is followed by a name. The separator is a '.'. The name may only
12747 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
12748
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020012749utime(<format>[,<offset>])
12750 Converts an integer supposed to contain a date since epoch to a string
12751 representing this date in UTC time using a format defined by the <format>
12752 string using strftime(3). The purpose is to allow any date format to be used
12753 in logs. An optional <offset> in seconds may be applied to the input date
12754 (positive or negative). See the strftime() man page for the format supported
12755 by your operating system. See also the ltime converter.
12756
12757 Example :
12758
12759 # Emit two colons, one with the UTC time and another with ip:port
12760 # Eg: 20140710162350 127.0.0.1:57325
12761 log-format %[date,utime(%Y%m%d%H%M%S)]\ %ci:%cp
12762
Emeric Brunc9a0f6d2014-11-25 14:09:01 +010012763word(<index>,<delimiters>)
12764 Extracts the nth word considering given delimiters from an input string.
12765 Indexes start at 1 and delimiters are a string formatted list of chars.
12766
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012767wt6([<avalanche>])
12768 Hashes a binary input sample into an unsigned 32-bit quantity using the WT6
12769 hash function. Optionally, it is possible to apply a full avalanche hash
12770 function to the output if the optional <avalanche> argument equals 1. This
12771 converter uses the same functions as used by the various hash-based load
12772 balancing algorithms, so it will provide exactly the same results. It is
12773 mostly intended for debugging, but can be used as a stick-table entry to
12774 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010012775 32-bit hash is trivial to break. See also "crc32", "djb2", "sdbm", and the
12776 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012777
Willy Tarreau97707872015-01-27 15:12:13 +010012778xor(<value>)
12779 Performs a bitwise "XOR" (exclusive OR) between <value> and the input value
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012780 of type signed integer, and returns the result as an signed integer.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012781 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010012782 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010012783 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010012784 "sess" : the variable is shared with the whole session
12785 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012786 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010012787 "req" : the variable is shared only during request processing,
12788 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012789 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010012790 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012791
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012792
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200127937.3.2. Fetching samples from internal states
Willy Tarreau74ca5042013-06-11 23:12:07 +020012794--------------------------------------------
12795
12796A first set of sample fetch methods applies to internal information which does
12797not even relate to any client information. These ones are sometimes used with
12798"monitor-fail" directives to report an internal status to external watchers.
12799The sample fetch methods described in this section are usable anywhere.
12800
12801always_false : boolean
12802 Always returns the boolean "false" value. It may be used with ACLs as a
12803 temporary replacement for another one when adjusting configurations.
12804
12805always_true : boolean
12806 Always returns the boolean "true" value. It may be used with ACLs as a
12807 temporary replacement for another one when adjusting configurations.
12808
12809avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010012810 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +020012811 divided by the number of active servers. The current backend is used if no
12812 backend is specified. This is very similar to "queue" except that the size of
12813 the farm is considered, in order to give a more accurate measurement of the
12814 time it may take for a new connection to be processed. The main usage is with
12815 ACL to return a sorry page to new users when it becomes certain they will get
12816 a degraded service, or to pass to the backend servers in a header so that
12817 they decide to work in degraded mode or to disable some functions to speed up
12818 the processing a bit. Note that in the event there would not be any active
12819 server anymore, twice the number of queued connections would be considered as
12820 the measured value. This is a fair estimate, as we expect one server to get
12821 back soon anyway, but we still prefer to send new traffic to another backend
12822 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
12823 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +010012824
Willy Tarreau74ca5042013-06-11 23:12:07 +020012825be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +020012826 Applies to the number of currently established connections on the backend,
12827 possibly including the connection being evaluated. If no backend name is
12828 specified, the current one is used. But it is also possible to check another
12829 backend. It can be used to use a specific farm when the nominal one is full.
12830 See also the "fe_conn", "queue" and "be_sess_rate" criteria.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012831
Willy Tarreau74ca5042013-06-11 23:12:07 +020012832be_sess_rate([<backend>]) : integer
12833 Returns an integer value corresponding to the sessions creation rate on the
12834 backend, in number of new sessions per second. This is used with ACLs to
12835 switch to an alternate backend when an expensive or fragile one reaches too
12836 high a session rate, or to limit abuse of service (eg. prevent sucking of an
12837 online dictionary). It can also be useful to add this element to logs using a
12838 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012839
12840 Example :
12841 # Redirect to an error page if the dictionary is requested too often
12842 backend dynamic
12843 mode http
12844 acl being_scanned be_sess_rate gt 100
12845 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +010012846
Thierry FOURNIERcc103292015-06-06 19:30:17 +020012847bin(<hexa>) : bin
12848 Returns a binary chain. The input is the hexadecimal representation
12849 of the string.
12850
12851bool(<bool>) : bool
12852 Returns a boolean value. <bool> can be 'true', 'false', '1' or '0'.
12853 'false' and '0' are the same. 'true' and '1' are the same.
12854
Willy Tarreau74ca5042013-06-11 23:12:07 +020012855connslots([<backend>]) : integer
12856 Returns an integer value corresponding to the number of connection slots
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012857 still available in the backend, by totaling the maximum amount of
Willy Tarreau74ca5042013-06-11 23:12:07 +020012858 connections on all servers and the maximum queue size. This is probably only
12859 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -050012860
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012861 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +020012862 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012863 usage; see "use_backend" keyword) can be redirected to a different backend.
12864
Willy Tarreau55165fe2009-05-10 12:02:55 +020012865 'connslots' = number of available server connection slots, + number of
12866 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012867
Willy Tarreaua36af912009-10-10 12:02:45 +020012868 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +020012869 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +020012870 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +020012871 you want to be able to differentiate between different backends, and their
12872 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +020012873 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +020012874 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012875
Willy Tarreau55165fe2009-05-10 12:02:55 +020012876 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
12877 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +020012878 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +020012879 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012880
Willy Tarreau6236d3a2013-07-25 14:28:25 +020012881date([<offset>]) : integer
12882 Returns the current date as the epoch (number of seconds since 01/01/1970).
12883 If an offset value is specified, then it is a number of seconds that is added
12884 to the current date before returning the value. This is particularly useful
12885 to compute relative dates, as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +020012886 It is useful combined with the http_date converter.
12887
12888 Example :
12889
12890 # set an expires header to now+1 hour in every response
12891 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +020012892
Willy Tarreau595ec542013-06-12 21:34:28 +020012893env(<name>) : string
12894 Returns a string containing the value of environment variable <name>. As a
12895 reminder, environment variables are per-process and are sampled when the
12896 process starts. This can be useful to pass some information to a next hop
12897 server, or with ACLs to take specific action when the process is started a
12898 certain way.
12899
12900 Examples :
12901 # Pass the Via header to next hop with the local hostname in it
12902 http-request add-header Via 1.1\ %[env(HOSTNAME)]
12903
12904 # reject cookie-less requests when the STOP environment variable is set
12905 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
12906
Willy Tarreau74ca5042013-06-11 23:12:07 +020012907fe_conn([<frontend>]) : integer
12908 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +010012909 possibly including the connection being evaluated. If no frontend name is
12910 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020012911 frontend. It can be used to return a sorry page before hard-blocking, or to
12912 use a specific backend to drain new requests when the farm is considered
12913 full. This is mostly used with ACLs but can also be used to pass some
12914 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
12915 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +020012916
Nenad Merdanovicad9a7e92016-10-03 04:57:37 +020012917fe_req_rate([<frontend>]) : integer
12918 Returns an integer value corresponding to the number of HTTP requests per
12919 second sent to a frontend. This number can differ from "fe_sess_rate" in
12920 situations where client-side keep-alive is enabled.
12921
Willy Tarreau74ca5042013-06-11 23:12:07 +020012922fe_sess_rate([<frontend>]) : integer
12923 Returns an integer value corresponding to the sessions creation rate on the
12924 frontend, in number of new sessions per second. This is used with ACLs to
12925 limit the incoming session rate to an acceptable range in order to prevent
12926 abuse of service at the earliest moment, for example when combined with other
12927 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
12928 down below the limit. It can also be useful to add this element to logs using
12929 a log-format directive. See also the "rate-limit sessions" directive for use
12930 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010012931
12932 Example :
12933 # This frontend limits incoming mails to 10/s with a max of 100
12934 # concurrent connections. We accept any connection below 10/s, and
12935 # force excess clients to wait for 100 ms. Since clients are limited to
12936 # 100 max, there cannot be more than 10 incoming mails per second.
12937 frontend mail
12938 bind :25
12939 mode tcp
12940 maxconn 100
12941 acl too_fast fe_sess_rate ge 10
12942 tcp-request inspect-delay 100ms
12943 tcp-request content accept if ! too_fast
12944 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +010012945
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012946int(<integer>) : signed integer
12947 Returns a signed integer.
12948
Thierry FOURNIERcc103292015-06-06 19:30:17 +020012949ipv4(<ipv4>) : ipv4
12950 Returns an ipv4.
12951
12952ipv6(<ipv6>) : ipv6
12953 Returns an ipv6.
12954
12955meth(<method>) : method
12956 Returns a method.
12957
Willy Tarreau0f30d262014-11-24 16:02:05 +010012958nbproc : integer
12959 Returns an integer value corresponding to the number of processes that were
12960 started (it equals the global "nbproc" setting). This is useful for logging
12961 and debugging purposes.
12962
Willy Tarreau74ca5042013-06-11 23:12:07 +020012963nbsrv([<backend>]) : integer
12964 Returns an integer value corresponding to the number of usable servers of
12965 either the current backend or the named backend. This is mostly used with
12966 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +010012967 switch to an alternate backend when the number of servers is too low to
12968 to handle some load. It is useful to report a failure when combined with
12969 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +010012970
Willy Tarreau0f30d262014-11-24 16:02:05 +010012971proc : integer
12972 Returns an integer value corresponding to the position of the process calling
12973 the function, between 1 and global.nbproc. This is useful for logging and
12974 debugging purposes.
12975
Willy Tarreau74ca5042013-06-11 23:12:07 +020012976queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010012977 Returns the total number of queued connections of the designated backend,
12978 including all the connections in server queues. If no backend name is
12979 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020012980 one. This is useful with ACLs or to pass statistics to backend servers. This
12981 can be used to take actions when queuing goes above a known level, generally
12982 indicating a surge of traffic or a massive slowdown on the servers. One
12983 possible action could be to reject new users but still accept old ones. See
12984 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
12985
Willy Tarreau84310e22014-02-14 11:59:04 +010012986rand([<range>]) : integer
12987 Returns a random integer value within a range of <range> possible values,
12988 starting at zero. If the range is not specified, it defaults to 2^32, which
12989 gives numbers between 0 and 4294967295. It can be useful to pass some values
12990 needed to take some routing decisions for example, or just for debugging
12991 purposes. This random must not be used for security purposes.
12992
Willy Tarreau74ca5042013-06-11 23:12:07 +020012993srv_conn([<backend>/]<server>) : integer
12994 Returns an integer value corresponding to the number of currently established
12995 connections on the designated server, possibly including the connection being
12996 evaluated. If <backend> is omitted, then the server is looked up in the
12997 current backend. It can be used to use a specific farm when one server is
12998 full, or to inform the server about our view of the number of active
12999 connections with it. See also the "fe_conn", "be_conn" and "queue" fetch
13000 methods.
13001
13002srv_is_up([<backend>/]<server>) : boolean
13003 Returns true when the designated server is UP, and false when it is either
13004 DOWN or in maintenance mode. If <backend> is omitted, then the server is
13005 looked up in the current backend. It is mainly used to take action based on
13006 an external status reported via a health check (eg: a geographical site's
13007 availability). Another possible use which is more of a hack consists in
13008 using dummy servers as boolean variables that can be enabled or disabled from
13009 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
13010
13011srv_sess_rate([<backend>/]<server>) : integer
13012 Returns an integer corresponding to the sessions creation rate on the
13013 designated server, in number of new sessions per second. If <backend> is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013014 omitted, then the server is looked up in the current backend. This is mostly
Willy Tarreau74ca5042013-06-11 23:12:07 +020013015 used with ACLs but can make sense with logs too. This is used to switch to an
13016 alternate backend when an expensive or fragile one reaches too high a session
13017 rate, or to limit abuse of service (eg. prevent latent requests from
13018 overloading servers).
13019
13020 Example :
13021 # Redirect to a separate back
13022 acl srv1_full srv_sess_rate(be1/srv1) gt 50
13023 acl srv2_full srv_sess_rate(be1/srv2) gt 50
13024 use_backend be2 if srv1_full or srv2_full
13025
Willy Tarreau0f30d262014-11-24 16:02:05 +010013026stopping : boolean
13027 Returns TRUE if the process calling the function is currently stopping. This
13028 can be useful for logging, or for relaxing certain checks or helping close
13029 certain connections upon graceful shutdown.
13030
Thierry FOURNIERcc103292015-06-06 19:30:17 +020013031str(<string>) : string
13032 Returns a string.
13033
Willy Tarreau74ca5042013-06-11 23:12:07 +020013034table_avl([<table>]) : integer
13035 Returns the total number of available entries in the current proxy's
13036 stick-table or in the designated stick-table. See also table_cnt.
13037
13038table_cnt([<table>]) : integer
13039 Returns the total number of entries currently in use in the current proxy's
13040 stick-table or in the designated stick-table. See also src_conn_cnt and
13041 table_avl for other entry counting methods.
13042
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013043var(<var-name>) : undefined
13044 Returns a variable with the stored type. If the variable is not set, the
Daniel Schneller0b547052016-03-21 20:46:57 +010013045 sample fetch fails. The name of the variable starts with an indication
13046 about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013047 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013048 "sess" : the variable is shared with the whole session
13049 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013050 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010013051 "req" : the variable is shared only during request processing,
13052 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013053 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013054 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013055
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200130567.3.3. Fetching samples at Layer 4
Willy Tarreau74ca5042013-06-11 23:12:07 +020013057----------------------------------
13058
13059The layer 4 usually describes just the transport layer which in haproxy is
13060closest to the connection, where no content is yet made available. The fetch
13061methods described here are usable as low as the "tcp-request connection" rule
13062sets unless they require some future information. Those generally include
13063TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013064the incoming connection. For retrieving a value from a sticky counters, the
13065counter number can be explicitly set as 0, 1, or 2 using the pre-defined
13066"sc0_", "sc1_", or "sc2_" prefix, or it can be specified as the first integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013067argument when using the "sc_" prefix. An optional table may be specified with
13068the "sc*" form, in which case the currently tracked key will be looked up into
13069this alternate table instead of the table currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013070
13071be_id : integer
13072 Returns an integer containing the current backend's id. It can be used in
13073 frontends with responses to check which backend processed the request.
13074
13075dst : ip
13076 This is the destination IPv4 address of the connection on the client side,
13077 which is the address the client connected to. It can be useful when running
13078 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
13079 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
13080 RFC 4291.
13081
13082dst_conn : integer
13083 Returns an integer value corresponding to the number of currently established
13084 connections on the same socket including the one being evaluated. It is
13085 normally used with ACLs but can as well be used to pass the information to
13086 servers in an HTTP header or in logs. It can be used to either return a sorry
13087 page before hard-blocking, or to use a specific backend to drain new requests
13088 when the socket is considered saturated. This offers the ability to assign
13089 different limits to different listening ports or addresses. See also the
13090 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013091
Willy Tarreau16e01562016-08-09 16:46:18 +020013092dst_is_local : boolean
13093 Returns true if the destination address of the incoming connection is local
13094 to the system, or false if the address doesn't exist on the system, meaning
13095 that it was intercepted in transparent mode. It can be useful to apply
13096 certain rules by default to forwarded traffic and other rules to the traffic
13097 targetting the real address of the machine. For example the stats page could
13098 be delivered only on this address, or SSH access could be locally redirected.
13099 Please note that the check involves a few system calls, so it's better to do
13100 it only once per connection.
13101
Willy Tarreau74ca5042013-06-11 23:12:07 +020013102dst_port : integer
13103 Returns an integer value corresponding to the destination TCP port of the
13104 connection on the client side, which is the port the client connected to.
13105 This might be used when running in transparent mode, when assigning dynamic
13106 ports to some clients for a whole application session, to stick all users to
13107 a same server, or to pass the destination port information to a server using
13108 an HTTP header.
13109
Thierry Fournier / OZON.IO6310bef2016-07-24 20:16:50 +020013110fc_rtt(<unit>) : integer
13111 Returns the Round Trip Time (RTT) measured by the kernel for the client
13112 connection. <unit> is facultative, by default the unit is milliseconds. <unit>
13113 can be set to "ms" for milliseconds or "us" for microseconds. If the server
13114 connection is not established, if the connection is not TCP or if the
13115 operating system does not support TCP_INFO, for example Linux kernels before
13116 2.4, the sample fetch fails.
13117
13118fc_rttvar(<unit>) : integer
13119 Returns the Round Trip Time (RTT) variance measured by the kernel for the
13120 client connection. <unit> is facultative, by default the unit is milliseconds.
13121 <unit> can be set to "ms" for milliseconds or "us" for microseconds. If the
13122 server connection is not established, if the connection is not TCP or if the
13123 operating system does not support TCP_INFO, for example Linux kernels before
13124 2.4, the sample fetch fails.
13125
Joe Williams30fcd392016-08-10 07:06:44 -070013126fc_unacked(<unit>) : integer
13127 Returns the unacked counter measured by the kernel for the client connection.
13128 If the server connection is not established, if the connection is not TCP or
13129 if the operating system does not support TCP_INFO, for example Linux kernels
13130 before 2.4, the sample fetch fails.
13131
13132fc_sacked(<unit>) : integer
13133 Returns the sacked counter measured by the kernel for the client connection.
13134 If the server connection is not established, if the connection is not TCP or
13135 if the operating system does not support TCP_INFO, for example Linux kernels
13136 before 2.4, the sample fetch fails.
13137
13138fc_retrans(<unit>) : integer
13139 Returns the retransmits counter measured by the kernel for the client
13140 connection. If the server connection is not established, if the connection is
13141 not TCP or if the operating system does not support TCP_INFO, for example
13142 Linux kernels before 2.4, the sample fetch fails.
13143
13144fc_fackets(<unit>) : integer
13145 Returns the fack counter measured by the kernel for the client
13146 connection. If the server connection is not established, if the connection is
13147 not TCP or if the operating system does not support TCP_INFO, for example
13148 Linux kernels before 2.4, the sample fetch fails.
13149
13150fc_lost(<unit>) : integer
13151 Returns the lost counter measured by the kernel for the client
13152 connection. If the server connection is not established, if the connection is
13153 not TCP or if the operating system does not support TCP_INFO, for example
13154 Linux kernels before 2.4, the sample fetch fails.
13155
13156fc_reordering(<unit>) : integer
13157 Returns the reordering counter measured by the kernel for the client
13158 connection. If the server connection is not established, if the connection is
13159 not TCP or if the operating system does not support TCP_INFO, for example
13160 Linux kernels before 2.4, the sample fetch fails.
13161
Willy Tarreau74ca5042013-06-11 23:12:07 +020013162fe_id : integer
13163 Returns an integer containing the current frontend's id. It can be used in
13164 backends to check from which backend it was called, or to stick all users
13165 coming via a same frontend to the same server.
13166
Cyril Bonté62ba8702014-04-22 23:52:25 +020013167sc_bytes_in_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013168sc0_bytes_in_rate([<table>]) : integer
13169sc1_bytes_in_rate([<table>]) : integer
13170sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020013171 Returns the average client-to-server bytes rate from the currently tracked
13172 counters, measured in amount of bytes over the period configured in the
13173 table. See also src_bytes_in_rate.
13174
Cyril Bonté62ba8702014-04-22 23:52:25 +020013175sc_bytes_out_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013176sc0_bytes_out_rate([<table>]) : integer
13177sc1_bytes_out_rate([<table>]) : integer
13178sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020013179 Returns the average server-to-client bytes rate from the currently tracked
13180 counters, measured in amount of bytes over the period configured in the
13181 table. See also src_bytes_out_rate.
