blob: a2eac7bb85b944be7b587ec2f50bb348756b1f52 [file] [log] [blame]
Willy Tarreau6a06a402007-07-15 20:15:28 +02001 ----------------------
Willy Tarreau8317b282014-04-23 01:49:41 +02002 HAProxy
Willy Tarreau6a06a402007-07-15 20:15:28 +02003 Configuration Manual
4 ----------------------
Willy Tarreau15480d72014-06-19 21:10:58 +02005 version 1.6
Willy Tarreau6a06a402007-07-15 20:15:28 +02006 willy tarreau
Willy Tarreau8747b6d2015-03-11 23:57:23 +01007 2015/03/11
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
Willy Tarreauc57f0e22009-05-10 13:12:33 +020053
544. Proxies
554.1. Proxy keywords matrix
564.2. Alphabetically sorted keywords reference
57
Willy Tarreau086fbf52012-09-24 20:34:51 +0200585. Bind and Server options
595.1. Bind options
605.2. Server and default-server options
Willy Tarreauc57f0e22009-05-10 13:12:33 +020061
626. HTTP header manipulation
63
Willy Tarreau74ca5042013-06-11 23:12:07 +0200647. Using ACLs and fetching samples
657.1. ACL basics
667.1.1. Matching booleans
677.1.2. Matching integers
687.1.3. Matching strings
697.1.4. Matching regular expressions (regexes)
707.1.5. Matching arbitrary data blocks
717.1.6. Matching IPv4 and IPv6 addresses
727.2. Using ACLs to form conditions
737.3. Fetching samples
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200747.3.1. Converters
757.3.2. Fetching samples from internal states
767.3.3. Fetching samples at Layer 4
777.3.4. Fetching samples at Layer 5
787.3.5. Fetching samples from buffer contents (Layer 6)
797.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +0200807.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020081
828. Logging
838.1. Log levels
848.2. Log formats
858.2.1. Default log format
868.2.2. TCP log format
878.2.3. HTTP log format
William Lallemand48940402012-01-30 16:47:22 +0100888.2.4. Custom log format
Willy Tarreau5f51e1a2012-12-03 18:40:10 +0100898.2.5. Error log format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200908.3. Advanced logging options
918.3.1. Disabling logging of external tests
928.3.2. Logging before waiting for the session to terminate
938.3.3. Raising log level upon errors
948.3.4. Disabling logging of successful connections
958.4. Timing events
968.5. Session state at disconnection
978.6. Non-printable characters
988.7. Capturing HTTP cookies
998.8. Capturing HTTP headers
1008.9. Examples of logs
101
1029. Statistics and monitoring
1039.1. CSV format
1049.2. Unix Socket commands
105
106
1071. Quick reminder about HTTP
108----------------------------
109
110When haproxy is running in HTTP mode, both the request and the response are
111fully analyzed and indexed, thus it becomes possible to build matching criteria
112on almost anything found in the contents.
113
114However, it is important to understand how HTTP requests and responses are
115formed, and how HAProxy decomposes them. It will then become easier to write
116correct rules and to debug existing configurations.
117
118
1191.1. The HTTP transaction model
120-------------------------------
121
122The HTTP protocol is transaction-driven. This means that each request will lead
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100123to one and only one response. Traditionally, a TCP connection is established
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200124from the client to the server, a request is sent by the client on the
125connection, the server responds and the connection is closed. A new request
126will involve a new connection :
127
128 [CON1] [REQ1] ... [RESP1] [CLO1] [CON2] [REQ2] ... [RESP2] [CLO2] ...
129
130In this mode, called the "HTTP close" mode, there are as many connection
131establishments as there are HTTP transactions. Since the connection is closed
132by the server after the response, the client does not need to know the content
133length.
134
135Due to the transactional nature of the protocol, it was possible to improve it
136to avoid closing a connection between two subsequent transactions. In this mode
137however, it is mandatory that the server indicates the content length for each
138response so that the client does not wait indefinitely. For this, a special
139header is used: "Content-length". This mode is called the "keep-alive" mode :
140
141 [CON] [REQ1] ... [RESP1] [REQ2] ... [RESP2] [CLO] ...
142
143Its advantages are a reduced latency between transactions, and less processing
144power required on the server side. It is generally better than the close mode,
145but not always because the clients often limit their concurrent connections to
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200146a smaller value.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200147
148A last improvement in the communications is the pipelining mode. It still uses
149keep-alive, but the client does not wait for the first response to send the
150second request. This is useful for fetching large number of images composing a
151page :
152
153 [CON] [REQ1] [REQ2] ... [RESP1] [RESP2] [CLO] ...
154
155This can obviously have a tremendous benefit on performance because the network
156latency is eliminated between subsequent requests. Many HTTP agents do not
157correctly support pipelining since there is no way to associate a response with
158the corresponding request in HTTP. For this reason, it is mandatory for the
Cyril Bonté78caf842010-03-10 22:41:43 +0100159server to reply in the exact same order as the requests were received.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200160
Willy Tarreau70dffda2014-01-30 03:07:23 +0100161By default HAProxy operates in keep-alive mode with regards to persistent
162connections: for each connection it processes each request and response, and
163leaves the connection idle on both sides between the end of a response and the
164start of a new request.