13182
Cyril Bonté62ba8702014-04-22 23:52:25 +020013183sc_clr_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013184sc0_clr_gpc0([<table>]) : integer
13185sc1_clr_gpc0([<table>]) : integer
13186sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +020013187 Clears the first General Purpose Counter associated to the currently tracked
13188 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +010013189 stored value is zero, so first invocation will always return zero. This is
13190 typically used as a second ACL in an expression in order to mark a connection
13191 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020013192
13193 # block if 5 consecutive requests continue to come faster than 10 sess
13194 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020013195 acl abuse sc0_http_req_rate gt 10
13196 acl kill sc0_inc_gpc0 gt 5
13197 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020013198 tcp-request connection accept if !abuse save
13199 tcp-request connection reject if abuse kill
13200
Cyril Bonté62ba8702014-04-22 23:52:25 +020013201sc_conn_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013202sc0_conn_cnt([<table>]) : integer
13203sc1_conn_cnt([<table>]) : integer
13204sc2_conn_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020013205 Returns the cumulated number of incoming connections from currently tracked
13206 counters. See also src_conn_cnt.
13207
Cyril Bonté62ba8702014-04-22 23:52:25 +020013208sc_conn_cur(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013209sc0_conn_cur([<table>]) : integer
13210sc1_conn_cur([<table>]) : integer
13211sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020013212 Returns the current amount of concurrent connections tracking the same
13213 tracked counters. This number is automatically incremented when tracking
13214 begins and decremented when tracking stops. See also src_conn_cur.
13215
Cyril Bonté62ba8702014-04-22 23:52:25 +020013216sc_conn_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013217sc0_conn_rate([<table>]) : integer
13218sc1_conn_rate([<table>]) : integer
13219sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020013220 Returns the average connection rate from the currently tracked counters,
13221 measured in amount of connections over the period configured in the table.
13222 See also src_conn_rate.
13223
Cyril Bonté62ba8702014-04-22 23:52:25 +020013224sc_get_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013225sc0_get_gpc0([<table>]) : integer
13226sc1_get_gpc0([<table>]) : integer
13227sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020013228 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013229 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020013230
Thierry FOURNIER236657b2015-08-19 08:25:14 +020013231sc_get_gpt0(<ctr>[,<table>]) : integer
13232sc0_get_gpt0([<table>]) : integer
13233sc1_get_gpt0([<table>]) : integer
13234sc2_get_gpt0([<table>]) : integer
13235 Returns the value of the first General Purpose Tag associated to the
13236 currently tracked counters. See also src_get_gpt0.
13237
Cyril Bonté62ba8702014-04-22 23:52:25 +020013238sc_gpc0_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013239sc0_gpc0_rate([<table>]) : integer
13240sc1_gpc0_rate([<table>]) : integer
13241sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020013242 Returns the average increment rate of the first General Purpose Counter
13243 associated to the currently tracked counters. It reports the frequency
13244 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013245 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
13246 that the "gpc0_rate" counter must be stored in the stick-table for a value to
13247 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +020013248
Cyril Bonté62ba8702014-04-22 23:52:25 +020013249sc_http_err_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013250sc0_http_err_cnt([<table>]) : integer
13251sc1_http_err_cnt([<table>]) : integer
13252sc2_http_err_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020013253 Returns the cumulated number of HTTP errors from the currently tracked
13254 counters. This includes the both request errors and 4xx error responses.
13255 See also src_http_err_cnt.
13256
Cyril Bonté62ba8702014-04-22 23:52:25 +020013257sc_http_err_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013258sc0_http_err_rate([<table>]) : integer
13259sc1_http_err_rate([<table>]) : integer
13260sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020013261 Returns the average rate of HTTP errors from the currently tracked counters,
13262 measured in amount of errors over the period configured in the table. This
13263 includes the both request errors and 4xx error responses. See also
13264 src_http_err_rate.
13265
Cyril Bonté62ba8702014-04-22 23:52:25 +020013266sc_http_req_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013267sc0_http_req_cnt([<table>]) : integer
13268sc1_http_req_cnt([<table>]) : integer
13269sc2_http_req_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020013270 Returns the cumulated number of HTTP requests from the currently tracked
13271 counters. This includes every started request, valid or not. See also
13272 src_http_req_cnt.
13273
Cyril Bonté62ba8702014-04-22 23:52:25 +020013274sc_http_req_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013275sc0_http_req_rate([<table>]) : integer
13276sc1_http_req_rate([<table>]) : integer
13277sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020013278 Returns the average rate of HTTP requests from the currently tracked
13279 counters, measured in amount of requests over the period configured in
13280 the table. This includes every started request, valid or not. See also
13281 src_http_req_rate.
13282
Cyril Bonté62ba8702014-04-22 23:52:25 +020013283sc_inc_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013284sc0_inc_gpc0([<table>]) : integer
13285sc1_inc_gpc0([<table>]) : integer
13286sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020013287 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +010013288 tracked counters, and returns its new value. Before the first invocation,
13289 the stored value is zero, so first invocation will increase it to 1 and will
13290 return 1. This is typically used as a second ACL in an expression in order
13291 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +020013292
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020013293 acl abuse sc0_http_req_rate gt 10
13294 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020013295 tcp-request connection reject if abuse kill
13296
Cyril Bonté62ba8702014-04-22 23:52:25 +020013297sc_kbytes_in(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013298sc0_kbytes_in([<table>]) : integer
13299sc1_kbytes_in([<table>]) : integer
13300sc2_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020013301 Returns the total amount of client-to-server data from the currently tracked
13302 counters, measured in kilobytes. The test is currently performed on 32-bit
13303 integers, which limits values to 4 terabytes. See also src_kbytes_in.
Willy Tarreaue9656522010-08-17 15:40:09 +020013304
Cyril Bonté62ba8702014-04-22 23:52:25 +020013305sc_kbytes_out(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013306sc0_kbytes_out([<table>]) : integer
13307sc1_kbytes_out([<table>]) : integer
13308sc2_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020013309 Returns the total amount of server-to-client data from the currently tracked
13310 counters, measured in kilobytes. The test is currently performed on 32-bit
13311 integers, which limits values to 4 terabytes. See also src_kbytes_out.
Willy Tarreaue9656522010-08-17 15:40:09 +020013312
Cyril Bonté62ba8702014-04-22 23:52:25 +020013313sc_sess_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013314sc0_sess_cnt([<table>]) : integer
13315sc1_sess_cnt([<table>]) : integer
13316sc2_sess_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020013317 Returns the cumulated number of incoming connections that were transformed
13318 into sessions, which means that they were accepted by a "tcp-request
13319 connection" rule, from the currently tracked counters. A backend may count
13320 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -040013321 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +020013322 with the client. See also src_sess_cnt.
13323
Cyril Bonté62ba8702014-04-22 23:52:25 +020013324sc_sess_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013325sc0_sess_rate([<table>]) : integer
13326sc1_sess_rate([<table>]) : integer
13327sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020013328 Returns the average session rate from the currently tracked counters,
13329 measured in amount of sessions over the period configured in the table. A
13330 session is a connection that got past the early "tcp-request connection"
13331 rules. A backend may count more sessions than connections because each
13332 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -040013333 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +020013334
Cyril Bonté62ba8702014-04-22 23:52:25 +020013335sc_tracked(<ctr>[,<table>]) : boolean
Willy Tarreau0f791d42013-07-23 19:56:43 +020013336sc0_tracked([<table>]) : boolean
13337sc1_tracked([<table>]) : boolean
13338sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +020013339 Returns true if the designated session counter is currently being tracked by
13340 the current session. This can be useful when deciding whether or not we want
13341 to set some values in a header passed to the server.
13342
Cyril Bonté62ba8702014-04-22 23:52:25 +020013343sc_trackers(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020013344sc0_trackers([<table>]) : integer
13345sc1_trackers([<table>]) : integer
13346sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +010013347 Returns the current amount of concurrent connections tracking the same
13348 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020013349 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +010013350 that it does not rely on any stored information but on the table's reference
13351 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020013352 may sometimes be more suited for layer7 tracking. It can be used to tell a
13353 server how many concurrent connections there are from a given address for
13354 example.
Willy Tarreau2406db42012-12-09 12:16:43 +010013355
Willy Tarreau74ca5042013-06-11 23:12:07 +020013356so_id : integer
13357 Returns an integer containing the current listening socket's id. It is useful
13358 in frontends involving many "bind" lines, or to stick all users coming via a
13359 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013360
Willy Tarreau74ca5042013-06-11 23:12:07 +020013361src : ip
13362 This is the source IPv4 address of the client of the session. It is of type
13363 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
13364 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
13365 TCP-level source address which is used, and not the address of a client
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010013366 behind a proxy. However if the "accept-proxy" or "accept-netscaler-cip" bind
13367 directive is used, it can be the address of a client behind another
13368 PROXY-protocol compatible component for all rule sets except
13369 "tcp-request connection" which sees the real address.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013370
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013371 Example:
13372 # add an HTTP header in requests with the originating address' country
13373 http-request set-header X-Country %[src,map_ip(geoip.lst)]
13374
Willy Tarreau74ca5042013-06-11 23:12:07 +020013375src_bytes_in_rate([<table>]) : integer
13376 Returns the average bytes rate from the incoming connection's source address
13377 in the current proxy's stick-table or in the designated stick-table, measured
13378 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013379 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013380
Willy Tarreau74ca5042013-06-11 23:12:07 +020013381src_bytes_out_rate([<table>]) : integer
13382 Returns the average bytes rate to the incoming connection's source address in
13383 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +020013384 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013385 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013386
Willy Tarreau74ca5042013-06-11 23:12:07 +020013387src_clr_gpc0([<table>]) : integer
13388 Clears the first General Purpose Counter associated to the incoming
13389 connection's source address in the current proxy's stick-table or in the
13390 designated stick-table, and returns its previous value. If the address is not
13391 found, an entry is created and 0 is returned. This is typically used as a
13392 second ACL in an expression in order to mark a connection when a first ACL
13393 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020013394
13395 # block if 5 consecutive requests continue to come faster than 10 sess
13396 # per second, and reset the counter as soon as the traffic slows down.
13397 acl abuse src_http_req_rate gt 10
13398 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +010013399 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020013400 tcp-request connection accept if !abuse save
13401 tcp-request connection reject if abuse kill
13402
Willy Tarreau74ca5042013-06-11 23:12:07 +020013403src_conn_cnt([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020013404 Returns the cumulated number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +020013405 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020013406 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013407 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013408
Willy Tarreau74ca5042013-06-11 23:12:07 +020013409src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020013410 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +020013411 current incoming connection's source address in the current proxy's
13412 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013413 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013414
Willy Tarreau74ca5042013-06-11 23:12:07 +020013415src_conn_rate([<table>]) : integer
13416 Returns the average connection rate from the incoming connection's source
13417 address in the current proxy's stick-table or in the designated stick-table,
13418 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013419 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013420
Willy Tarreau74ca5042013-06-11 23:12:07 +020013421src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020013422 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +020013423 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020013424 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013425 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013426
Thierry FOURNIER236657b2015-08-19 08:25:14 +020013427src_get_gpt0([<table>]) : integer
13428 Returns the value of the first General Purpose Tag associated to the
13429 incoming connection's source address in the current proxy's stick-table or in
13430 the designated stick-table. If the address is not found, zero is returned.
13431 See also sc/sc0/sc1/sc2_get_gpt0.
13432
Willy Tarreau74ca5042013-06-11 23:12:07 +020013433src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020013434 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +020013435 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +020013436 stick-table or in the designated stick-table. It reports the frequency
13437 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013438 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
13439 that the "gpc0_rate" counter must be stored in the stick-table for a value to
13440 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020013441
Willy Tarreau74ca5042013-06-11 23:12:07 +020013442src_http_err_cnt([<table>]) : integer
13443 Returns the cumulated number of HTTP errors from the incoming connection's
13444 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +020013445 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013446 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +020013447 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013448
Willy Tarreau74ca5042013-06-11 23:12:07 +020013449src_http_err_rate([<table>]) : integer
13450 Returns the average rate of HTTP errors from the incoming connection's source
13451 address in the current proxy's stick-table or in the designated stick-table,
13452 measured in amount of errors over the period configured in the table. This
13453 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013454 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013455
Willy Tarreau74ca5042013-06-11 23:12:07 +020013456src_http_req_cnt([<table>]) : integer
13457 Returns the cumulated number of HTTP requests from the incoming connection's
13458 source address in the current proxy's stick-table or in the designated stick-
13459 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013460 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013461
Willy Tarreau74ca5042013-06-11 23:12:07 +020013462src_http_req_rate([<table>]) : integer
13463 Returns the average rate of HTTP requests from the incoming connection's
13464 source address in the current proxy's stick-table or in the designated stick-
13465 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +020013466 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013467 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013468
Willy Tarreau74ca5042013-06-11 23:12:07 +020013469src_inc_gpc0([<table>]) : integer
13470 Increments the first General Purpose Counter associated to the incoming
13471 connection's source address in the current proxy's stick-table or in the
13472 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020013473 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013474 This is typically used as a second ACL in an expression in order to mark a
13475 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +020013476
13477 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +010013478 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020013479 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +020013480
Willy Tarreau16e01562016-08-09 16:46:18 +020013481src_is_local : boolean
13482 Returns true if the source address of the incoming connection is local to the
13483 system, or false if the address doesn't exist on the system, meaning that it
13484 comes from a remote machine. Note that UNIX addresses are considered local.
13485 It can be useful to apply certain access restrictions based on where the
13486 client comes from (eg: require auth or https for remote machines). Please
13487 note that the check involves a few system calls, so it's better to do it only
13488 once per connection.
13489
Willy Tarreau74ca5042013-06-11 23:12:07 +020013490src_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020013491 Returns the total amount of data received from the incoming connection's
13492 source address in the current proxy's stick-table or in the designated
13493 stick-table, measured in kilobytes. If the address is not found, zero is
13494 returned. The test is currently performed on 32-bit integers, which limits
13495 values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013496
Willy Tarreau74ca5042013-06-11 23:12:07 +020013497src_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020013498 Returns the total amount of data sent to the incoming connection's source
13499 address in the current proxy's stick-table or in the designated stick-table,
13500 measured in kilobytes. If the address is not found, zero is returned. The
13501 test is currently performed on 32-bit integers, which limits values to 4
13502 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020013503
Willy Tarreau74ca5042013-06-11 23:12:07 +020013504src_port : integer
13505 Returns an integer value corresponding to the TCP source port of the
13506 connection on the client side, which is the port the client connected from.
13507 Usage of this function is very limited as modern protocols do not care much
13508 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010013509
Willy Tarreau74ca5042013-06-11 23:12:07 +020013510src_sess_cnt([<table>]) : integer
13511 Returns the cumulated number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +020013512 connection's source IPv4 address in the current proxy's stick-table or in the
13513 designated stick-table, that were transformed into sessions, which means that
13514 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013515 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013516
Willy Tarreau74ca5042013-06-11 23:12:07 +020013517src_sess_rate([<table>]) : integer
13518 Returns the average session rate from the incoming connection's source
13519 address in the current proxy's stick-table or in the designated stick-table,
13520 measured in amount of sessions over the period configured in the table. A
13521 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020013522 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020013523
Willy Tarreau74ca5042013-06-11 23:12:07 +020013524src_updt_conn_cnt([<table>]) : integer
13525 Creates or updates the entry associated to the incoming connection's source
13526 address in the current proxy's stick-table or in the designated stick-table.
13527 This table must be configured to store the "conn_cnt" data type, otherwise
13528 the match will be ignored. The current count is incremented by one, and the
13529 expiration timer refreshed. The updated count is returned, so this match
13530 can't return zero. This was used to reject service abusers based on their
13531 source address. Note: it is recommended to use the more complete "track-sc*"
13532 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020013533
13534 Example :
13535 # This frontend limits incoming SSH connections to 3 per 10 second for
13536 # each source address, and rejects excess connections until a 10 second
13537 # silence is observed. At most 20 addresses are tracked.
13538 listen ssh
13539 bind :22
13540 mode tcp
13541 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +020013542 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +020013543 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +020013544 server local 127.0.0.1:22
13545
Willy Tarreau74ca5042013-06-11 23:12:07 +020013546srv_id : integer
13547 Returns an integer containing the server's id when processing the response.
13548 While it's almost only used with ACLs, it may be used for logging or
13549 debugging.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +020013550
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200135517.3.4. Fetching samples at Layer 5
Willy Tarreau74ca5042013-06-11 23:12:07 +020013552----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +020013553
Willy Tarreau74ca5042013-06-11 23:12:07 +020013554The layer 5 usually describes just the session layer which in haproxy is
13555closest to the session once all the connection handshakes are finished, but
13556when no content is yet made available. The fetch methods described here are
13557usable as low as the "tcp-request content" rule sets unless they require some
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013558future information. Those generally include the results of SSL negotiations.
Willy Tarreauc735a072011-03-29 00:57:02 +020013559
Emeric Brun645ae792014-04-30 14:21:06 +020013560ssl_bc : boolean
13561 Returns true when the back connection was made via an SSL/TLS transport
13562 layer and is locally deciphered. This means the outgoing connection was made
13563 other a server with the "ssl" option.
13564
13565ssl_bc_alg_keysize : integer
13566 Returns the symmetric cipher key size supported in bits when the outgoing
13567 connection was made over an SSL/TLS transport layer.
13568
13569ssl_bc_cipher : string
13570 Returns the name of the used cipher when the outgoing connection was made
13571 over an SSL/TLS transport layer.
13572
13573ssl_bc_protocol : string
13574 Returns the name of the used protocol when the outgoing connection was made
13575 over an SSL/TLS transport layer.
13576
Emeric Brunb73a9b02014-04-30 18:49:19 +020013577ssl_bc_unique_id : binary
Emeric Brun645ae792014-04-30 14:21:06 +020013578 When the outgoing connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020013579 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
13580 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
Emeric Brun645ae792014-04-30 14:21:06 +020013581
13582ssl_bc_session_id : binary
13583 Returns the SSL ID of the back connection when the outgoing connection was
13584 made over an SSL/TLS transport layer. It is useful to log if we want to know
13585 if session was reused or not.
13586
13587ssl_bc_use_keysize : integer
13588 Returns the symmetric cipher key size used in bits when the outgoing
13589 connection was made over an SSL/TLS transport layer.
13590
Willy Tarreau74ca5042013-06-11 23:12:07 +020013591ssl_c_ca_err : integer
13592 When the incoming connection was made over an SSL/TLS transport layer,
13593 returns the ID of the first error detected during verification of the client
13594 certificate at depth > 0, or 0 if no error was encountered during this
13595 verification process. Please refer to your SSL library's documentation to
13596 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +020013597
Willy Tarreau74ca5042013-06-11 23:12:07 +020013598ssl_c_ca_err_depth : integer
13599 When the incoming connection was made over an SSL/TLS transport layer,
13600 returns the depth in the CA chain of the first error detected during the
13601 verification of the client certificate. If no error is encountered, 0 is
13602 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +010013603
Emeric Brun43e79582014-10-29 19:03:26 +010013604ssl_c_der : binary
13605 Returns the DER formatted certificate presented by the client when the
13606 incoming connection was made over an SSL/TLS transport layer. When used for
13607 an ACL, the value(s) to match against can be passed in hexadecimal form.
13608
Willy Tarreau74ca5042013-06-11 23:12:07 +020013609ssl_c_err : integer
13610 When the incoming connection was made over an SSL/TLS transport layer,
13611 returns the ID of the first error detected during verification at depth 0, or
13612 0 if no error was encountered during this verification process. Please refer
13613 to your SSL library's documentation to find the exhaustive list of error
13614 codes.
Willy Tarreau62644772008-07-16 18:36:06 +020013615
Willy Tarreau74ca5042013-06-11 23:12:07 +020013616ssl_c_i_dn([<entry>[,<occ>]]) : string
13617 When the incoming connection was made over an SSL/TLS transport layer,
13618 returns the full distinguished name of the issuer of the certificate
13619 presented by the client when no <entry> is specified, or the value of the
13620 first given entry found from the beginning of the DN. If a positive/negative
13621 occurrence number is specified as the optional second argument, it returns
13622 the value of the nth given entry value from the beginning/end of the DN.
13623 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
13624 "ssl_c_i_dn(CN)" retrieves the common name.
Willy Tarreau62644772008-07-16 18:36:06 +020013625
Willy Tarreau74ca5042013-06-11 23:12:07 +020013626ssl_c_key_alg : string
13627 Returns the name of the algorithm used to generate the key of the certificate
13628 presented by the client when the incoming connection was made over an SSL/TLS
13629 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +020013630
Willy Tarreau74ca5042013-06-11 23:12:07 +020013631ssl_c_notafter : string
13632 Returns the end date presented by the client as a formatted string
13633 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
13634 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +020013635
Willy Tarreau74ca5042013-06-11 23:12:07 +020013636ssl_c_notbefore : string
13637 Returns the start date presented by the client as a formatted string
13638 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
13639 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +010013640
Willy Tarreau74ca5042013-06-11 23:12:07 +020013641ssl_c_s_dn([<entry>[,<occ>]]) : string
13642 When the incoming connection was made over an SSL/TLS transport layer,
13643 returns the full distinguished name of the subject of the certificate
13644 presented by the client when no <entry> is specified, or the value of the
13645 first given entry found from the beginning of the DN. If a positive/negative
13646 occurrence number is specified as the optional second argument, it returns
13647 the value of the nth given entry value from the beginning/end of the DN.
13648 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
13649 "ssl_c_s_dn(CN)" retrieves the common name.
Willy Tarreaub6672b52011-12-12 17:23:41 +010013650
Willy Tarreau74ca5042013-06-11 23:12:07 +020013651ssl_c_serial : binary
13652 Returns the serial of the certificate presented by the client when the
13653 incoming connection was made over an SSL/TLS transport layer. When used for
13654 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +020013655
Willy Tarreau74ca5042013-06-11 23:12:07 +020013656ssl_c_sha1 : binary
13657 Returns the SHA-1 fingerprint of the certificate presented by the client when
13658 the incoming connection was made over an SSL/TLS transport layer. This can be
13659 used to stick a client to a server, or to pass this information to a server.
Willy Tarreau2d0caa32014-07-02 19:01:22 +020013660 Note that the output is binary, so if you want to pass that signature to the
13661 server, you need to encode it in hex or base64, such as in the example below:
13662
13663 http-request set-header X-SSL-Client-SHA1 %[ssl_c_sha1,hex]
Emeric Brun2525b6b2012-10-18 15:59:43 +020013664
Willy Tarreau74ca5042013-06-11 23:12:07 +020013665ssl_c_sig_alg : string
13666 Returns the name of the algorithm used to sign the certificate presented by
13667 the client when the incoming connection was made over an SSL/TLS transport
13668 layer.
Emeric Brun87855892012-10-17 17:39:35 +020013669
Willy Tarreau74ca5042013-06-11 23:12:07 +020013670ssl_c_used : boolean
13671 Returns true if current SSL session uses a client certificate even if current
13672 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +020013673
Willy Tarreau74ca5042013-06-11 23:12:07 +020013674ssl_c_verify : integer
13675 Returns the verify result error ID when the incoming connection was made over
13676 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
13677 refer to your SSL library's documentation for an exhaustive list of error
13678 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +020013679
Willy Tarreau74ca5042013-06-11 23:12:07 +020013680ssl_c_version : integer
13681 Returns the version of the certificate presented by the client when the
13682 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +020013683
Emeric Brun43e79582014-10-29 19:03:26 +010013684ssl_f_der : binary
13685 Returns the DER formatted certificate presented by the frontend when the
13686 incoming connection was made over an SSL/TLS transport layer. When used for
13687 an ACL, the value(s) to match against can be passed in hexadecimal form.
13688
Willy Tarreau74ca5042013-06-11 23:12:07 +020013689ssl_f_i_dn([<entry>[,<occ>]]) : string
13690 When the incoming connection was made over an SSL/TLS transport layer,
13691 returns the full distinguished name of the issuer of the certificate
13692 presented by the frontend when no <entry> is specified, or the value of the
13693 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +020013694 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +020013695 the value of the nth given entry value from the beginning/end of the DN.
13696 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
13697 "ssl_f_i_dn(CN)" retrieves the common name.
Emeric Brun87855892012-10-17 17:39:35 +020013698
Willy Tarreau74ca5042013-06-11 23:12:07 +020013699ssl_f_key_alg : string
13700 Returns the name of the algorithm used to generate the key of the certificate
13701 presented by the frontend when the incoming connection was made over an
13702 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020013703
Willy Tarreau74ca5042013-06-11 23:12:07 +020013704ssl_f_notafter : string
13705 Returns the end date presented by the frontend as a formatted string
13706 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
13707 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +020013708
Willy Tarreau74ca5042013-06-11 23:12:07 +020013709ssl_f_notbefore : string
13710 Returns the start date presented by the frontend as a formatted string
13711 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
13712 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +020013713
Willy Tarreau74ca5042013-06-11 23:12:07 +020013714ssl_f_s_dn([<entry>[,<occ>]]) : string
13715 When the incoming connection was made over an SSL/TLS transport layer,
13716 returns the full distinguished name of the subject of the certificate
13717 presented by the frontend when no <entry> is specified, or the value of the
13718 first given entry found from the beginning of the DN. If a positive/negative
13719 occurrence number is specified as the optional second argument, it returns
13720 the value of the nth given entry value from the beginning/end of the DN.
13721 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
13722 "ssl_f_s_dn(CN)" retrieves the common name.
Emeric Brunce5ad802012-10-22 14:11:22 +020013723
Willy Tarreau74ca5042013-06-11 23:12:07 +020013724ssl_f_serial : binary
13725 Returns the serial of the certificate presented by the frontend when the
13726 incoming connection was made over an SSL/TLS transport layer. When used for
13727 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +020013728
Emeric Brun55f4fa82014-04-30 17:11:25 +020013729ssl_f_sha1 : binary
13730 Returns the SHA-1 fingerprint of the certificate presented by the frontend
13731 when the incoming connection was made over an SSL/TLS transport layer. This
13732 can be used to know which certificate was chosen using SNI.
13733
Willy Tarreau74ca5042013-06-11 23:12:07 +020013734ssl_f_sig_alg : string
13735 Returns the name of the algorithm used to sign the certificate presented by
13736 the frontend when the incoming connection was made over an SSL/TLS transport
13737 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020013738
Willy Tarreau74ca5042013-06-11 23:12:07 +020013739ssl_f_version : integer
13740 Returns the version of the certificate presented by the frontend when the
13741 incoming connection was made over an SSL/TLS transport layer.
13742
13743ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020013744 Returns true when the front connection was made via an SSL/TLS transport
13745 layer and is locally deciphered. This means it has matched a socket declared
13746 with a "bind" line having the "ssl" option.
13747
Willy Tarreau74ca5042013-06-11 23:12:07 +020013748 Example :
13749 # This passes "X-Proto: https" to servers when client connects over SSL
13750 listen http-https
13751 bind :80
13752 bind :443 ssl crt /etc/haproxy.pem
13753 http-request add-header X-Proto https if { ssl_fc }
13754
13755ssl_fc_alg_keysize : integer
13756 Returns the symmetric cipher key size supported in bits when the incoming
13757 connection was made over an SSL/TLS transport layer.
13758
13759ssl_fc_alpn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013760 This extracts the Application Layer Protocol Negotiation field from an
Willy Tarreau74ca5042013-06-11 23:12:07 +020013761 incoming connection made via a TLS transport layer and locally deciphered by
13762 haproxy. The result is a string containing the protocol name advertised by
13763 the client. The SSL library must have been built with support for TLS
13764 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
13765 not advertised unless the "alpn" keyword on the "bind" line specifies a
13766 protocol list. Also, nothing forces the client to pick a protocol from this
13767 list, any other one may be requested. The TLS ALPN extension is meant to
13768 replace the TLS NPN extension. See also "ssl_fc_npn".
13769
Willy Tarreau74ca5042013-06-11 23:12:07 +020013770ssl_fc_cipher : string
13771 Returns the name of the used cipher when the incoming connection was made
13772 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +020013773
Willy Tarreau74ca5042013-06-11 23:12:07 +020013774ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020013775 Returns true if a client certificate is present in an incoming connection over
13776 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +010013777 Note: on SSL session resumption with Session ID or TLS ticket, client
13778 certificate is not present in the current connection but may be retrieved
13779 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
13780 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +020013781
Willy Tarreau74ca5042013-06-11 23:12:07 +020013782ssl_fc_has_sni : boolean
13783 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020013784 in an incoming connection was made over an SSL/TLS transport layer. Returns
13785 true when the incoming connection presents a TLS SNI field. This requires
13786 that the SSL library is build with support for TLS extensions enabled (check
13787 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +020013788
Nenad Merdanovic1516fe32016-05-17 03:31:21 +020013789ssl_fc_is_resumed : boolean
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020013790 Returns true if the SSL/TLS session has been resumed through the use of
13791 SSL session cache or TLS tickets.
13792
Willy Tarreau74ca5042013-06-11 23:12:07 +020013793ssl_fc_npn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013794 This extracts the Next Protocol Negotiation field from an incoming connection
Willy Tarreau74ca5042013-06-11 23:12:07 +020013795 made via a TLS transport layer and locally deciphered by haproxy. The result
13796 is a string containing the protocol name advertised by the client. The SSL
13797 library must have been built with support for TLS extensions enabled (check
13798 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
13799 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
13800 forces the client to pick a protocol from this list, any other one may be
13801 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +020013802
Willy Tarreau74ca5042013-06-11 23:12:07 +020013803ssl_fc_protocol : string
13804 Returns the name of the used protocol when the incoming connection was made
13805 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +020013806
Emeric Brunb73a9b02014-04-30 18:49:19 +020013807ssl_fc_unique_id : binary
David Sc1ad52e2014-04-08 18:48:47 -040013808 When the incoming connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020013809 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
13810 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
David Sc1ad52e2014-04-08 18:48:47 -040013811
Willy Tarreau74ca5042013-06-11 23:12:07 +020013812ssl_fc_session_id : binary
13813 Returns the SSL ID of the front connection when the incoming connection was
13814 made over an SSL/TLS transport layer. It is useful to stick a given client to
13815 a server. It is important to note that some browsers refresh their session ID
13816 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +020013817
Willy Tarreau74ca5042013-06-11 23:12:07 +020013818ssl_fc_sni : string
13819 This extracts the Server Name Indication TLS extension (SNI) field from an
13820 incoming connection made via an SSL/TLS transport layer and locally
13821 deciphered by haproxy. The result (when present) typically is a string
13822 matching the HTTPS host name (253 chars or less). The SSL library must have
13823 been built with support for TLS extensions enabled (check haproxy -vv).
13824
13825 This fetch is different from "req_ssl_sni" above in that it applies to the
13826 connection being deciphered by haproxy and not to SSL contents being blindly
13827 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +020013828 requires that the SSL library is build with support for TLS extensions
13829 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +020013830
Willy Tarreau74ca5042013-06-11 23:12:07 +020013831 ACL derivatives :
Willy Tarreau74ca5042013-06-11 23:12:07 +020013832 ssl_fc_sni_end : suffix match
13833 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +020013834
Willy Tarreau74ca5042013-06-11 23:12:07 +020013835ssl_fc_use_keysize : integer
13836 Returns the symmetric cipher key size used in bits when the incoming
13837 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020013838
Willy Tarreaub6fb4202008-07-20 11:18:28 +020013839
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200138407.3.5. Fetching samples from buffer contents (Layer 6)
Willy Tarreau74ca5042013-06-11 23:12:07 +020013841------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +020013842
Willy Tarreau74ca5042013-06-11 23:12:07 +020013843Fetching samples from buffer contents is a bit different from the previous
13844sample fetches above because the sampled data are ephemeral. These data can
13845only be used when they're available and will be lost when they're forwarded.
13846For this reason, samples fetched from buffer contents during a request cannot
13847be used in a response for example. Even while the data are being fetched, they
13848can change. Sometimes it is necessary to set some delays or combine multiple
13849sample fetch methods to ensure that the expected data are complete and usable,
13850for example through TCP request content inspection. Please see the "tcp-request
13851content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +020013852
Willy Tarreau74ca5042013-06-11 23:12:07 +020013853payload(<offset>,<length>) : binary (deprecated)
13854 This is an alias for "req.payload" when used in the context of a request (eg:
13855 "stick on", "stick match"), and for "res.payload" when used in the context of
13856 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010013857
Willy Tarreau74ca5042013-06-11 23:12:07 +020013858payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
13859 This is an alias for "req.payload_lv" when used in the context of a request
13860 (eg: "stick on", "stick match"), and for "res.payload_lv" when used in the
13861 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010013862
Willy Tarreau74ca5042013-06-11 23:12:07 +020013863req.len : integer
13864req_len : integer (deprecated)
13865 Returns an integer value corresponding to the number of bytes present in the
13866 request buffer. This is mostly used in ACL. It is important to understand
13867 that this test does not return false as long as the buffer is changing. This
13868 means that a check with equality to zero will almost always immediately match
13869 at the beginning of the session, while a test for more data will wait for
13870 that data to come in and return false only when haproxy is certain that no
13871 more data will come in. This test was designed to be used with TCP request
13872 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013873
Willy Tarreau74ca5042013-06-11 23:12:07 +020013874req.payload(<offset>,<length>) : binary
13875 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020013876 in the request buffer. As a special case, if the <length> argument is zero,
13877 the the whole buffer from <offset> to the end is extracted. This can be used
13878 with ACLs in order to check for the presence of some content in a buffer at
13879 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013880
Willy Tarreau74ca5042013-06-11 23:12:07 +020013881 ACL alternatives :
13882 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013883
Willy Tarreau74ca5042013-06-11 23:12:07 +020013884req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
13885 This extracts a binary block whose size is specified at <offset1> for <length>
13886 bytes, and which starts at <offset2> if specified or just after the length in
13887 the request buffer. The <offset2> parameter also supports relative offsets if
13888 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013889
Willy Tarreau74ca5042013-06-11 23:12:07 +020013890 ACL alternatives :
13891 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013892
Willy Tarreau74ca5042013-06-11 23:12:07 +020013893 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013894
Willy Tarreau74ca5042013-06-11 23:12:07 +020013895req.proto_http : boolean
13896req_proto_http : boolean (deprecated)
13897 Returns true when data in the request buffer look like HTTP and correctly
13898 parses as such. It is the same parser as the common HTTP request parser which
13899 is used so there should be no surprises. The test does not match until the
13900 request is complete, failed or timed out. This test may be used to report the
13901 protocol in TCP logs, but the biggest use is to block TCP request analysis
13902 until a complete HTTP request is present in the buffer, for example to track
13903 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013904
Willy Tarreau74ca5042013-06-11 23:12:07 +020013905 Example:
13906 # track request counts per "base" (concatenation of Host+URL)
13907 tcp-request inspect-delay 10s
13908 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020013909 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013910
Willy Tarreau74ca5042013-06-11 23:12:07 +020013911req.rdp_cookie([<name>]) : string
13912rdp_cookie([<name>]) : string (deprecated)
13913 When the request buffer looks like the RDP protocol, extracts the RDP cookie
13914 <name>, or any cookie if unspecified. The parser only checks for the first
13915 cookie, as illustrated in the RDP protocol specification. The cookie name is
13916 case insensitive. Generally the "MSTS" cookie name will be used, as it can
13917 contain the user name of the client connecting to the server if properly
13918 configured on the client. The "MSTSHASH" cookie is often used as well for
13919 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013920
Willy Tarreau74ca5042013-06-11 23:12:07 +020013921 This differs from "balance rdp-cookie" in that any balancing algorithm may be
13922 used and thus the distribution of clients to backend servers is not linked to
13923 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
13924 such as "balance roundrobin" or "balance leastconn" will lead to a more even
13925 distribution of clients to backend servers than the hash used by "balance
13926 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013927
Willy Tarreau74ca5042013-06-11 23:12:07 +020013928 ACL derivatives :
13929 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013930
Willy Tarreau74ca5042013-06-11 23:12:07 +020013931 Example :
13932 listen tse-farm
13933 bind 0.0.0.0:3389
13934 # wait up to 5s for an RDP cookie in the request
13935 tcp-request inspect-delay 5s
13936 tcp-request content accept if RDP_COOKIE
13937 # apply RDP cookie persistence
13938 persist rdp-cookie
13939 # Persist based on the mstshash cookie
13940 # This is only useful makes sense if
13941 # balance rdp-cookie is not used
13942 stick-table type string size 204800
13943 stick on req.rdp_cookie(mstshash)
13944 server srv1 1.1.1.1:3389
13945 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013946
Willy Tarreau74ca5042013-06-11 23:12:07 +020013947 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
13948 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013949
Willy Tarreau74ca5042013-06-11 23:12:07 +020013950req.rdp_cookie_cnt([name]) : integer
13951rdp_cookie_cnt([name]) : integer (deprecated)
13952 Tries to parse the request buffer as RDP protocol, then returns an integer
13953 corresponding to the number of RDP cookies found. If an optional cookie name
13954 is passed, only cookies matching this name are considered. This is mostly
13955 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013956
Willy Tarreau74ca5042013-06-11 23:12:07 +020013957 ACL derivatives :
13958 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013959
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020013960req.ssl_ec_ext : boolean
13961 Returns a boolean identifying if client sent the Supported Elliptic Curves
13962 Extension as defined in RFC4492, section 5.1. within the SSL ClientHello
Cyril Bonté307ee1e2015-09-28 23:16:06 +020013963 message. This can be used to present ECC compatible clients with EC
13964 certificate and to use RSA for all others, on the same IP address. Note that
13965 this only applies to raw contents found in the request buffer and not to
13966 contents deciphered via an SSL data layer, so this will not work with "bind"
13967 lines having the "ssl" option.
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020013968
Willy Tarreau74ca5042013-06-11 23:12:07 +020013969req.ssl_hello_type : integer
13970req_ssl_hello_type : integer (deprecated)
13971 Returns an integer value containing the type of the SSL hello message found
13972 in the request buffer if the buffer contains data that parse as a complete
13973 SSL (v3 or superior) client hello message. Note that this only applies to raw
13974 contents found in the request buffer and not to contents deciphered via an
13975 SSL data layer, so this will not work with "bind" lines having the "ssl"
13976 option. This is mostly used in ACL to detect presence of an SSL hello message
13977 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013978
Willy Tarreau74ca5042013-06-11 23:12:07 +020013979req.ssl_sni : string
13980req_ssl_sni : string (deprecated)
13981 Returns a string containing the value of the Server Name TLS extension sent
13982 by a client in a TLS stream passing through the request buffer if the buffer
13983 contains data that parse as a complete SSL (v3 or superior) client hello
13984 message. Note that this only applies to raw contents found in the request
13985 buffer and not to contents deciphered via an SSL data layer, so this will not
13986 work with "bind" lines having the "ssl" option. SNI normally contains the
13987 name of the host the client tries to connect to (for recent browsers). SNI is
13988 useful for allowing or denying access to certain hosts when SSL/TLS is used
13989 by the client. This test was designed to be used with TCP request content
13990 inspection. If content switching is needed, it is recommended to first wait
13991 for a complete client hello (type 1), like in the example below. See also
13992 "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013993
Willy Tarreau74ca5042013-06-11 23:12:07 +020013994 ACL derivatives :
13995 req_ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013996
Willy Tarreau74ca5042013-06-11 23:12:07 +020013997 Examples :
13998 # Wait for a client hello for at most 5 seconds
13999 tcp-request inspect-delay 5s
14000 tcp-request content accept if { req_ssl_hello_type 1 }
14001 use_backend bk_allow if { req_ssl_sni -f allowed_sites }
14002 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +020014003
Pradeep Jindalbb2acf52015-09-29 10:12:57 +053014004req.ssl_st_ext : integer
14005 Returns 0 if the client didn't send a SessionTicket TLS Extension (RFC5077)
14006 Returns 1 if the client sent SessionTicket TLS Extension
14007 Returns 2 if the client also sent non-zero length TLS SessionTicket
14008 Note that this only applies to raw contents found in the request buffer and
14009 not to contents deciphered via an SSL data layer, so this will not work with
14010 "bind" lines having the "ssl" option. This can for example be used to detect
14011 whether the client sent a SessionTicket or not and stick it accordingly, if
14012 no SessionTicket then stick on SessionID or don't stick as there's no server
14013 side state is there when SessionTickets are in use.