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200165
Willy Tarreau70dffda2014-01-30 03:07:23 +0100166HAProxy supports 5 connection modes :
167 - keep alive : all requests and responses are processed (default)
168 - tunnel : only the first request and response are processed,
169 everything else is forwarded with no analysis.
170 - passive close : tunnel with "Connection: close" added in both directions.
171 - server close : the server-facing connection is closed after the response.
172 - forced close : the connection is actively closed after end of response.
173
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200174
1751.2. HTTP request
176-----------------
177
178First, let's consider this HTTP request :
179
180 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100181 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200182 1 GET /serv/login.php?lang=en&profile=2 HTTP/1.1
183 2 Host: www.mydomain.com
184 3 User-agent: my small browser
185 4 Accept: image/jpeg, image/gif
186 5 Accept: image/png
187
188
1891.2.1. The Request line
190-----------------------
191
192Line 1 is the "request line". It is always composed of 3 fields :
193
194 - a METHOD : GET
195 - a URI : /serv/login.php?lang=en&profile=2
196 - a version tag : HTTP/1.1
197
198All of them are delimited by what the standard calls LWS (linear white spaces),
199which are commonly spaces, but can also be tabs or line feeds/carriage returns
200followed by spaces/tabs. The method itself cannot contain any colon (':') and
201is limited to alphabetic letters. All those various combinations make it
202desirable that HAProxy performs the splitting itself rather than leaving it to
203the user to write a complex or inaccurate regular expression.
204
205The URI itself can have several forms :
206
207 - A "relative URI" :
208
209 /serv/login.php?lang=en&profile=2
210
211 It is a complete URL without the host part. This is generally what is
212 received by servers, reverse proxies and transparent proxies.
213
214 - An "absolute URI", also called a "URL" :
215
216 http://192.168.0.12:8080/serv/login.php?lang=en&profile=2
217
218 It is composed of a "scheme" (the protocol name followed by '://'), a host
219 name or address, optionally a colon (':') followed by a port number, then
220 a relative URI beginning at the first slash ('/') after the address part.
221 This is generally what proxies receive, but a server supporting HTTP/1.1
222 must accept this form too.
223
224 - a star ('*') : this form is only accepted in association with the OPTIONS
225 method and is not relayable. It is used to inquiry a next hop's
226 capabilities.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100227
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200228 - an address:port combination : 192.168.0.12:80
229 This is used with the CONNECT method, which is used to establish TCP
230 tunnels through HTTP proxies, generally for HTTPS, but sometimes for
231 other protocols too.
232
233In a relative URI, two sub-parts are identified. The part before the question
234mark is called the "path". It is typically the relative path to static objects
235on the server. The part after the question mark is called the "query string".
236It is mostly used with GET requests sent to dynamic scripts and is very
237specific to the language, framework or application in use.
238
239
2401.2.2. The request headers
241--------------------------
242
243The headers start at the second line. They are composed of a name at the
244beginning of the line, immediately followed by a colon (':'). Traditionally,
245an LWS is added after the colon but that's not required. Then come the values.
246Multiple identical headers may be folded into one single line, delimiting the
247values with commas, provided that their order is respected. This is commonly
248encountered in the "Cookie:" field. A header may span over multiple lines if
249the subsequent lines begin with an LWS. In the example in 1.2, lines 4 and 5
250define a total of 3 values for the "Accept:" header.
251
252Contrary to a common mis-conception, header names are not case-sensitive, and
253their values are not either if they refer to other header names (such as the
254"Connection:" header).
255
256The end of the headers is indicated by the first empty line. People often say
257that it's a double line feed, which is not exact, even if a double line feed
258is one valid form of empty line.
259
260Fortunately, HAProxy takes care of all these complex combinations when indexing
261headers, checking values and counting them, so there is no reason to worry
262about the way they could be written, but it is important not to accuse an
263application of being buggy if it does unusual, valid things.
264
265Important note:
266 As suggested by RFC2616, HAProxy normalizes headers by replacing line breaks
267 in the middle of headers by LWS in order to join multi-line headers. This
268 is necessary for proper analysis and helps less capable HTTP parsers to work
269 correctly and not to be fooled by such complex constructs.
270
271
2721.3. HTTP response
273------------------
274
275An HTTP response looks very much like an HTTP request. Both are called HTTP
276messages. Let's consider this HTTP response :
277
278 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100279 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200280 1 HTTP/1.1 200 OK
281 2 Content-length: 350
282 3 Content-Type: text/html
283
Willy Tarreau816b9792009-09-15 21:25:21 +0200284As a special case, HTTP supports so called "Informational responses" as status
285codes 1xx. These messages are special in that they don't convey any part of the
286response, they're just used as sort of a signaling message to ask a client to
Willy Tarreau5843d1a2010-02-01 15:13:32 +0100287continue to post its request for instance. In the case of a status 100 response
288the requested information will be carried by the next non-100 response message
289following the informational one. This implies that multiple responses may be
290sent to a single request, and that this only works when keep-alive is enabled
291(1xx messages are HTTP/1.1 only). HAProxy handles these messages and is able to
292correctly forward and skip them, and only process the next non-100 response. As
293such, these messages are neither logged nor transformed, unless explicitly
294state otherwise. Status 101 messages indicate that the protocol is changing
295over the same connection and that haproxy must switch to tunnel mode, just as
296if a CONNECT had occurred. Then the Upgrade header would contain additional
297information about the type of protocol the connection is switching to.