14014
Willy Tarreau74ca5042013-06-11 23:12:07 +020014015req.ssl_ver : integer
14016req_ssl_ver : integer (deprecated)
14017 Returns an integer value containing the version of the SSL/TLS protocol of a
14018 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
14019 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
14020 composed of the major version multiplied by 65536, added to the minor
14021 version. Note that this only applies to raw contents found in the request
14022 buffer and not to contents deciphered via an SSL data layer, so this will not
14023 work with "bind" lines having the "ssl" option. The ACL version of the test
14024 matches against a decimal notation in the form MAJOR.MINOR (eg: 3.1). This
14025 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014026
Willy Tarreau74ca5042013-06-11 23:12:07 +020014027 ACL derivatives :
14028 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +010014029
Willy Tarreau47e8eba2013-09-11 23:28:46 +020014030res.len : integer
14031 Returns an integer value corresponding to the number of bytes present in the
14032 response buffer. This is mostly used in ACL. It is important to understand
14033 that this test does not return false as long as the buffer is changing. This
14034 means that a check with equality to zero will almost always immediately match
14035 at the beginning of the session, while a test for more data will wait for
14036 that data to come in and return false only when haproxy is certain that no
14037 more data will come in. This test was designed to be used with TCP response
14038 content inspection.
14039
Willy Tarreau74ca5042013-06-11 23:12:07 +020014040res.payload(<offset>,<length>) : binary
14041 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020014042 in the response buffer. As a special case, if the <length> argument is zero,
14043 the the whole buffer from <offset> to the end is extracted. This can be used
14044 with ACLs in order to check for the presence of some content in a buffer at
14045 any location.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014046
Willy Tarreau74ca5042013-06-11 23:12:07 +020014047res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
14048 This extracts a binary block whose size is specified at <offset1> for <length>
14049 bytes, and which starts at <offset2> if specified or just after the length in
14050 the response buffer. The <offset2> parameter also supports relative offsets
14051 if prepended with a '+' or '-' sign.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014052
Willy Tarreau74ca5042013-06-11 23:12:07 +020014053 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014054
Willy Tarreau971f7b62015-09-29 14:06:59 +020014055res.ssl_hello_type : integer
14056rep_ssl_hello_type : integer (deprecated)
14057 Returns an integer value containing the type of the SSL hello message found
14058 in the response buffer if the buffer contains data that parses as a complete
14059 SSL (v3 or superior) hello message. Note that this only applies to raw
14060 contents found in the response buffer and not to contents deciphered via an
14061 SSL data layer, so this will not work with "server" lines having the "ssl"
14062 option. This is mostly used in ACL to detect presence of an SSL hello message
14063 that is supposed to contain an SSL session ID usable for stickiness.
14064
Willy Tarreau74ca5042013-06-11 23:12:07 +020014065wait_end : boolean
14066 This fetch either returns true when the inspection period is over, or does
14067 not fetch. It is only used in ACLs, in conjunction with content analysis to
14068 avoid returning a wrong verdict early. It may also be used to delay some
14069 actions, such as a delayed reject for some special addresses. Since it either
14070 stops the rules evaluation or immediately returns true, it is recommended to
14071 use this acl as the last one in a rule. Please note that the default ACL
14072 "WAIT_END" is always usable without prior declaration. This test was designed
14073 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014074
Willy Tarreau74ca5042013-06-11 23:12:07 +020014075 Examples :
14076 # delay every incoming request by 2 seconds
14077 tcp-request inspect-delay 2s
14078 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +010014079
Willy Tarreau74ca5042013-06-11 23:12:07 +020014080 # don't immediately tell bad guys they are rejected
14081 tcp-request inspect-delay 10s
14082 acl goodguys src 10.0.0.0/24
14083 acl badguys src 10.0.1.0/24
14084 tcp-request content accept if goodguys
14085 tcp-request content reject if badguys WAIT_END
14086 tcp-request content reject
14087
14088
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200140897.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +020014090--------------------------------------
14091
14092It is possible to fetch samples from HTTP contents, requests and responses.
14093This application layer is also called layer 7. It is only possible to fetch the
14094data in this section when a full HTTP request or response has been parsed from
14095its respective request or response buffer. This is always the case with all
14096HTTP specific rules and for sections running with "mode http". When using TCP
14097content inspection, it may be necessary to support an inspection delay in order
14098to let the request or response come in first. These fetches may require a bit
14099more CPU resources than the layer 4 ones, but not much since the request and
14100response are indexed.
14101
14102base : string
14103 This returns the concatenation of the first Host header and the path part of
14104 the request, which starts at the first slash and ends before the question
14105 mark. It can be useful in virtual hosted environments to detect URL abuses as
14106 well as to improve shared caches efficiency. Using this with a limited size
14107 stick table also allows one to collect statistics about most commonly
14108 requested objects by host/path. With ACLs it can allow simple content
14109 switching rules involving the host and the path at the same time, such as
14110 "www.example.com/favicon.ico". See also "path" and "uri".
14111
14112 ACL derivatives :
14113 base : exact string match
14114 base_beg : prefix match
14115 base_dir : subdir match
14116 base_dom : domain match
14117 base_end : suffix match
14118 base_len : length match
14119 base_reg : regex match
14120 base_sub : substring match
14121
14122base32 : integer
14123 This returns a 32-bit hash of the value returned by the "base" fetch method
14124 above. This is useful to track per-URL activity on high traffic sites without
14125 having to store all URLs. Instead a shorter hash is stored, saving a lot of
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014126 memory. The output type is an unsigned integer. The hash function used is
14127 SDBM with full avalanche on the output. Technically, base32 is exactly equal
14128 to "base,sdbm(1)".
Willy Tarreau74ca5042013-06-11 23:12:07 +020014129
14130base32+src : binary
14131 This returns the concatenation of the base32 fetch above and the src fetch
14132 below. The resulting type is of type binary, with a size of 8 or 20 bytes
14133 depending on the source address family. This can be used to track per-IP,
14134 per-URL counters.
14135
William Lallemand65ad6e12014-01-31 15:08:02 +010014136capture.req.hdr(<idx>) : string
14137 This extracts the content of the header captured by the "capture request
14138 header", idx is the position of the capture keyword in the configuration.
14139 The first entry is an index of 0. See also: "capture request header".
14140
14141capture.req.method : string
14142 This extracts the METHOD of an HTTP request. It can be used in both request
14143 and response. Unlike "method", it can be used in both request and response
14144 because it's allocated.
14145
14146capture.req.uri : string
14147 This extracts the request's URI, which starts at the first slash and ends
14148 before the first space in the request (without the host part). Unlike "path"
14149 and "url", it can be used in both request and response because it's
14150 allocated.
14151
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020014152capture.req.ver : string
14153 This extracts the request's HTTP version and returns either "HTTP/1.0" or
14154 "HTTP/1.1". Unlike "req.ver", it can be used in both request, response, and
14155 logs because it relies on a persistent flag.
14156
William Lallemand65ad6e12014-01-31 15:08:02 +010014157capture.res.hdr(<idx>) : string
14158 This extracts the content of the header captured by the "capture response
14159 header", idx is the position of the capture keyword in the configuration.
14160 The first entry is an index of 0.
14161 See also: "capture response header"
14162
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020014163capture.res.ver : string
14164 This extracts the response's HTTP version and returns either "HTTP/1.0" or
14165 "HTTP/1.1". Unlike "res.ver", it can be used in logs because it relies on a
14166 persistent flag.
14167
Willy Tarreaua5910cc2015-05-02 00:46:08 +020014168req.body : binary
14169 This returns the HTTP request's available body as a block of data. It
14170 requires that the request body has been buffered made available using
14171 "option http-buffer-request". In case of chunked-encoded body, currently only
14172 the first chunk is analyzed.
14173
Thierry FOURNIER9826c772015-05-20 15:50:54 +020014174req.body_param([<name>) : string
14175 This fetch assumes that the body of the POST request is url-encoded. The user
14176 can check if the "content-type" contains the value
14177 "application/x-www-form-urlencoded". This extracts the first occurrence of the
14178 parameter <name> in the body, which ends before '&'. The parameter name is
14179 case-sensitive. If no name is given, any parameter will match, and the first
14180 one will be returned. The result is a string corresponding to the value of the
14181 parameter <name> as presented in the request body (no URL decoding is
14182 performed). Note that the ACL version of this fetch iterates over multiple
14183 parameters and will iteratively report all parameters values if no name is
14184 given.
14185
Willy Tarreaua5910cc2015-05-02 00:46:08 +020014186req.body_len : integer
14187 This returns the length of the HTTP request's available body in bytes. It may
14188 be lower than the advertised length if the body is larger than the buffer. It
14189 requires that the request body has been buffered made available using
14190 "option http-buffer-request".
14191
14192req.body_size : integer
14193 This returns the advertised length of the HTTP request's body in bytes. It
14194 will represent the advertised Content-Length header, or the size of the first
14195 chunk in case of chunked encoding. In order to parse the chunks, it requires
14196 that the request body has been buffered made available using
14197 "option http-buffer-request".
14198
Willy Tarreau74ca5042013-06-11 23:12:07 +020014199req.cook([<name>]) : string
14200cook([<name>]) : string (deprecated)
14201 This extracts the last occurrence of the cookie name <name> on a "Cookie"
14202 header line from the request, and returns its value as string. If no name is
14203 specified, the first cookie value is returned. When used with ACLs, all
14204 matching cookies are evaluated. Spaces around the name and the value are
14205 ignored as requested by the Cookie header specification (RFC6265). The cookie
14206 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
14207 well return an empty value if it is present. Use the "found" match to detect
14208 presence. Use the res.cook() variant for response cookies sent by the server.
14209
14210 ACL derivatives :
14211 cook([<name>]) : exact string match
14212 cook_beg([<name>]) : prefix match
14213 cook_dir([<name>]) : subdir match
14214 cook_dom([<name>]) : domain match
14215 cook_end([<name>]) : suffix match
14216 cook_len([<name>]) : length match
14217 cook_reg([<name>]) : regex match
14218 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +010014219
Willy Tarreau74ca5042013-06-11 23:12:07 +020014220req.cook_cnt([<name>]) : integer
14221cook_cnt([<name>]) : integer (deprecated)
14222 Returns an integer value representing the number of occurrences of the cookie
14223 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014224
Willy Tarreau74ca5042013-06-11 23:12:07 +020014225req.cook_val([<name>]) : integer
14226cook_val([<name>]) : integer (deprecated)
14227 This extracts the last occurrence of the cookie name <name> on a "Cookie"
14228 header line from the request, and converts its value to an integer which is
14229 returned. If no name is specified, the first cookie value is returned. When
14230 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +020014231
Willy Tarreau74ca5042013-06-11 23:12:07 +020014232cookie([<name>]) : string (deprecated)
14233 This extracts the last occurrence of the cookie name <name> on a "Cookie"
14234 header line from the request, or a "Set-Cookie" header from the response, and
14235 returns its value as a string. A typical use is to get multiple clients
14236 sharing a same profile use the same server. This can be similar to what
Willy Tarreau294d0f02015-08-10 19:40:12 +020014237 "appsession" did with the "request-learn" statement, but with support for
Willy Tarreau74ca5042013-06-11 23:12:07 +020014238 multi-peer synchronization and state keeping across restarts. If no name is
14239 specified, the first cookie value is returned. This fetch should not be used
14240 anymore and should be replaced by req.cook() or res.cook() instead as it
14241 ambiguously uses the direction based on the context where it is used.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014242
Willy Tarreau74ca5042013-06-11 23:12:07 +020014243hdr([<name>[,<occ>]]) : string
14244 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
14245 used on responses. Please refer to these respective fetches for more details.
14246 In case of doubt about the fetch direction, please use the explicit ones.
14247 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014248 unambiguously apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014249
Willy Tarreau74ca5042013-06-11 23:12:07 +020014250req.fhdr(<name>[,<occ>]) : string
14251 This extracts the last occurrence of header <name> in an HTTP request. When
14252 used from an ACL, all occurrences are iterated over until a match is found.
14253 Optionally, a specific occurrence might be specified as a position number.
14254 Positive values indicate a position from the first occurrence, with 1 being
14255 the first one. Negative values indicate positions relative to the last one,
14256 with -1 being the last one. It differs from req.hdr() in that any commas
14257 present in the value are returned and are not used as delimiters. This is
14258 sometimes useful with headers such as User-Agent.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014259
Willy Tarreau74ca5042013-06-11 23:12:07 +020014260req.fhdr_cnt([<name>]) : integer
14261 Returns an integer value representing the number of occurrences of request
14262 header field name <name>, or the total number of header fields if <name> is
14263 not specified. Contrary to its req.hdr_cnt() cousin, this function returns
14264 the number of full line headers and does not stop on commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014265
Willy Tarreau74ca5042013-06-11 23:12:07 +020014266req.hdr([<name>[,<occ>]]) : string
14267 This extracts the last occurrence of header <name> in an HTTP request. When
14268 used from an ACL, all occurrences are iterated over until a match is found.
14269 Optionally, a specific occurrence might be specified as a position number.
14270 Positive values indicate a position from the first occurrence, with 1 being
14271 the first one. Negative values indicate positions relative to the last one,
14272 with -1 being the last one. A typical use is with the X-Forwarded-For header
14273 once converted to IP, associated with an IP stick-table. The function
14274 considers any comma as a delimiter for distinct values. If full-line headers
14275 are desired instead, use req.fhdr(). Please carefully check RFC2616 to know
14276 how certain headers are supposed to be parsed. Also, some of them are case
14277 insensitive (eg: Connection).
Willy Tarreaud63335a2010-02-26 12:56:52 +010014278
Willy Tarreau74ca5042013-06-11 23:12:07 +020014279 ACL derivatives :
14280 hdr([<name>[,<occ>]]) : exact string match
14281 hdr_beg([<name>[,<occ>]]) : prefix match
14282 hdr_dir([<name>[,<occ>]]) : subdir match
14283 hdr_dom([<name>[,<occ>]]) : domain match
14284 hdr_end([<name>[,<occ>]]) : suffix match
14285 hdr_len([<name>[,<occ>]]) : length match
14286 hdr_reg([<name>[,<occ>]]) : regex match
14287 hdr_sub([<name>[,<occ>]]) : substring match
14288
14289req.hdr_cnt([<name>]) : integer
14290hdr_cnt([<header>]) : integer (deprecated)
14291 Returns an integer value representing the number of occurrences of request
14292 header field name <name>, or the total number of header field values if
14293 <name> is not specified. It is important to remember that one header line may
14294 count as several headers if it has several values. The function considers any
14295 comma as a delimiter for distinct values. If full-line headers are desired
14296 instead, req.fhdr_cnt() should be used instead. With ACLs, it can be used to
14297 detect presence, absence or abuse of a specific header, as well as to block
14298 request smuggling attacks by rejecting requests which contain more than one
14299 of certain headers. See "req.hdr" for more information on header matching.
14300
14301req.hdr_ip([<name>[,<occ>]]) : ip
14302hdr_ip([<name>[,<occ>]]) : ip (deprecated)
14303 This extracts the last occurrence of header <name> in an HTTP request,
14304 converts it to an IPv4 or IPv6 address and returns this address. When used
14305 with ACLs, all occurrences are checked, and if <name> is omitted, every value
14306 of every header is checked. Optionally, a specific occurrence might be
14307 specified as a position number. Positive values indicate a position from the
14308 first occurrence, with 1 being the first one. Negative values indicate
14309 positions relative to the last one, with -1 being the last one. A typical use
14310 is with the X-Forwarded-For and X-Client-IP headers.
14311
14312req.hdr_val([<name>[,<occ>]]) : integer
14313hdr_val([<name>[,<occ>]]) : integer (deprecated)
14314 This extracts the last occurrence of header <name> in an HTTP request, and
14315 converts it to an integer value. When used with ACLs, all occurrences are
14316 checked, and if <name> is omitted, every value of every header is checked.
14317 Optionally, a specific occurrence might be specified as a position number.
14318 Positive values indicate a position from the first occurrence, with 1 being
14319 the first one. Negative values indicate positions relative to the last one,
14320 with -1 being the last one. A typical use is with the X-Forwarded-For header.
14321
14322http_auth(<userlist>) : boolean
14323 Returns a boolean indicating whether the authentication data received from
14324 the client match a username & password stored in the specified userlist. This
14325 fetch function is not really useful outside of ACLs. Currently only http
14326 basic auth is supported.
14327
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010014328http_auth_group(<userlist>) : string
14329 Returns a string corresponding to the user name found in the authentication
14330 data received from the client if both the user name and password are valid
14331 according to the specified userlist. The main purpose is to use it in ACLs
14332 where it is then checked whether the user belongs to any group within a list.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014333 This fetch function is not really useful outside of ACLs. Currently only http
14334 basic auth is supported.
14335
14336 ACL derivatives :
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010014337 http_auth_group(<userlist>) : group ...
14338 Returns true when the user extracted from the request and whose password is
14339 valid according to the specified userlist belongs to at least one of the
14340 groups.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014341
14342http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020014343 Returns true when the request being processed is the first one of the
14344 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020014345 from some requests when a request is not the first one, or to help grouping
14346 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020014347
Willy Tarreau74ca5042013-06-11 23:12:07 +020014348method : integer + string
14349 Returns an integer value corresponding to the method in the HTTP request. For
14350 example, "GET" equals 1 (check sources to establish the matching). Value 9
14351 means "other method" and may be converted to a string extracted from the
14352 stream. This should not be used directly as a sample, this is only meant to
14353 be used from ACLs, which transparently convert methods from patterns to these
14354 integer + string values. Some predefined ACL already check for most common
14355 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014356
Willy Tarreau74ca5042013-06-11 23:12:07 +020014357 ACL derivatives :
14358 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020014359
Willy Tarreau74ca5042013-06-11 23:12:07 +020014360 Example :
14361 # only accept GET and HEAD requests
14362 acl valid_method method GET HEAD
14363 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020014364
Willy Tarreau74ca5042013-06-11 23:12:07 +020014365path : string
14366 This extracts the request's URL path, which starts at the first slash and
14367 ends before the question mark (without the host part). A typical use is with
14368 prefetch-capable caches, and with portals which need to aggregate multiple
14369 information from databases and keep them in caches. Note that with outgoing
14370 caches, it would be wiser to use "url" instead. With ACLs, it's typically
14371 used to match exact file names (eg: "/login.php"), or directory parts using
14372 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014373
Willy Tarreau74ca5042013-06-11 23:12:07 +020014374 ACL derivatives :
14375 path : exact string match
14376 path_beg : prefix match
14377 path_dir : subdir match
14378 path_dom : domain match
14379 path_end : suffix match
14380 path_len : length match
14381 path_reg : regex match
14382 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020014383
Willy Tarreau49ad95c2015-01-19 15:06:26 +010014384query : string
14385 This extracts the request's query string, which starts after the first
14386 question mark. If no question mark is present, this fetch returns nothing. If
14387 a question mark is present but nothing follows, it returns an empty string.
14388 This means it's possible to easily know whether a query string is present
14389 using the "found" matching method. This fetch is the completemnt of "path"
14390 which stops before the question mark.
14391
Willy Tarreaueb27ec72015-02-20 13:55:29 +010014392req.hdr_names([<delim>]) : string
14393 This builds a string made from the concatenation of all header names as they
14394 appear in the request when the rule is evaluated. The default delimiter is
14395 the comma (',') but it may be overridden as an optional argument <delim>. In
14396 this case, only the first character of <delim> is considered.
14397
Willy Tarreau74ca5042013-06-11 23:12:07 +020014398req.ver : string
14399req_ver : string (deprecated)
14400 Returns the version string from the HTTP request, for example "1.1". This can
14401 be useful for logs, but is mostly there for ACL. Some predefined ACL already
14402 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014403
Willy Tarreau74ca5042013-06-11 23:12:07 +020014404 ACL derivatives :
14405 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020014406
Willy Tarreau74ca5042013-06-11 23:12:07 +020014407res.comp : boolean
14408 Returns the boolean "true" value if the response has been compressed by
14409 HAProxy, otherwise returns boolean "false". This may be used to add
14410 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014411
Willy Tarreau74ca5042013-06-11 23:12:07 +020014412res.comp_algo : string
14413 Returns a string containing the name of the algorithm used if the response
14414 was compressed by HAProxy, for example : "deflate". This may be used to add
14415 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014416
Willy Tarreau74ca5042013-06-11 23:12:07 +020014417res.cook([<name>]) : string
14418scook([<name>]) : string (deprecated)
14419 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
14420 header line from the response, and returns its value as string. If no name is
14421 specified, the first cookie value is returned.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020014422
Willy Tarreau74ca5042013-06-11 23:12:07 +020014423 ACL derivatives :
14424 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020014425
Willy Tarreau74ca5042013-06-11 23:12:07 +020014426res.cook_cnt([<name>]) : integer
14427scook_cnt([<name>]) : integer (deprecated)
14428 Returns an integer value representing the number of occurrences of the cookie
14429 <name> in the response, or all cookies if <name> is not specified. This is
14430 mostly useful when combined with ACLs to detect suspicious responses.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014431
Willy Tarreau74ca5042013-06-11 23:12:07 +020014432res.cook_val([<name>]) : integer
14433scook_val([<name>]) : integer (deprecated)
14434 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
14435 header line from the response, and converts its value to an integer which is
14436 returned. If no name is specified, the first cookie value is returned.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014437
Willy Tarreau74ca5042013-06-11 23:12:07 +020014438res.fhdr([<name>[,<occ>]]) : string
14439 This extracts the last occurrence of header <name> in an HTTP response, or of
14440 the last header if no <name> is specified. Optionally, a specific occurrence
14441 might be specified as a position number. Positive values indicate a position
14442 from the first occurrence, with 1 being the first one. Negative values
14443 indicate positions relative to the last one, with -1 being the last one. It
14444 differs from res.hdr() in that any commas present in the value are returned
14445 and are not used as delimiters. If this is not desired, the res.hdr() fetch
14446 should be used instead. This is sometimes useful with headers such as Date or
14447 Expires.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014448
Willy Tarreau74ca5042013-06-11 23:12:07 +020014449res.fhdr_cnt([<name>]) : integer
14450 Returns an integer value representing the number of occurrences of response
14451 header field name <name>, or the total number of header fields if <name> is
14452 not specified. Contrary to its res.hdr_cnt() cousin, this function returns
14453 the number of full line headers and does not stop on commas. If this is not
14454 desired, the res.hdr_cnt() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014455
Willy Tarreau74ca5042013-06-11 23:12:07 +020014456res.hdr([<name>[,<occ>]]) : string
14457shdr([<name>[,<occ>]]) : string (deprecated)
14458 This extracts the last occurrence of header <name> in an HTTP response, or of
14459 the last header if no <name> is specified. Optionally, a specific occurrence
14460 might be specified as a position number. Positive values indicate a position
14461 from the first occurrence, with 1 being the first one. Negative values
14462 indicate positions relative to the last one, with -1 being the last one. This
14463 can be useful to learn some data into a stick-table. The function considers
14464 any comma as a delimiter for distinct values. If this is not desired, the
14465 res.fhdr() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014466
Willy Tarreau74ca5042013-06-11 23:12:07 +020014467 ACL derivatives :
14468 shdr([<name>[,<occ>]]) : exact string match
14469 shdr_beg([<name>[,<occ>]]) : prefix match
14470 shdr_dir([<name>[,<occ>]]) : subdir match
14471 shdr_dom([<name>[,<occ>]]) : domain match
14472 shdr_end([<name>[,<occ>]]) : suffix match
14473 shdr_len([<name>[,<occ>]]) : length match
14474 shdr_reg([<name>[,<occ>]]) : regex match
14475 shdr_sub([<name>[,<occ>]]) : substring match
14476
14477res.hdr_cnt([<name>]) : integer
14478shdr_cnt([<name>]) : integer (deprecated)
14479 Returns an integer value representing the number of occurrences of response
14480 header field name <name>, or the total number of header fields if <name> is
14481 not specified. The function considers any comma as a delimiter for distinct
14482 values. If this is not desired, the res.fhdr_cnt() fetch should be used
14483 instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014484
Willy Tarreau74ca5042013-06-11 23:12:07 +020014485res.hdr_ip([<name>[,<occ>]]) : ip
14486shdr_ip([<name>[,<occ>]]) : ip (deprecated)
14487 This extracts the last occurrence of header <name> in an HTTP response,
14488 convert it to an IPv4 or IPv6 address and returns this address. Optionally, a
14489 specific occurrence might be specified as a position number. Positive values
14490 indicate a position from the first occurrence, with 1 being the first one.
14491 Negative values indicate positions relative to the last one, with -1 being
14492 the last one. This can be useful to learn some data into a stick table.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014493
Willy Tarreaueb27ec72015-02-20 13:55:29 +010014494res.hdr_names([<delim>]) : string
14495 This builds a string made from the concatenation of all header names as they
14496 appear in the response when the rule is evaluated. The default delimiter is
14497 the comma (',') but it may be overridden as an optional argument <delim>. In
14498 this case, only the first character of <delim> is considered.
14499
Willy Tarreau74ca5042013-06-11 23:12:07 +020014500res.hdr_val([<name>[,<occ>]]) : integer
14501shdr_val([<name>[,<occ>]]) : integer (deprecated)
14502 This extracts the last occurrence of header <name> in an HTTP response, and
14503 converts it to an integer value. Optionally, a specific occurrence might be
14504 specified as a position number. Positive values indicate a position from the
14505 first occurrence, with 1 being the first one. Negative values indicate
14506 positions relative to the last one, with -1 being the last one. This can be
14507 useful to learn some data into a stick table.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010014508
Willy Tarreau74ca5042013-06-11 23:12:07 +020014509res.ver : string
14510resp_ver : string (deprecated)
14511 Returns the version string from the HTTP response, for example "1.1". This
14512 can be useful for logs, but is mostly there for ACL.
Willy Tarreau0e698542011-09-16 08:32:32 +020014513
Willy Tarreau74ca5042013-06-11 23:12:07 +020014514 ACL derivatives :
14515 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010014516
Willy Tarreau74ca5042013-06-11 23:12:07 +020014517set-cookie([<name>]) : string (deprecated)
14518 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
14519 header line from the response and uses the corresponding value to match. This
Willy Tarreau294d0f02015-08-10 19:40:12 +020014520 can be comparable to what "appsession" did with default options, but with
Willy Tarreau74ca5042013-06-11 23:12:07 +020014521 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010014522
Willy Tarreau74ca5042013-06-11 23:12:07 +020014523 This fetch function is deprecated and has been superseded by the "res.cook"
14524 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010014525
Willy Tarreau74ca5042013-06-11 23:12:07 +020014526status : integer
14527 Returns an integer containing the HTTP status code in the HTTP response, for
14528 example, 302. It is mostly used within ACLs and integer ranges, for example,
14529 to remove any Location header if the response is not a 3xx.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014530
Thierry Fournier0e00dca2016-04-07 15:47:40 +020014531unique-id : string
14532 Returns the unique-id attached to the request. The directive
14533 "unique-id-format" must be set. If it is not set, the unique-id sample fetch
14534 fails. Note that the unique-id is usually used with HTTP requests, however this
14535 sample fetch can be used with other protocols. Obviously, if it is used with
14536 other protocols than HTTP, the unique-id-format directive must not contain
14537 HTTP parts. See: unique-id-format and unique-id-header
14538
Willy Tarreau74ca5042013-06-11 23:12:07 +020014539url : string
14540 This extracts the request's URL as presented in the request. A typical use is
14541 with prefetch-capable caches, and with portals which need to aggregate
14542 multiple information from databases and keep them in caches. With ACLs, using
14543 "path" is preferred over using "url", because clients may send a full URL as
14544 is normally done with proxies. The only real use is to match "*" which does
14545 not match in "path", and for which there is already a predefined ACL. See
14546 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014547
Willy Tarreau74ca5042013-06-11 23:12:07 +020014548 ACL derivatives :
14549 url : exact string match
14550 url_beg : prefix match
14551 url_dir : subdir match
14552 url_dom : domain match
14553 url_end : suffix match
14554 url_len : length match
14555 url_reg : regex match
14556 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014557
Willy Tarreau74ca5042013-06-11 23:12:07 +020014558url_ip : ip
14559 This extracts the IP address from the request's URL when the host part is
14560 presented as an IP address. Its use is very limited. For instance, a
14561 monitoring system might use this field as an alternative for the source IP in
14562 order to test what path a given source address would follow, or to force an
14563 entry in a table for a given source address. With ACLs it can be used to
14564 restrict access to certain systems through a proxy, for example when combined
14565 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014566
Willy Tarreau74ca5042013-06-11 23:12:07 +020014567url_port : integer
14568 This extracts the port part from the request's URL. Note that if the port is
14569 not specified in the request, port 80 is assumed. With ACLs it can be used to
14570 restrict access to certain systems through a proxy, for example when combined
14571 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014572
Willy Tarreau1ede1da2015-05-07 16:06:18 +020014573urlp([<name>[,<delim>]]) : string
14574url_param([<name>[,<delim>]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020014575 This extracts the first occurrence of the parameter <name> in the query
14576 string, which begins after either '?' or <delim>, and which ends before '&',
Willy Tarreau1ede1da2015-05-07 16:06:18 +020014577 ';' or <delim>. The parameter name is case-sensitive. If no name is given,
14578 any parameter will match, and the first one will be returned. The result is
14579 a string corresponding to the value of the parameter <name> as presented in
14580 the request (no URL decoding is performed). This can be used for session
Willy Tarreau74ca5042013-06-11 23:12:07 +020014581 stickiness based on a client ID, to extract an application cookie passed as a
14582 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
Willy Tarreau1ede1da2015-05-07 16:06:18 +020014583 this fetch iterates over multiple parameters and will iteratively report all
14584 parameters values if no name is given
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014585
Willy Tarreau74ca5042013-06-11 23:12:07 +020014586 ACL derivatives :
14587 urlp(<name>[,<delim>]) : exact string match
14588 urlp_beg(<name>[,<delim>]) : prefix match
14589 urlp_dir(<name>[,<delim>]) : subdir match
14590 urlp_dom(<name>[,<delim>]) : domain match
14591 urlp_end(<name>[,<delim>]) : suffix match
14592 urlp_len(<name>[,<delim>]) : length match
14593 urlp_reg(<name>[,<delim>]) : regex match
14594 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014595
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014596
Willy Tarreau74ca5042013-06-11 23:12:07 +020014597 Example :
14598 # match http://example.com/foo?PHPSESSIONID=some_id
14599 stick on urlp(PHPSESSIONID)
14600 # match http://example.com/foo;JSESSIONID=some_id
14601 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020014602
Willy Tarreau1ede1da2015-05-07 16:06:18 +020014603urlp_val([<name>[,<delim>])] : integer
Willy Tarreau74ca5042013-06-11 23:12:07 +020014604 See "urlp" above. This one extracts the URL parameter <name> in the request
14605 and converts it to an integer value. This can be used for session stickiness
14606 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020014607
Dragan Dosen0070cd52016-06-16 12:19:49 +020014608url32 : integer
14609 This returns a 32-bit hash of the value obtained by concatenating the first
14610 Host header and the whole URL including parameters (not only the path part of
14611 the request, as in the "base32" fetch above). This is useful to track per-URL
14612 activity. A shorter hash is stored, saving a lot of memory. The output type
14613 is an unsigned integer.
14614
14615url32+src : binary
14616 This returns the concatenation of the "url32" fetch and the "src" fetch. The
14617 resulting type is of type binary, with a size of 8 or 20 bytes depending on
14618 the source address family. This can be used to track per-IP, per-URL counters.
14619
Willy Tarreau198a7442008-01-17 12:05:32 +010014620
Willy Tarreau74ca5042013-06-11 23:12:07 +0200146217.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014622---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010014623
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014624Some predefined ACLs are hard-coded so that they do not have to be declared in
14625every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020014626order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010014627
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014628ACL name Equivalent to Usage
14629---------------+-----------------------------+---------------------------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014630FALSE always_false never match
Willy Tarreau2492d5b2009-07-11 00:06:00 +020014631HTTP req_proto_http match if protocol is valid HTTP
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014632HTTP_1.0 req_ver 1.0 match HTTP version 1.0
14633HTTP_1.1 req_ver 1.1 match HTTP version 1.1
Willy Tarreaud63335a2010-02-26 12:56:52 +010014634HTTP_CONTENT hdr_val(content-length) gt 0 match an existing content-length
14635HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
14636HTTP_URL_SLASH url_beg / match URL beginning with "/"
14637HTTP_URL_STAR url * match URL equal to "*"
14638LOCALHOST src 127.0.0.1/8 match connection from local host
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014639METH_CONNECT method CONNECT match HTTP CONNECT method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020014640METH_DELETE method DELETE match HTTP DELETE method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014641METH_GET method GET HEAD match HTTP GET or HEAD method
14642METH_HEAD method HEAD match HTTP HEAD method
14643METH_OPTIONS method OPTIONS match HTTP OPTIONS method
14644METH_POST method POST match HTTP POST method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020014645METH_PUT method PUT match HTTP PUT method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014646METH_TRACE method TRACE match HTTP TRACE method
Emeric Brunbede3d02009-06-30 17:54:00 +020014647RDP_COOKIE req_rdp_cookie_cnt gt 0 match presence of an RDP cookie
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014648REQ_CONTENT req_len gt 0 match data in the request buffer
Willy Tarreaud63335a2010-02-26 12:56:52 +010014649TRUE always_true always match
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014650WAIT_END wait_end wait for end of content analysis
14651---------------+-----------------------------+---------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010014652
Willy Tarreaub937b7e2010-01-12 15:27:54 +010014653
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200146548. Logging
14655----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010014656
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014657One of HAProxy's strong points certainly lies is its precise logs. It probably
14658provides the finest level of information available for such a product, which is
14659very important for troubleshooting complex environments. Standard information
14660provided in logs include client ports, TCP/HTTP state timers, precise session
14661state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010014662to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014663headers.
14664
14665In order to improve administrators reactivity, it offers a great transparency
14666about encountered problems, both internal and external, and it is possible to
14667send logs to different sources at the same time with different level filters :
14668
14669 - global process-level logs (system errors, start/stop, etc..)
14670 - per-instance system and internal errors (lack of resource, bugs, ...)
14671 - per-instance external troubles (servers up/down, max connections)
14672 - per-instance activity (client connections), either at the establishment or
14673 at the termination.
Jim Freeman9e8714b2015-05-26 09:16:34 -060014674 - per-request control of log-level, eg:
14675 http-request set-log-level silent if sensitive_request
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014676
14677The ability to distribute different levels of logs to different log servers
14678allow several production teams to interact and to fix their problems as soon
14679as possible. For example, the system team might monitor system-wide errors,
14680while the application team might be monitoring the up/down for their servers in
14681real time, and the security team might analyze the activity logs with one hour
14682delay.
14683
14684
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200146858.1. Log levels
14686---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014687
Simon Hormandf791f52011-05-29 15:01:10 +090014688TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014689source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090014690HTTP request, HTTP return code, number of bytes transmitted, conditions
14691in which the session ended, and even exchanged cookies values. For example
14692track a particular user's problems. All messages may be sent to up to two
14693syslog servers. Check the "log" keyword in section 4.2 for more information
14694about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014695
14696
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200146978.2. Log formats
14698----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014699
William Lallemand48940402012-01-30 16:47:22 +010014700HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090014701and will be detailed in the following sections. A few of them may vary
14702slightly with the configuration, due to indicators specific to certain
14703options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014704
14705 - the default format, which is very basic and very rarely used. It only
14706 provides very basic information about the incoming connection at the moment
14707 it is accepted : source IP:port, destination IP:port, and frontend-name.
14708 This mode will eventually disappear so it will not be described to great
14709 extents.
14710
14711 - the TCP format, which is more advanced. This format is enabled when "option
14712 tcplog" is set on the frontend. HAProxy will then usually wait for the
14713 connection to terminate before logging. This format provides much richer
14714 information, such as timers, connection counts, queue size, etc... This
14715 format is recommended for pure TCP proxies.
14716
14717 - the HTTP format, which is the most advanced for HTTP proxying. This format
14718 is enabled when "option httplog" is set on the frontend. It provides the
14719 same information as the TCP format with some HTTP-specific fields such as
14720 the request, the status code, and captures of headers and cookies. This
14721 format is recommended for HTTP proxies.
14722
Emeric Brun3a058f32009-06-30 18:26:00 +020014723 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
14724 fields arranged in the same order as the CLF format. In this mode, all
14725 timers, captures, flags, etc... appear one per field after the end of the
14726 common fields, in the same order they appear in the standard HTTP format.
14727
William Lallemand48940402012-01-30 16:47:22 +010014728 - the custom log format, allows you to make your own log line.
14729
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014730Next sections will go deeper into details for each of these formats. Format
14731specification will be performed on a "field" basis. Unless stated otherwise, a
14732field is a portion of text delimited by any number of spaces. Since syslog
14733servers are susceptible of inserting fields at the beginning of a line, it is
14734always assumed that the first field is the one containing the process name and
14735identifier.
14736
14737Note : Since log lines may be quite long, the log examples in sections below
14738 might be broken into multiple lines. The example log lines will be
14739 prefixed with 3 closing angle brackets ('>>>') and each time a log is
14740 broken into multiple lines, each non-final line will end with a
14741 backslash ('\') and the next line will start indented by two characters.
14742
14743
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200147448.2.1. Default log format
14745-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014746
14747This format is used when no specific option is set. The log is emitted as soon
14748as the connection is accepted. One should note that this currently is the only
14749format which logs the request's destination IP and ports.
14750
14751 Example :
14752 listen www
14753 mode http
14754 log global
14755 server srv1 127.0.0.1:8000
14756
14757 >>> Feb 6 12:12:09 localhost \
14758 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
14759 (www/HTTP)
14760
14761 Field Format Extract from the example above
14762 1 process_name '[' pid ']:' haproxy[14385]:
14763 2 'Connect from' Connect from
14764 3 source_ip ':' source_port 10.0.1.2:33312
14765 4 'to' to
14766 5 destination_ip ':' destination_port 10.0.3.31:8012
14767 6 '(' frontend_name '/' mode ')' (www/HTTP)
14768
14769Detailed fields description :
14770 - "source_ip" is the IP address of the client which initiated the connection.
14771 - "source_port" is the TCP port of the client which initiated the connection.
14772 - "destination_ip" is the IP address the client connected to.
14773 - "destination_port" is the TCP port the client connected to.
14774 - "frontend_name" is the name of the frontend (or listener) which received
14775 and processed the connection.
14776 - "mode is the mode the frontend is operating (TCP or HTTP).
14777
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014778In case of a UNIX socket, the source and destination addresses are marked as
14779"unix:" and the ports reflect the internal ID of the socket which accepted the
14780connection (the same ID as reported in the stats).
14781
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014782It is advised not to use this deprecated format for newer installations as it
14783will eventually disappear.
14784
14785
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200147868.2.2. TCP log format
14787---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014788
14789The TCP format is used when "option tcplog" is specified in the frontend, and
14790is the recommended format for pure TCP proxies. It provides a lot of precious
14791information for troubleshooting. Since this format includes timers and byte
14792counts, the log is normally emitted at the end of the session. It can be
14793emitted earlier if "option logasap" is specified, which makes sense in most
14794environments with long sessions such as remote terminals. Sessions which match
14795the "monitor" rules are never logged. It is also possible not to emit logs for
14796sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020014797specifying "option dontlognull" in the frontend. Successful connections will
14798not be logged if "option dontlog-normal" is specified in the frontend. A few
14799fields may slightly vary depending on some configuration options, those are
14800marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014801
14802 Example :
14803 frontend fnt
14804 mode tcp
14805 option tcplog
14806 log global
14807 default_backend bck
14808
14809 backend bck
14810 server srv1 127.0.0.1:8000
14811
14812 >>> Feb 6 12:12:56 localhost \
14813 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
14814 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
14815
14816 Field Format Extract from the example above
14817 1 process_name '[' pid ']:' haproxy[14387]:
14818 2 client_ip ':' client_port 10.0.1.2:33313
14819 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
14820 4 frontend_name fnt
14821 5 backend_name '/' server_name bck/srv1
14822 6 Tw '/' Tc '/' Tt* 0/0/5007
14823 7 bytes_read* 212
14824 8 termination_state --
14825 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
14826 10 srv_queue '/' backend_queue 0/0
14827
14828Detailed fields description :
14829 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014830 connection to haproxy. If the connection was accepted on a UNIX socket
14831 instead, the IP address would be replaced with the word "unix". Note that
14832 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010014833 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
14834 and the NetScaler Client IP insetion protocol is correctly used, then the
14835 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014836
14837 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014838 If the connection was accepted on a UNIX socket instead, the port would be
14839 replaced with the ID of the accepting socket, which is also reported in the
14840 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014841
14842 - "accept_date" is the exact date when the connection was received by haproxy
14843 (which might be very slightly different from the date observed on the
14844 network if there was some queuing in the system's backlog). This is usually
14845 the same date which may appear in any upstream firewall's log.
14846
14847 - "frontend_name" is the name of the frontend (or listener) which received
14848 and processed the connection.
14849
14850 - "backend_name" is the name of the backend (or listener) which was selected
14851 to manage the connection to the server. This will be the same as the
14852 frontend if no switching rule has been applied, which is common for TCP
14853 applications.
14854
14855 - "server_name" is the name of the last server to which the connection was
14856 sent, which might differ from the first one if there were connection errors
14857 and a redispatch occurred. Note that this server belongs to the backend
14858 which processed the request. If the connection was aborted before reaching
14859 a server, "<NOSRV>" is indicated instead of a server name.
14860
14861 - "Tw" is the total time in milliseconds spent waiting in the various queues.
14862 It can be "-1" if the connection was aborted before reaching the queue.
14863 See "Timers" below for more details.
14864
14865 - "Tc" is the total time in milliseconds spent waiting for the connection to
14866 establish to the final server, including retries. It can be "-1" if the
14867 connection was aborted before a connection could be established. See
14868 "Timers" below for more details.
14869
14870 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014871 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014872 "option logasap" was specified, then the time counting stops at the moment
14873 the log is emitted. In this case, a '+' sign is prepended before the value,
14874 indicating that the final one will be larger. See "Timers" below for more
14875 details.
14876
14877 - "bytes_read" is the total number of bytes transmitted from the server to
14878 the client when the log is emitted. If "option logasap" is specified, the
14879 this value will be prefixed with a '+' sign indicating that the final one
14880 may be larger. Please note that this value is a 64-bit counter, so log
14881 analysis tools must be able to handle it without overflowing.
14882
14883 - "termination_state" is the condition the session was in when the session
14884 ended. This indicates the session state, which side caused the end of
14885 session to happen, and for what reason (timeout, error, ...). The normal
14886 flags should be "--", indicating the session was closed by either end with
14887 no data remaining in buffers. See below "Session state at disconnection"
14888 for more details.
14889
14890 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014891 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014892 limits have been reached. For instance, if actconn is close to 512 when
14893 multiple connection errors occur, chances are high that the system limits
14894 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014895 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014896
14897 - "feconn" is the total number of concurrent connections on the frontend when
14898 the session was logged. It is useful to estimate the amount of resource
14899 required to sustain high loads, and to detect when the frontend's "maxconn"
14900 has been reached. Most often when this value increases by huge jumps, it is
14901 because there is congestion on the backend servers, but sometimes it can be
14902 caused by a denial of service attack.
14903
14904 - "beconn" is the total number of concurrent connections handled by the
14905 backend when the session was logged. It includes the total number of
14906 concurrent connections active on servers as well as the number of
14907 connections pending in queues. It is useful to estimate the amount of
14908 additional servers needed to support high loads for a given application.
14909 Most often when this value increases by huge jumps, it is because there is
14910 congestion on the backend servers, but sometimes it can be caused by a
14911 denial of service attack.
14912
14913 - "srv_conn" is the total number of concurrent connections still active on
14914 the server when the session was logged. It can never exceed the server's
14915 configured "maxconn" parameter. If this value is very often close or equal
14916 to the server's "maxconn", it means that traffic regulation is involved a
14917 lot, meaning that either the server's maxconn value is too low, or that
14918 there aren't enough servers to process the load with an optimal response
14919 time. When only one of the server's "srv_conn" is high, it usually means
14920 that this server has some trouble causing the connections to take longer to
14921 be processed than on other servers.
14922
14923 - "retries" is the number of connection retries experienced by this session
14924 when trying to connect to the server. It must normally be zero, unless a
14925 server is being stopped at the same moment the connection was attempted.
14926 Frequent retries generally indicate either a network problem between
14927 haproxy and the server, or a misconfigured system backlog on the server
14928 preventing new connections from being queued. This field may optionally be
14929 prefixed with a '+' sign, indicating that the session has experienced a
14930 redispatch after the maximal retry count has been reached on the initial
14931 server. In this case, the server name appearing in the log is the one the
14932 connection was redispatched to, and not the first one, though both may
14933 sometimes be the same in case of hashing for instance. So as a general rule
14934 of thumb, when a '+' is present in front of the retry count, this count
14935 should not be attributed to the logged server.
14936
14937 - "srv_queue" is the total number of requests which were processed before
14938 this one in the server queue. It is zero when the request has not gone
14939 through the server queue. It makes it possible to estimate the approximate
14940 server's response time by dividing the time spent in queue by the number of
14941 requests in the queue. It is worth noting that if a session experiences a
14942 redispatch and passes through two server queues, their positions will be
14943 cumulated. A request should not pass through both the server queue and the
14944 backend queue unless a redispatch occurs.
14945
14946 - "backend_queue" is the total number of requests which were processed before
14947 this one in the backend's global queue. It is zero when the request has not
14948 gone through the global queue. It makes it possible to estimate the average
14949 queue length, which easily translates into a number of missing servers when
14950 divided by a server's "maxconn" parameter. It is worth noting that if a
14951 session experiences a redispatch, it may pass twice in the backend's queue,
14952 and then both positions will be cumulated. A request should not pass
14953 through both the server queue and the backend queue unless a redispatch
14954 occurs.
14955
14956
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200149578.2.3. HTTP log format
14958----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014959
14960The HTTP format is the most complete and the best suited for HTTP proxies. It
14961is enabled by when "option httplog" is specified in the frontend. It provides
14962the same level of information as the TCP format with additional features which
14963are specific to the HTTP protocol. Just like the TCP format, the log is usually
14964emitted at the end of the session, unless "option logasap" is specified, which
14965generally only makes sense for download sites. A session which matches the
14966"monitor" rules will never logged. It is also possible not to log sessions for
14967which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020014968frontend. Successful connections will not be logged if "option dontlog-normal"
14969is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014970
14971Most fields are shared with the TCP log, some being different. A few fields may
14972slightly vary depending on some configuration options. Those ones are marked
14973with a star ('*') after the field name below.
14974
14975 Example :
14976 frontend http-in
14977 mode http
14978 option httplog
14979 log global
14980 default_backend bck
14981
14982 backend static
14983 server srv1 127.0.0.1:8000
14984
14985 >>> Feb 6 12:14:14 localhost \
14986 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
14987 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 +010014988 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014989
14990 Field Format Extract from the example above
14991 1 process_name '[' pid ']:' haproxy[14389]:
14992 2 client_ip ':' client_port 10.0.1.2:33317
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020014993 3 '[' request_date ']' [06/Feb/2009:12:14:14.655]
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014994 4 frontend_name http-in
14995 5 backend_name '/' server_name static/srv1
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020014996 6 TR '/' Tw '/' Tc '/' Tr '/' Ta* 10/0/30/69/109
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014997 7 status_code 200
14998 8 bytes_read* 2750
14999 9 captured_request_cookie -
15000 10 captured_response_cookie -
15001 11 termination_state ----
15002 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
15003 13 srv_queue '/' backend_queue 0/0
15004 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
15005 15 '{' captured_response_headers* '}' {}
15006 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010015007
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015008Detailed fields description :
15009 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010015010 connection to haproxy. If the connection was accepted on a UNIX socket
15011 instead, the IP address would be replaced with the word "unix". Note that
15012 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010015013 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
15014 and the NetScaler Client IP insetion protocol is correctly used, then the
15015 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015016
15017 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010015018 If the connection was accepted on a UNIX socket instead, the port would be
15019 replaced with the ID of the accepting socket, which is also reported in the
15020 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015021
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015022 - "request_date" is the exact date when the first byte of the HTTP request
15023 was received by haproxy (log field %tr).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015024
15025 - "frontend_name" is the name of the frontend (or listener) which received
15026 and processed the connection.
15027
15028 - "backend_name" is the name of the backend (or listener) which was selected
15029 to manage the connection to the server. This will be the same as the
15030 frontend if no switching rule has been applied.
15031
15032 - "server_name" is the name of the last server to which the connection was
15033 sent, which might differ from the first one if there were connection errors
15034 and a redispatch occurred. Note that this server belongs to the backend
15035 which processed the request. If the request was aborted before reaching a
15036 server, "<NOSRV>" is indicated instead of a server name. If the request was
15037 intercepted by the stats subsystem, "<STATS>" is indicated instead.
15038
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015039 - "TR" is the total time in milliseconds spent waiting for a full HTTP
15040 request from the client (not counting body) after the first byte was
15041 received. It can be "-1" if the connection was aborted before a complete
15042 request could be received or the a bad request was received. It should
15043 always be very small because a request generally fits in one single packet.
15044 Large times here generally indicate network issues between the client and
15045 haproxy or requests being typed by hand. See "Timers" below for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015046
15047 - "Tw" is the total time in milliseconds spent waiting in the various queues.
15048 It can be "-1" if the connection was aborted before reaching the queue.
15049 See "Timers" below for more details.
15050
15051 - "Tc" is the total time in milliseconds spent waiting for the connection to
15052 establish to the final server, including retries. It can be "-1" if the
15053 request was aborted before a connection could be established. See "Timers"
15054 below for more details.
15055
15056 - "Tr" is the total time in milliseconds spent waiting for the server to send
15057 a full HTTP response, not counting data. It can be "-1" if the request was
15058 aborted before a complete response could be received. It generally matches
15059 the server's processing time for the request, though it may be altered by
15060 the amount of data sent by the client to the server. Large times here on
15061 "GET" requests generally indicate an overloaded server. See "Timers" below
15062 for more details.
15063
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015064 - "Ta" is the time the request remained active in haproxy, which is the total
15065 time in milliseconds elapsed between the first byte of the request was
15066 received and the last byte of response was sent. It covers all possible
15067 processing except the handshake (see Th) and idle time (see Ti). There is
15068 one exception, if "option logasap" was specified, then the time counting
15069 stops at the moment the log is emitted. In this case, a '+' sign is
15070 prepended before the value, indicating that the final one will be larger.
15071 See "Timers" below for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015072
15073 - "status_code" is the HTTP status code returned to the client. This status
15074 is generally set by the server, but it might also be set by haproxy when
15075 the server cannot be reached or when its response is blocked by haproxy.
15076
15077 - "bytes_read" is the total number of bytes transmitted to the client when
15078 the log is emitted. This does include HTTP headers. If "option logasap" is
15079 specified, the this value will be prefixed with a '+' sign indicating that
15080 the final one may be larger. Please note that this value is a 64-bit
15081 counter, so log analysis tools must be able to handle it without
15082 overflowing.
15083
15084 - "captured_request_cookie" is an optional "name=value" entry indicating that
15085 the client had this cookie in the request. The cookie name and its maximum
15086 length are defined by the "capture cookie" statement in the frontend
15087 configuration. The field is a single dash ('-') when the option is not
15088 set. Only one cookie may be captured, it is generally used to track session
15089 ID exchanges between a client and a server to detect session crossing
15090 between clients due to application bugs. For more details, please consult
15091 the section "Capturing HTTP headers and cookies" below.
15092
15093 - "captured_response_cookie" is an optional "name=value" entry indicating
15094 that the server has returned a cookie with its response. The cookie name
15095 and its maximum length are defined by the "capture cookie" statement in the
15096 frontend configuration. The field is a single dash ('-') when the option is
15097 not set. Only one cookie may be captured, it is generally used to track
15098 session ID exchanges between a client and a server to detect session
15099 crossing between clients due to application bugs. For more details, please
15100 consult the section "Capturing HTTP headers and cookies" below.
15101
15102 - "termination_state" is the condition the session was in when the session
15103 ended. This indicates the session state, which side caused the end of
15104 session to happen, for what reason (timeout, error, ...), just like in TCP
15105 logs, and information about persistence operations on cookies in the last
15106 two characters. The normal flags should begin with "--", indicating the
15107 session was closed by either end with no data remaining in buffers. See
15108 below "Session state at disconnection" for more details.
15109
15110 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015111 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015112 limits have been reached. For instance, if actconn is close to 512 or 1024
15113 when multiple connection errors occur, chances are high that the system
15114 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015115 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015116 system.
15117
15118 - "feconn" is the total number of concurrent connections on the frontend when
15119 the session was logged. It is useful to estimate the amount of resource
15120 required to sustain high loads, and to detect when the frontend's "maxconn"
15121 has been reached. Most often when this value increases by huge jumps, it is
15122 because there is congestion on the backend servers, but sometimes it can be
15123 caused by a denial of service attack.
15124
15125 - "beconn" is the total number of concurrent connections handled by the
15126 backend when the session was logged. It includes the total number of
15127 concurrent connections active on servers as well as the number of
15128 connections pending in queues. It is useful to estimate the amount of
15129 additional servers needed to support high loads for a given application.
15130 Most often when this value increases by huge jumps, it is because there is
15131 congestion on the backend servers, but sometimes it can be caused by a
15132 denial of service attack.
15133
15134 - "srv_conn" is the total number of concurrent connections still active on
15135 the server when the session was logged. It can never exceed the server's
15136 configured "maxconn" parameter. If this value is very often close or equal
15137 to the server's "maxconn", it means that traffic regulation is involved a
15138 lot, meaning that either the server's maxconn value is too low, or that
15139 there aren't enough servers to process the load with an optimal response
15140 time. When only one of the server's "srv_conn" is high, it usually means
15141 that this server has some trouble causing the requests to take longer to be
15142 processed than on other servers.
15143
15144 - "retries" is the number of connection retries experienced by this session
15145 when trying to connect to the server. It must normally be zero, unless a
15146 server is being stopped at the same moment the connection was attempted.
15147 Frequent retries generally indicate either a network problem between
15148 haproxy and the server, or a misconfigured system backlog on the server
15149 preventing new connections from being queued. This field may optionally be
15150 prefixed with a '+' sign, indicating that the session has experienced a
15151 redispatch after the maximal retry count has been reached on the initial
15152 server. In this case, the server name appearing in the log is the one the
15153 connection was redispatched to, and not the first one, though both may
15154 sometimes be the same in case of hashing for instance. So as a general rule
15155 of thumb, when a '+' is present in front of the retry count, this count
15156 should not be attributed to the logged server.
15157
15158 - "srv_queue" is the total number of requests which were processed before
15159 this one in the server queue. It is zero when the request has not gone
15160 through the server queue. It makes it possible to estimate the approximate
15161 server's response time by dividing the time spent in queue by the number of
15162 requests in the queue. It is worth noting that if a session experiences a
15163 redispatch and passes through two server queues, their positions will be
15164 cumulated. A request should not pass through both the server queue and the
15165 backend queue unless a redispatch occurs.
15166
15167 - "backend_queue" is the total number of requests which were processed before
15168 this one in the backend's global queue. It is zero when the request has not
15169 gone through the global queue. It makes it possible to estimate the average
15170 queue length, which easily translates into a number of missing servers when
15171 divided by a server's "maxconn" parameter. It is worth noting that if a
15172 session experiences a redispatch, it may pass twice in the backend's queue,
15173 and then both positions will be cumulated. A request should not pass
15174 through both the server queue and the backend queue unless a redispatch
15175 occurs.
15176
15177 - "captured_request_headers" is a list of headers captured in the request due
15178 to the presence of the "capture request header" statement in the frontend.
15179 Multiple headers can be captured, they will be delimited by a vertical bar
15180 ('|'). When no capture is enabled, the braces do not appear, causing a
15181 shift of remaining fields. It is important to note that this field may
15182 contain spaces, and that using it requires a smarter log parser than when
15183 it's not used. Please consult the section "Capturing HTTP headers and
15184 cookies" below for more details.
15185
15186 - "captured_response_headers" is a list of headers captured in the response
15187 due to the presence of the "capture response header" statement in the
15188 frontend. Multiple headers can be captured, they will be delimited by a
15189 vertical bar ('|'). When no capture is enabled, the braces do not appear,
15190 causing a shift of remaining fields. It is important to note that this
15191 field may contain spaces, and that using it requires a smarter log parser
15192 than when it's not used. Please consult the section "Capturing HTTP headers
15193 and cookies" below for more details.
15194
15195 - "http_request" is the complete HTTP request line, including the method,
15196 request and HTTP version string. Non-printable characters are encoded (see
15197 below the section "Non-printable characters"). This is always the last
15198 field, and it is always delimited by quotes and is the only one which can
15199 contain quotes. If new fields are added to the log format, they will be
15200 added before this field. This field might be truncated if the request is
15201 huge and does not fit in the standard syslog buffer (1024 characters). This
15202 is the reason why this field must always remain the last one.
15203
15204
Cyril Bontédc4d9032012-04-08 21:57:39 +0200152058.2.4. Custom log format
15206------------------------
William Lallemand48940402012-01-30 16:47:22 +010015207
Willy Tarreau2beef582012-12-20 17:22:52 +010015208The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010015209mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010015210
15211HAproxy understands some log format variables. % precedes log format variables.
15212Variables can take arguments using braces ('{}'), and multiple arguments are
15213separated by commas within the braces. Flags may be added or removed by
15214prefixing them with a '+' or '-' sign.
15215
15216Special variable "%o" may be used to propagate its flags to all other
15217variables on the same format string. This is particularly handy with quoted
Dragan Dosen835b9212016-02-12 13:23:03 +010015218("Q") and escaped ("E") string formats.
William Lallemand48940402012-01-30 16:47:22 +010015219
Willy Tarreauc8368452012-12-21 00:09:23 +010015220If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreaube722a22014-06-13 16:31:59 +020015221as a sample expression rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010015222less common information such as the client's SSL certificate's DN, or to log
15223the key that would be used to store an entry into a stick table.
15224
William Lallemand48940402012-01-30 16:47:22 +010015225Note: spaces must be escaped. A space character is considered as a separator.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015226In order to emit a verbatim '%', it must be preceded by another '%' resulting
Willy Tarreau06d97f92013-12-02 17:45:48 +010015227in '%%'. HAProxy will automatically merge consecutive separators.
William Lallemand48940402012-01-30 16:47:22 +010015228
Dragan Dosen835b9212016-02-12 13:23:03 +010015229Note: when using the RFC5424 syslog message format, the characters '"',
15230'\' and ']' inside PARAM-VALUE should be escaped with '\' as prefix (see
15231https://tools.ietf.org/html/rfc5424#section-6.3.3 for more details). In
15232such cases, the use of the flag "E" should be considered.
15233
William Lallemand48940402012-01-30 16:47:22 +010015234Flags are :
15235 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015236 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
Dragan Dosen835b9212016-02-12 13:23:03 +010015237 * E: escape characters '"', '\' and ']' in a string with '\' as prefix
15238 (intended purpose is for the RFC5424 structured-data log formats)
William Lallemand48940402012-01-30 16:47:22 +010015239
15240 Example:
15241
15242 log-format %T\ %t\ Some\ Text
15243 log-format %{+Q}o\ %t\ %s\ %{-Q}r
15244
Dragan Dosen835b9212016-02-12 13:23:03 +010015245 log-format-sd %{+Q,+E}o\ [exampleSDID@1234\ header=%[capture.req.hdr(0)]]
15246
William Lallemand48940402012-01-30 16:47:22 +010015247At the moment, the default HTTP format is defined this way :
15248
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015249 log-format "%ci:%cp [%tr] %ft %b/%s %TR/%Tw/%Tc/%Tr/%Ta %ST %B %CC \
15250 %CS %tsc %ac/%fc/%bc/%sc/%rc %sq/%bq %hr %hs %{+Q}r"
William Lallemand48940402012-01-30 16:47:22 +010015251
William Lallemandbddd4fd2012-02-27 11:23:10 +010015252the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010015253
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015254 log-format "%{+Q}o %{-Q}ci - - [%trg] %r %ST %B \"\" \"\" %cp \
15255 %ms %ft %b %s %TR %Tw %Tc %Tr %Ta %tsc %ac %fc \
15256 %bc %sc %rc %sq %bq %CC %CS %hrl %hsl"
William Lallemand48940402012-01-30 16:47:22 +010015257
William Lallemandbddd4fd2012-02-27 11:23:10 +010015258and the default TCP format is defined this way :
15259
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015260 log-format "%ci:%cp [%t] %ft %b/%s %Tw/%Tc/%Tt %B %ts \
15261 %ac/%fc/%bc/%sc/%rc %sq/%bq"
William Lallemandbddd4fd2012-02-27 11:23:10 +010015262
William Lallemand48940402012-01-30 16:47:22 +010015263Please refer to the table below for currently defined variables :
15264
William Lallemandbddd4fd2012-02-27 11:23:10 +010015265 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020015266 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010015267 +---+------+-----------------------------------------------+-------------+
15268 | | %o | special variable, apply flags on all next var | |
15269 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010015270 | | %B | bytes_read (from server to client) | numeric |
15271 | H | %CC | captured_request_cookie | string |
15272 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020015273 | | %H | hostname | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000015274 | H | %HM | HTTP method (ex: POST) | string |
15275 | H | %HP | HTTP request URI without query string (path) | string |
Andrew Hayworthe63ac872015-07-31 16:14:16 +000015276 | H | %HQ | HTTP request URI query string (ex: ?bar=baz) | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000015277 | H | %HU | HTTP request URI (ex: /foo?bar=baz) | string |
15278 | H | %HV | HTTP version (ex: HTTP/1.0) | string |
William Lallemanda73203e2012-03-12 12:48:57 +010015279 | | %ID | unique-id | string |
Willy Tarreau4bf99632014-06-13 12:21:40 +020015280 | | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020015281 | | %T | gmt_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015282 | | %Ta | Active time of the request (from TR to end) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010015283 | | %Tc | Tc | numeric |
Willy Tarreau27b639d2016-05-17 17:55:27 +020015284 | | %Td | Td = Tt - (Tq + Tw + Tc + Tr) | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080015285 | | %Tl | local_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015286 | | %Th | connection handshake time (SSL, PROXY proto) | numeric |
15287 | H | %Ti | idle time before the HTTP request | numeric |
15288 | H | %Tq | Th + Ti + TR | numeric |
15289 | H | %TR | time to receive the full request from 1st byte| numeric |
15290 | H | %Tr | Tr (response time) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020015291 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010015292 | | %Tt | Tt | numeric |
15293 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010015294 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010015295 | | %ac | actconn | numeric |
15296 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010015297 | | %bc | beconn (backend concurrent connections) | numeric |
15298 | | %bi | backend_source_ip (connecting address) | IP |
15299 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010015300 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010015301 | | %ci | client_ip (accepted address) | IP |
15302 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010015303 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010015304 | | %fc | feconn (frontend concurrent connections) | numeric |
15305 | | %fi | frontend_ip (accepting address) | IP |
15306 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020015307 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreau7346acb2014-08-28 15:03:15 +020015308 | | %lc | frontend_log_counter | numeric |
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020015309 | | %hr | captured_request_headers default style | string |
15310 | | %hrl | captured_request_headers CLF style | string list |
15311 | | %hs | captured_response_headers default style | string |
15312 | | %hsl | captured_response_headers CLF style | string list |
Willy Tarreau812c88e2015-08-09 10:56:35 +020015313 | | %ms | accept date milliseconds (left-padded with 0) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020015314 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020015315 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010015316 | | %rc | retries | numeric |
Willy Tarreau1f0da242014-01-25 11:01:50 +010015317 | | %rt | request_counter (HTTP req or TCP session) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010015318 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010015319 | | %sc | srv_conn (server concurrent connections) | numeric |
15320 | | %si | server_IP (target address) | IP |
15321 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010015322 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020015323 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
15324 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010015325 | | %t | date_time (with millisecond resolution) | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015326 | H | %tr | date_time of HTTP request | date |
15327 | H | %trg | gmt_date_time of start of HTTP request | date |
15328 | H | %trl | locla_date_time of start of HTTP request | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010015329 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020015330 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010015331 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010015332
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020015333 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010015334
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010015335
153368.2.5. Error log format
15337-----------------------
15338
15339When an incoming connection fails due to an SSL handshake or an invalid PROXY
15340protocol header, haproxy will log the event using a shorter, fixed line format.
15341By default, logs are emitted at the LOG_INFO level, unless the option
15342"log-separate-errors" is set in the backend, in which case the LOG_ERR level
15343will be used. Connections on which no data are exchanged (eg: probes) are not
15344logged if the "dontlognull" option is set.
15345
15346The format looks like this :
15347
15348 >>> Dec 3 18:27:14 localhost \
15349 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
15350 Connection error during SSL handshake
15351
15352 Field Format Extract from the example above
15353 1 process_name '[' pid ']:' haproxy[6103]:
15354 2 client_ip ':' client_port 127.0.0.1:56059
15355 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
15356 4 frontend_name "/" bind_name ":" frt/f1:
15357 5 message Connection error during SSL handshake
15358
15359These fields just provide minimal information to help debugging connection
15360failures.
15361
15362
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200153638.3. Advanced logging options
15364-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015365
15366Some advanced logging options are often looked for but are not easy to find out
15367just by looking at the various options. Here is an entry point for the few
15368options which can enable better logging. Please refer to the keywords reference
15369for more information about their usage.
15370
15371
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200153728.3.1. Disabling logging of external tests
15373------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015374
15375It is quite common to have some monitoring tools perform health checks on
15376haproxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
15377commercial load-balancer, and sometimes it will simply be a more complete
15378monitoring system such as Nagios. When the tests are very frequent, users often
15379ask how to disable logging for those checks. There are three possibilities :
15380
15381 - if connections come from everywhere and are just TCP probes, it is often
15382 desired to simply disable logging of connections without data exchange, by
15383 setting "option dontlognull" in the frontend. It also disables logging of
15384 port scans, which may or may not be desired.
15385
15386 - if the connection come from a known source network, use "monitor-net" to
15387 declare this network as monitoring only. Any host in this network will then
15388 only be able to perform health checks, and their requests will not be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015389 logged. This is generally appropriate to designate a list of equipment
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015390 such as other load-balancers.
15391
15392 - if the tests are performed on a known URI, use "monitor-uri" to declare
15393 this URI as dedicated to monitoring. Any host sending this request will
15394 only get the result of a health-check, and the request will not be logged.
15395
15396
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200153978.3.2. Logging before waiting for the session to terminate
15398----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015399
15400The problem with logging at end of connection is that you have no clue about
15401what is happening during very long sessions, such as remote terminal sessions
15402or large file downloads. This problem can be worked around by specifying
15403"option logasap" in the frontend. Haproxy will then log as soon as possible,
15404just before data transfer begins. This means that in case of TCP, it will still
15405log the connection status to the server, and in case of HTTP, it will log just
15406after processing the server headers. In this case, the number of bytes reported
15407is the number of header bytes sent to the client. In order to avoid confusion
15408with normal logs, the total time field and the number of bytes are prefixed
15409with a '+' sign which means that real numbers are certainly larger.
15410
15411
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200154128.3.3. Raising log level upon errors
15413------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020015414
15415Sometimes it is more convenient to separate normal traffic from errors logs,
15416for instance in order to ease error monitoring from log files. When the option
15417"log-separate-errors" is used, connections which experience errors, timeouts,
15418retries, redispatches or HTTP status codes 5xx will see their syslog level
15419raised from "info" to "err". This will help a syslog daemon store the log in
15420a separate file. It is very important to keep the errors in the normal traffic
15421file too, so that log ordering is not altered. You should also be careful if
15422you already have configured your syslog daemon to store all logs higher than
15423"notice" in an "admin" file, because the "err" level is higher than "notice".
15424
15425
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200154268.3.4. Disabling logging of successful connections
15427--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020015428
15429Although this may sound strange at first, some large sites have to deal with
15430multiple thousands of logs per second and are experiencing difficulties keeping
15431them intact for a long time or detecting errors within them. If the option
15432"dontlog-normal" is set on the frontend, all normal connections will not be
15433logged. In this regard, a normal connection is defined as one without any
15434error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
15435and a response with a status 5xx is not considered normal and will be logged
15436too. Of course, doing is is really discouraged as it will remove most of the
15437useful information from the logs. Do this only if you have no other
15438alternative.
15439
15440
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200154418.4. Timing events
15442------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015443
15444Timers provide a great help in troubleshooting network problems. All values are
15445reported in milliseconds (ms). These timers should be used in conjunction with
15446the session termination flags. In TCP mode with "option tcplog" set on the
15447frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015448mode, 5 control points are reported under the form "TR/Tw/Tc/Tr/Ta". In
15449addition, three other measures are provided, "Th", "Ti", and "Tq".
15450
15451 - Th: total time to accept tcp connection and execute handshakes for low level
15452 protocols. Currently, these protocoles are proxy-protocol and SSL. This may
15453 only happen once during the whole connection's lifetime. A large time here
15454 may indicate that the client only pre-established the connection without
15455 speaking, that it is experiencing network issues preventing it from
15456 completing a handshake in a reasonable time (eg: MTU issues), or that an
15457 SSL handshake was very expensive to compute.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015458
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015459 - Ti: is the idle time before the HTTP request (HTTP mode only). This timer
15460 counts between the end of the handshakes and the first byte of the HTTP
15461 request. When dealing with a second request in keep-alive mode, it starts
15462 to count after the end of the transmission the previous response. Some
15463 browsers pre-establish connections to a server in order to reduce the
15464 latency of a future request, and keep them pending until they need it. This
15465 delay will be reported as the idle time. A value of -1 indicates that
15466 nothing was received on the connection.
15467
15468 - TR: total time to get the client request (HTTP mode only). It's the time
15469 elapsed between the first bytes received and the moment the proxy received
15470 the empty line marking the end of the HTTP headers. The value "-1"
15471 indicates that the end of headers has never been seen. This happens when
15472 the client closes prematurely or times out. This time is usually very short
15473 since most requests fit in a single packet. A large time may indicate a
15474 request typed by hand during a test.
15475
15476 - Tq: total time to get the client request from the accept date or since the
15477 emission of the last byte of the previous response (HTTP mode only). It's
15478 exactly equalt to Th + Ti + TR unless any of them is -1, in which case it
15479 returns -1 as well. This timer used to be very useful before the arrival of
15480 HTTP keep-alive and browsers' pre-connect feature. It's recommended to drop
15481 it in favor of TR nowadays, as the idle time adds a lot of noise to the
15482 reports.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015483
15484 - Tw: total time spent in the queues waiting for a connection slot. It
15485 accounts for backend queue as well as the server queues, and depends on the
15486 queue size, and the time needed for the server to complete previous
15487 requests. The value "-1" means that the request was killed before reaching
15488 the queue, which is generally what happens with invalid or denied requests.
15489
15490 - Tc: total time to establish the TCP connection to the server. It's the time
15491 elapsed between the moment the proxy sent the connection request, and the
15492 moment it was acknowledged by the server, or between the TCP SYN packet and
15493 the matching SYN/ACK packet in return. The value "-1" means that the
15494 connection never established.
15495
15496 - Tr: server response time (HTTP mode only). It's the time elapsed between
15497 the moment the TCP connection was established to the server and the moment
15498 the server sent its complete response headers. It purely shows its request
15499 processing time, without the network overhead due to the data transmission.
15500 It is worth noting that when the client has data to send to the server, for
15501 instance during a POST request, the time already runs, and this can distort
15502 apparent response time. For this reason, it's generally wise not to trust
15503 too much this field for POST requests initiated from clients behind an
15504 untrusted network. A value of "-1" here means that the last the response
15505 header (empty line) was never seen, most likely because the server timeout
15506 stroke before the server managed to process the request.
15507
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015508 - Ta: total active time for the HTTP request, between the moment the proxy
15509 received the first byte of the request header and the emission of the last
15510 byte of the response body. The exception is when the "logasap" option is
15511 specified. In this case, it only equals (TR+Tw+Tc+Tr), and is prefixed with
15512 a '+' sign. From this field, we can deduce "Td", the data transmission time,
15513 by subtracting other timers when valid :
15514
15515 Td = Ta - (TR + Tw + Tc + Tr)
15516
15517 Timers with "-1" values have to be excluded from this equation. Note that
15518 "Ta" can never be negative.
15519
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015520 - Tt: total session duration time, between the moment the proxy accepted it
15521 and the moment both ends were closed. The exception is when the "logasap"
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015522 option is specified. In this case, it only equals (Th+Ti+TR+Tw+Tc+Tr), and
15523 is prefixed with a '+' sign. From this field, we can deduce "Td", the data
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015524 transmission time, by subtracting other timers when valid :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015525
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015526 Td = Tt - (Th + Ti + TR + Tw + Tc + Tr)
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015527
15528 Timers with "-1" values have to be excluded from this equation. In TCP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015529 mode, "Ti", "Tq" and "Tr" have to be excluded too. Note that "Tt" can never
15530 be negative and that for HTTP, Tt is simply equal to (Th+Ti+Ta).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015531
15532These timers provide precious indications on trouble causes. Since the TCP
15533protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
15534that timers close to multiples of 3s are nearly always related to lost packets
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015535due to network problems (wires, negotiation, congestion). Moreover, if "Ta" or
15536"Tt" is close to a timeout value specified in the configuration, it often means
15537that a session has been aborted on timeout.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015538
15539Most common cases :
15540
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015541 - If "Th" or "Ti" are close to 3000, a packet has probably been lost between
15542 the client and the proxy. This is very rare on local networks but might
15543 happen when clients are on far remote networks and send large requests. It
15544 may happen that values larger than usual appear here without any network
15545 cause. Sometimes, during an attack or just after a resource starvation has
15546 ended, haproxy may accept thousands of connections in a few milliseconds.
15547 The time spent accepting these connections will inevitably slightly delay
15548 processing of other connections, and it can happen that request times in the
15549 order of a few tens of milliseconds are measured after a few thousands of
15550 new connections have been accepted at once. Using one of the keep-alive
15551 modes may display larger idle times since "Ti" measures the time spent
Patrick Mezard105faca2010-06-12 17:02:46 +020015552 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015553
15554 - If "Tc" is close to 3000, a packet has probably been lost between the
15555 server and the proxy during the server connection phase. This value should
15556 always be very low, such as 1 ms on local networks and less than a few tens
15557 of ms on remote networks.
15558
Willy Tarreau55165fe2009-05-10 12:02:55 +020015559 - If "Tr" is nearly always lower than 3000 except some rare values which seem
15560 to be the average majored by 3000, there are probably some packets lost
15561 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015562
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015563 - If "Ta" is large even for small byte counts, it generally is because
15564 neither the client nor the server decides to close the connection while
15565 haproxy is running in tunnel mode and both have agreed on a keep-alive
15566 connection mode. In order to solve this issue, it will be needed to specify
15567 one of the HTTP options to manipulate keep-alive or close options on either
15568 the frontend or the backend. Having the smallest possible 'Ta' or 'Tt' is
15569 important when connection regulation is used with the "maxconn" option on
15570 the servers, since no new connection will be sent to the server until
15571 another one is released.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015572
15573Other noticeable HTTP log cases ('xx' means any value to be ignored) :
15574
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015575 TR/Tw/Tc/Tr/+Ta The "option logasap" is present on the frontend and the log
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015576 was emitted before the data phase. All the timers are valid
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015577 except "Ta" which is shorter than reality.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015578
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015579 -1/xx/xx/xx/Ta The client was not able to send a complete request in time
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015580 or it aborted too early. Check the session termination flags
15581 then "timeout http-request" and "timeout client" settings.
15582
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015583 TR/-1/xx/xx/Ta It was not possible to process the request, maybe because
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015584 servers were out of order, because the request was invalid
15585 or forbidden by ACL rules. Check the session termination
15586 flags.
15587
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015588 TR/Tw/-1/xx/Ta The connection could not establish on the server. Either it
15589 actively refused it or it timed out after Ta-(TR+Tw) ms.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015590 Check the session termination flags, then check the
15591 "timeout connect" setting. Note that the tarpit action might
15592 return similar-looking patterns, with "Tw" equal to the time
15593 the client connection was maintained open.
15594
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015595 TR/Tw/Tc/-1/Ta The server has accepted the connection but did not return
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015596 a complete response in time, or it closed its connection
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020015597 unexpectedly after Ta-(TR+Tw+Tc) ms. Check the session
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015598 termination flags, then check the "timeout server" setting.
15599
15600
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200156018.5. Session state at disconnection
15602-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015603
15604TCP and HTTP logs provide a session termination indicator in the
15605"termination_state" field, just before the number of active connections. It is
156062-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
15607each of which has a special meaning :
15608
15609 - On the first character, a code reporting the first event which caused the
15610 session to terminate :
15611
15612 C : the TCP session was unexpectedly aborted by the client.
15613
15614 S : the TCP session was unexpectedly aborted by the server, or the
15615 server explicitly refused it.
15616
15617 P : the session was prematurely aborted by the proxy, because of a
15618 connection limit enforcement, because a DENY filter was matched,
15619 because of a security check which detected and blocked a dangerous
15620 error in server response which might have caused information leak
Willy Tarreau570f2212013-06-10 16:42:09 +020015621 (eg: cacheable cookie).
15622
15623 L : the session was locally processed by haproxy and was not passed to
15624 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015625
15626 R : a resource on the proxy has been exhausted (memory, sockets, source
15627 ports, ...). Usually, this appears during the connection phase, and
15628 system logs should contain a copy of the precise error. If this
15629 happens, it must be considered as a very serious anomaly which
15630 should be fixed as soon as possible by any means.
15631
15632 I : an internal error was identified by the proxy during a self-check.
15633 This should NEVER happen, and you are encouraged to report any log
15634 containing this, because this would almost certainly be a bug. It
15635 would be wise to preventively restart the process after such an
15636 event too, in case it would be caused by memory corruption.
15637
Simon Horman752dc4a2011-06-21 14:34:59 +090015638 D : the session was killed by haproxy because the server was detected
15639 as down and was configured to kill all connections when going down.
15640
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070015641 U : the session was killed by haproxy on this backup server because an
15642 active server was detected as up and was configured to kill all
15643 backup connections when going up.
15644
Willy Tarreaua2a64e92011-09-07 23:01:56 +020015645 K : the session was actively killed by an admin operating on haproxy.
15646
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015647 c : the client-side timeout expired while waiting for the client to
15648 send or receive data.
15649
15650 s : the server-side timeout expired while waiting for the server to
15651 send or receive data.
15652
15653 - : normal session completion, both the client and the server closed
15654 with nothing left in the buffers.
15655
15656 - on the second character, the TCP or HTTP session state when it was closed :
15657
Willy Tarreauf7b30a92010-12-06 22:59:17 +010015658 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015659 (HTTP mode only). Nothing was sent to any server.
15660
15661 Q : the proxy was waiting in the QUEUE for a connection slot. This can
15662 only happen when servers have a 'maxconn' parameter set. It can
15663 also happen in the global queue after a redispatch consecutive to
15664 a failed attempt to connect to a dying server. If no redispatch is
15665 reported, then no connection attempt was made to any server.
15666
15667 C : the proxy was waiting for the CONNECTION to establish on the
15668 server. The server might at most have noticed a connection attempt.
15669
15670 H : the proxy was waiting for complete, valid response HEADERS from the
15671 server (HTTP only).
15672
15673 D : the session was in the DATA phase.
15674
15675 L : the proxy was still transmitting LAST data to the client while the
15676 server had already finished. This one is very rare as it can only
15677 happen when the client dies while receiving the last packets.
15678
15679 T : the request was tarpitted. It has been held open with the client
15680 during the whole "timeout tarpit" duration or until the client
15681 closed, both of which will be reported in the "Tw" timer.
15682
15683 - : normal session completion after end of data transfer.
15684
15685 - the third character tells whether the persistence cookie was provided by
15686 the client (only in HTTP mode) :
15687
15688 N : the client provided NO cookie. This is usually the case for new
15689 visitors, so counting the number of occurrences of this flag in the
15690 logs generally indicate a valid trend for the site frequentation.
15691
15692 I : the client provided an INVALID cookie matching no known server.
15693 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020015694 cookies between HTTP/HTTPS sites, persistence conditionally
15695 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015696
15697 D : the client provided a cookie designating a server which was DOWN,
15698 so either "option persist" was used and the client was sent to
15699 this server, or it was not set and the client was redispatched to
15700 another server.
15701
Willy Tarreau996a92c2010-10-13 19:30:47 +020015702 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015703 server.
15704
Willy Tarreau996a92c2010-10-13 19:30:47 +020015705 E : the client provided a valid cookie, but with a last date which was
15706 older than what is allowed by the "maxidle" cookie parameter, so
15707 the cookie is consider EXPIRED and is ignored. The request will be
15708 redispatched just as if there was no cookie.
15709
15710 O : the client provided a valid cookie, but with a first date which was
15711 older than what is allowed by the "maxlife" cookie parameter, so
15712 the cookie is consider too OLD and is ignored. The request will be
15713 redispatched just as if there was no cookie.
15714
Willy Tarreauc89ccb62012-04-05 21:18:22 +020015715 U : a cookie was present but was not used to select the server because
15716 some other server selection mechanism was used instead (typically a
15717 "use-server" rule).
15718
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015719 - : does not apply (no cookie set in configuration).
15720
15721 - the last character reports what operations were performed on the persistence
15722 cookie returned by the server (only in HTTP mode) :
15723
15724 N : NO cookie was provided by the server, and none was inserted either.
15725
15726 I : no cookie was provided by the server, and the proxy INSERTED one.
15727 Note that in "cookie insert" mode, if the server provides a cookie,
15728 it will still be overwritten and reported as "I" here.
15729
Willy Tarreau996a92c2010-10-13 19:30:47 +020015730 U : the proxy UPDATED the last date in the cookie that was presented by
15731 the client. This can only happen in insert mode with "maxidle". It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015732 happens every time there is activity at a different date than the
Willy Tarreau996a92c2010-10-13 19:30:47 +020015733 date indicated in the cookie. If any other change happens, such as
15734 a redispatch, then the cookie will be marked as inserted instead.
15735
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015736 P : a cookie was PROVIDED by the server and transmitted as-is.
15737
15738 R : the cookie provided by the server was REWRITTEN by the proxy, which
15739 happens in "cookie rewrite" or "cookie prefix" modes.
15740
15741 D : the cookie provided by the server was DELETED by the proxy.
15742
15743 - : does not apply (no cookie set in configuration).
15744
Willy Tarreau996a92c2010-10-13 19:30:47 +020015745The combination of the two first flags gives a lot of information about what
15746was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015747helpful to detect server saturation, network troubles, local system resource
15748starvation, attacks, etc...
15749
15750The most common termination flags combinations are indicated below. They are
15751alphabetically sorted, with the lowercase set just after the upper case for
15752easier finding and understanding.
15753
15754 Flags Reason
15755
15756 -- Normal termination.
15757
15758 CC The client aborted before the connection could be established to the
15759 server. This can happen when haproxy tries to connect to a recently
15760 dead (or unchecked) server, and the client aborts while haproxy is
15761 waiting for the server to respond or for "timeout connect" to expire.
15762
15763 CD The client unexpectedly aborted during data transfer. This can be
15764 caused by a browser crash, by an intermediate equipment between the
15765 client and haproxy which decided to actively break the connection,
15766 by network routing issues between the client and haproxy, or by a
15767 keep-alive session between the server and the client terminated first
15768 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010015769
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015770 cD The client did not send nor acknowledge any data for as long as the
15771 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020015772 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015773
15774 CH The client aborted while waiting for the server to start responding.
15775 It might be the server taking too long to respond or the client
15776 clicking the 'Stop' button too fast.
15777
15778 cH The "timeout client" stroke while waiting for client data during a
15779 POST request. This is sometimes caused by too large TCP MSS values
15780 for PPPoE networks which cannot transport full-sized packets. It can
15781 also happen when client timeout is smaller than server timeout and
15782 the server takes too long to respond.
15783
15784 CQ The client aborted while its session was queued, waiting for a server
15785 with enough empty slots to accept it. It might be that either all the
15786 servers were saturated or that the assigned server was taking too
15787 long a time to respond.
15788
15789 CR The client aborted before sending a full HTTP request. Most likely
15790 the request was typed by hand using a telnet client, and aborted
15791 too early. The HTTP status code is likely a 400 here. Sometimes this
15792 might also be caused by an IDS killing the connection between haproxy
Willy Tarreau0f228a02015-05-01 15:37:53 +020015793 and the client. "option http-ignore-probes" can be used to ignore
15794 connections without any data transfer.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015795
15796 cR The "timeout http-request" stroke before the client sent a full HTTP
15797 request. This is sometimes caused by too large TCP MSS values on the
15798 client side for PPPoE networks which cannot transport full-sized
15799 packets, or by clients sending requests by hand and not typing fast
15800 enough, or forgetting to enter the empty line at the end of the
Willy Tarreau2705a612014-05-23 17:38:34 +020015801 request. The HTTP status code is likely a 408 here. Note: recently,
Willy Tarreau0f228a02015-05-01 15:37:53 +020015802 some browsers started to implement a "pre-connect" feature consisting
15803 in speculatively connecting to some recently visited web sites just
15804 in case the user would like to visit them. This results in many
15805 connections being established to web sites, which end up in 408
15806 Request Timeout if the timeout strikes first, or 400 Bad Request when
15807 the browser decides to close them first. These ones pollute the log
15808 and feed the error counters. Some versions of some browsers have even
15809 been reported to display the error code. It is possible to work
15810 around the undesirable effects of this behaviour by adding "option
15811 http-ignore-probes" in the frontend, resulting in connections with
15812 zero data transfer to be totally ignored. This will definitely hide
15813 the errors of people experiencing connectivity issues though.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015814
15815 CT The client aborted while its session was tarpitted. It is important to
15816 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020015817 wrong tarpit rules have been written. If a lot of them happen, it
15818 might make sense to lower the "timeout tarpit" value to something
15819 closer to the average reported "Tw" timer, in order not to consume
15820 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015821
Willy Tarreau570f2212013-06-10 16:42:09 +020015822 LR The request was intercepted and locally handled by haproxy. Generally
15823 it means that this was a redirect or a stats request.
15824
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010015825 SC The server or an equipment between it and haproxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015826 the TCP connection (the proxy received a TCP RST or an ICMP message
15827 in return). Under some circumstances, it can also be the network
15828 stack telling the proxy that the server is unreachable (eg: no route,
15829 or no ARP response on local network). When this happens in HTTP mode,
15830 the status code is likely a 502 or 503 here.
15831
15832 sC The "timeout connect" stroke before a connection to the server could
15833 complete. When this happens in HTTP mode, the status code is likely a
15834 503 or 504 here.
15835
15836 SD The connection to the server died with an error during the data
15837 transfer. This usually means that haproxy has received an RST from
15838 the server or an ICMP message from an intermediate equipment while
15839 exchanging data with the server. This can be caused by a server crash
15840 or by a network issue on an intermediate equipment.
15841
15842 sD The server did not send nor acknowledge any data for as long as the
15843 "timeout server" setting during the data phase. This is often caused
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010015844 by too short timeouts on L4 equipments before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015845 load-balancers, ...), as well as keep-alive sessions maintained
15846 between the client and the server expiring first on haproxy.
15847
15848 SH The server aborted before sending its full HTTP response headers, or
15849 it crashed while processing the request. Since a server aborting at
15850 this moment is very rare, it would be wise to inspect its logs to
15851 control whether it crashed and why. The logged request may indicate a
15852 small set of faulty requests, demonstrating bugs in the application.
15853 Sometimes this might also be caused by an IDS killing the connection
15854 between haproxy and the server.
15855
15856 sH The "timeout server" stroke before the server could return its
15857 response headers. This is the most common anomaly, indicating too
15858 long transactions, probably caused by server or database saturation.
15859 The immediate workaround consists in increasing the "timeout server"
15860 setting, but it is important to keep in mind that the user experience
15861 will suffer from these long response times. The only long term
15862 solution is to fix the application.
15863
15864 sQ The session spent too much time in queue and has been expired. See
15865 the "timeout queue" and "timeout connect" settings to find out how to
15866 fix this if it happens too often. If it often happens massively in
15867 short periods, it may indicate general problems on the affected
15868 servers due to I/O or database congestion, or saturation caused by
15869 external attacks.
15870
15871 PC The proxy refused to establish a connection to the server because the
15872 process' socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020015873 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015874 so that it does not happen anymore. This status is very rare and
15875 might happen when the global "ulimit-n" parameter is forced by hand.
15876
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010015877 PD The proxy blocked an incorrectly formatted chunked encoded message in
15878 a request or a response, after the server has emitted its headers. In
15879 most cases, this will indicate an invalid message from the server to
Willy Tarreauf3a3e132013-08-31 08:16:26 +020015880 the client. Haproxy supports chunk sizes of up to 2GB - 1 (2147483647
15881 bytes). Any larger size will be considered as an error.
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010015882
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015883 PH The proxy blocked the server's response, because it was invalid,
15884 incomplete, dangerous (cache control), or matched a security filter.
15885 In any case, an HTTP 502 error is sent to the client. One possible
15886 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010015887 containing unauthorized characters. It is also possible but quite
15888 rare, that the proxy blocked a chunked-encoding request from the
15889 client due to an invalid syntax, before the server responded. In this
15890 case, an HTTP 400 error is sent to the client and reported in the
15891 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015892
15893 PR The proxy blocked the client's HTTP request, either because of an
15894 invalid HTTP syntax, in which case it returned an HTTP 400 error to
15895 the client, or because a deny filter matched, in which case it
15896 returned an HTTP 403 error.
15897
15898 PT The proxy blocked the client's request and has tarpitted its
15899 connection before returning it a 500 server error. Nothing was sent
15900 to the server. The connection was maintained open for as long as
15901 reported by the "Tw" timer field.
15902
15903 RC A local resource has been exhausted (memory, sockets, source ports)
15904 preventing the connection to the server from establishing. The error
15905 logs will tell precisely what was missing. This is very rare and can
15906 only be solved by proper system tuning.
15907
Willy Tarreau996a92c2010-10-13 19:30:47 +020015908The combination of the two last flags gives a lot of information about how
15909persistence was handled by the client, the server and by haproxy. This is very
15910important to troubleshoot disconnections, when users complain they have to
15911re-authenticate. The commonly encountered flags are :
15912
15913 -- Persistence cookie is not enabled.
15914
15915 NN No cookie was provided by the client, none was inserted in the
15916 response. For instance, this can be in insert mode with "postonly"
15917 set on a GET request.
15918
15919 II A cookie designating an invalid server was provided by the client,
15920 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015921 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020015922 value can be presented by a client when no other server knows it.
15923
15924 NI No cookie was provided by the client, one was inserted in the
15925 response. This typically happens for first requests from every user
15926 in "insert" mode, which makes it an easy way to count real users.
15927
15928 VN A cookie was provided by the client, none was inserted in the
15929 response. This happens for most responses for which the client has
15930 already got a cookie.
15931
15932 VU A cookie was provided by the client, with a last visit date which is
15933 not completely up-to-date, so an updated cookie was provided in
15934 response. This can also happen if there was no date at all, or if
15935 there was a date but the "maxidle" parameter was not set, so that the
15936 cookie can be switched to unlimited time.
15937
15938 EI A cookie was provided by the client, with a last visit date which is
15939 too old for the "maxidle" parameter, so the cookie was ignored and a
15940 new cookie was inserted in the response.
15941
15942 OI A cookie was provided by the client, with a first visit date which is
15943 too old for the "maxlife" parameter, so the cookie was ignored and a
15944 new cookie was inserted in the response.
15945
15946 DI The server designated by the cookie was down, a new server was
15947 selected and a new cookie was emitted in the response.
15948
15949 VI The server designated by the cookie was not marked dead but could not
15950 be reached. A redispatch happened and selected another one, which was
15951 then advertised in the response.
15952
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015953
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200159548.6. Non-printable characters
15955-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015956
15957In order not to cause trouble to log analysis tools or terminals during log
15958consulting, non-printable characters are not sent as-is into log files, but are
15959converted to the two-digits hexadecimal representation of their ASCII code,
15960prefixed by the character '#'. The only characters that can be logged without
15961being escaped are comprised between 32 and 126 (inclusive). Obviously, the
15962escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
15963is the same for the character '"' which becomes "#22", as well as '{', '|' and
15964'}' when logging headers.
15965
15966Note that the space character (' ') is not encoded in headers, which can cause
15967issues for tools relying on space count to locate fields. A typical header
15968containing spaces is "User-Agent".
15969
15970Last, it has been observed that some syslog daemons such as syslog-ng escape
15971the quote ('"') with a backslash ('\'). The reverse operation can safely be
15972performed since no quote may appear anywhere else in the logs.
15973
15974
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200159758.7. Capturing HTTP cookies
15976---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015977
15978Cookie capture simplifies the tracking a complete user session. This can be
15979achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015980section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015981cookie will simultaneously be checked in the request ("Cookie:" header) and in
15982the response ("Set-Cookie:" header). The respective values will be reported in
15983the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015984locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015985not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
15986user switches to a new session for example, because the server will reassign it
15987a new cookie. It is also possible to detect if a server unexpectedly sets a
15988wrong cookie to a client, leading to session crossing.
15989
15990 Examples :
15991 # capture the first cookie whose name starts with "ASPSESSION"
15992 capture cookie ASPSESSION len 32
15993
15994 # capture the first cookie whose name is exactly "vgnvisitor"
15995 capture cookie vgnvisitor= len 32
15996
15997
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200159988.8. Capturing HTTP headers
15999---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016000
16001Header captures are useful to track unique request identifiers set by an upper
16002proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
16003the response, one can search for information about the response length, how the
16004server asked the cache to behave, or an object location during a redirection.
16005
16006Header captures are performed using the "capture request header" and "capture
16007response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016008section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016009
16010It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010016011time. Non-existent headers are logged as empty strings, and if one header
16012appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016013are grouped within braces '{' and '}' in the same order as they were declared,
16014and delimited with a vertical bar '|' without any space. Response headers
16015follow the same representation, but are displayed after a space following the
16016request headers block. These blocks are displayed just before the HTTP request
16017in the logs.
16018
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020016019As a special case, it is possible to specify an HTTP header capture in a TCP
16020frontend. The purpose is to enable logging of headers which will be parsed in
16021an HTTP backend if the request is then switched to this HTTP backend.
16022
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016023 Example :
16024 # This instance chains to the outgoing proxy
16025 listen proxy-out
16026 mode http
16027 option httplog
16028 option logasap
16029 log global
16030 server cache1 192.168.1.1:3128
16031
16032 # log the name of the virtual server
16033 capture request header Host len 20
16034
16035 # log the amount of data uploaded during a POST
16036 capture request header Content-Length len 10
16037
16038 # log the beginning of the referrer
16039 capture request header Referer len 20
16040
16041 # server name (useful for outgoing proxies only)
16042 capture response header Server len 20
16043
16044 # logging the content-length is useful with "option logasap"
16045 capture response header Content-Length len 10
16046
16047 # log the expected cache behaviour on the response
16048 capture response header Cache-Control len 8
16049
16050 # the Via header will report the next proxy's name
16051 capture response header Via len 20
16052
16053 # log the URL location during a redirection
16054 capture response header Location len 20
16055
16056 >>> Aug 9 20:26:09 localhost \
16057 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
16058 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
16059 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
16060 "GET http://fr.adserver.yahoo.com/"
16061
16062 >>> Aug 9 20:30:46 localhost \
16063 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
16064 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
16065 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010016066 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016067
16068 >>> Aug 9 20:30:46 localhost \
16069 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
16070 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
16071 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
16072 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010016073 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016074
16075
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200160768.9. Examples of logs
16077---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016078
16079These are real-world examples of logs accompanied with an explanation. Some of
16080them have been made up by hand. The syslog part has been removed for better
16081reading. Their sole purpose is to explain how to decipher them.
16082
16083 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
16084 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
16085 "HEAD / HTTP/1.0"
16086
16087 => long request (6.5s) entered by hand through 'telnet'. The server replied
16088 in 147 ms, and the session ended normally ('----')
16089
16090 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
16091 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
16092 0/9 "HEAD / HTTP/1.0"
16093
16094 => Idem, but the request was queued in the global queue behind 9 other
16095 requests, and waited there for 1230 ms.
16096
16097 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
16098 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
16099 "GET /image.iso HTTP/1.0"
16100
16101 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010016102 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016103 14 ms, 243 bytes of headers were sent to the client, and total time from
16104 accept to first data byte is 30 ms.
16105
16106 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
16107 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
16108 "GET /cgi-bin/bug.cgi? HTTP/1.0"
16109
16110 => the proxy blocked a server response either because of an "rspdeny" or
16111 "rspideny" filter, or because the response was improperly formatted and
Willy Tarreau3c92c5f2011-08-28 09:45:47 +020016112 not HTTP-compliant, or because it blocked sensitive information which
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016113 risked being cached. In this case, the response is replaced with a "502
16114 bad gateway". The flags ("PH--") tell us that it was haproxy who decided
16115 to return the 502 and not the server.
16116
16117 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010016118 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 +010016119
16120 => the client never completed its request and aborted itself ("C---") after
16121 8.5s, while the proxy was waiting for the request headers ("-R--").
16122 Nothing was sent to any server.
16123
16124 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
16125 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
16126
16127 => The client never completed its request, which was aborted by the
16128 time-out ("c---") after 50s, while the proxy was waiting for the request
16129 headers ("-R--"). Nothing was sent to any server, but the proxy could
16130 send a 408 return code to the client.
16131
16132 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
16133 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
16134
16135 => This log was produced with "option tcplog". The client timed out after
16136 5 seconds ("c----").
16137
16138 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
16139 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010016140 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016141
16142 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016143 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016144 (config says 'retries 3'), and no redispatch (otherwise we would have
16145 seen "/+3"). Status code 503 was returned to the client. There were 115
16146 connections on this server, 202 connections on this proxy, and 205 on
16147 the global process. It is possible that the server refused the
16148 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010016149
Willy Tarreau52b2d222011-09-07 23:48:48 +020016150
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200161519. Supported filters
16152--------------------
16153
16154Here are listed officially supported filters with the list of parameters they
16155accept. Depending on compile options, some of these filters might be
16156unavailable. The list of available filters is reported in haproxy -vv.
16157
16158See also : "filter"
16159
161609.1. Trace
16161----------
16162
16163filter trace [name <name>] [random-parsing] [random-forwarding]
16164
16165 Arguments:
16166 <name> is an arbitrary name that will be reported in
16167 messages. If no name is provided, "TRACE" is used.
16168
16169 <random-parsing> enables the random parsing of data exchanged between
16170 the client and the server. By default, this filter
16171 parses all available data. With this parameter, it
16172 only parses a random amount of the available data.
16173
16174 <random-forwarding> enables the random forwading of parsed data. By
16175 default, this filter forwards all previously parsed
16176 data. With this parameter, it only forwards a random
16177 amount of the parsed data.
16178
16179This filter can be used as a base to develop new filters. It defines all
16180callbacks and print a message on the standard error stream (stderr) with useful
16181information for all of them. It may be useful to debug the activity of other
16182filters or, quite simply, HAProxy's activity.
16183
16184Using <random-parsing> and/or <random-forwarding> parameters is a good way to
16185tests the behavior of a filter that parses data exchanged between a client and
16186a server by adding some latencies in the processing.
16187
16188
161899.2. HTTP compression
16190---------------------
16191
16192filter compression
16193
16194The HTTP compression has been moved in a filter in HAProxy 1.7. "compression"
16195keyword must still be used to enable and configure the HTTP compression. And
16196when no other filter is used, it is enough. But it is mandatory to explicitly
16197use a filter line to enable the HTTP compression when two or more filters are
16198used for the same listener/frontend/backend. This is important to know the
16199filters evaluation order.
16200
16201See also : "compression"
16202
16203
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +0200162049.3. Stream Processing Offload Engine (SPOE)
16205--------------------------------------------
16206
16207filter spoe [engine <name>] config <file>
16208
16209 Arguments :
16210
16211 <name> is the engine name that will be used to find the right scope in
16212 the configuration file. If not provided, all the file will be
16213 parsed.
16214
16215 <file> is the path of the engine configuration file. This file can
16216 contain configuration of several engines. In this case, each
16217 part must be placed in its own scope.
16218
16219The Stream Processing Offload Engine (SPOE) is a filter communicating with
16220external components. It allows the offload of some specifics processing on the
16221streams in tierce applications. These external components and information
16222exchanged with them are configured in dedicated files, for the main part. It
16223also requires dedicated backends, defined in HAProxy configuration.
16224
16225SPOE communicates with external components using an in-house binary protocol,
16226the Stream Processing Offload Protocol (SPOP).
16227
16228For all information about the SPOE configuation and the SPOP specification, see
16229"doc/SPOE.txt".
16230
16231Important note:
16232 The SPOE filter is highly experimental for now and was not heavily
16233 tested. It is really not production ready. So use it carefully.
16234
Willy Tarreau0ba27502007-12-24 16:55:16 +010016235/*
16236 * Local variables:
16237 * fill-column: 79
16238 * End:
16239 */