Willy Tarreau816b9792009-09-15 21:25:21 +0200298
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200299
3001.3.1. The Response line
301------------------------
302
303Line 1 is the "response line". It is always composed of 3 fields :
304
305 - a version tag : HTTP/1.1
306 - a status code : 200
307 - a reason : OK
308
309The status code is always 3-digit. The first digit indicates a general status :
Willy Tarreau816b9792009-09-15 21:25:21 +0200310 - 1xx = informational message to be skipped (eg: 100, 101)
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200311 - 2xx = OK, content is following (eg: 200, 206)
312 - 3xx = OK, no content following (eg: 302, 304)
313 - 4xx = error caused by the client (eg: 401, 403, 404)
314 - 5xx = error caused by the server (eg: 500, 502, 503)
315
316Please refer to RFC2616 for the detailed meaning of all such codes. The
Willy Tarreaud72758d2010-01-12 10:42:19 +0100317"reason" field is just a hint, but is not parsed by clients. Anything can be
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200318found there, but it's a common practice to respect the well-established
319messages. It can be composed of one or multiple words, such as "OK", "Found",
320or "Authentication Required".
321
322Haproxy may emit the following status codes by itself :
323
324 Code When / reason
325 200 access to stats page, and when replying to monitoring requests
326 301 when performing a redirection, depending on the configured code
327 302 when performing a redirection, depending on the configured code
328 303 when performing a redirection, depending on the configured code
Willy Tarreaub67fdc42013-03-29 19:28:11 +0100329 307 when performing a redirection, depending on the configured code
330 308 when performing a redirection, depending on the configured code
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200331 400 for an invalid or too large request
332 401 when an authentication is required to perform the action (when
333 accessing the stats page)
334 403 when a request is forbidden by a "block" ACL or "reqdeny" filter
335 408 when the request timeout strikes before the request is complete
336 500 when haproxy encounters an unrecoverable internal error, such as a
337 memory allocation failure, which should never happen
338 502 when the server returns an empty, invalid or incomplete response, or
339 when an "rspdeny" filter blocks the response.
340 503 when no server was available to handle the request, or in response to
341 monitoring requests which match the "monitor fail" condition
342 504 when the response timeout strikes before the server responds
343
344The error 4xx and 5xx codes above may be customized (see "errorloc" in section
3454.2).
346
347
3481.3.2. The response headers
349---------------------------
350
351Response headers work exactly like request headers, and as such, HAProxy uses
352the same parsing function for both. Please refer to paragraph 1.2.2 for more
353details.
354
355
3562. Configuring HAProxy
357----------------------
358
3592.1. Configuration file format
360------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +0200361
362HAProxy's configuration process involves 3 major sources of parameters :
363
364 - the arguments from the command-line, which always take precedence
365 - the "global" section, which sets process-wide parameters
366 - the proxies sections which can take form of "defaults", "listen",
367 "frontend" and "backend".
368
Willy Tarreau0ba27502007-12-24 16:55:16 +0100369The configuration file syntax consists in lines beginning with a keyword
370referenced in this manual, optionally followed by one or several parameters
William Lallemandf9873ba2015-05-05 17:37:14 +0200371delimited by spaces.
Willy Tarreau0ba27502007-12-24 16:55:16 +0100372
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200373
William Lallemandf9873ba2015-05-05 17:37:14 +02003742.2. Quoting and escaping
375-------------------------
376
377HAProxy's configuration introduces a quoting and escaping system similar to
378many programming languages. The configuration file supports 3 types: escaping
379with a backslash, weak quoting with double quotes, and strong quoting with
380single quotes.
381
382If spaces have to be entered in strings, then they must be escaped by preceding
383them by a backslash ('\') or by quoting them. Backslashes also have to be
384escaped by doubling or strong quoting them.
385
386Escaping is achieved by preceding a special character by a backslash ('\'):
387
388 \ to mark a space and differentiate it from a delimiter
389 \# to mark a hash and differentiate it from a comment
390 \\ to use a backslash
391 \' to use a single quote and differentiate it from strong quoting
392 \" to use a double quote and differentiate it from weak quoting
393
394Weak quoting is achieved by using double quotes (""). Weak quoting prevents
395the interpretation of:
396
397 space as a parameter separator
398 ' single quote as a strong quoting delimiter
399 # hash as a comment start
400
William Lallemandb2f07452015-05-12 14:27:13 +0200401Weak quoting permits the interpretation of variables, if you want to use a non
402-interpreted dollar within a double quoted string, you should escape it with a
403backslash ("\$"), it does not work outside weak quoting.
404
405Interpretation of escaping and special characters are not prevented by weak
William Lallemandf9873ba2015-05-05 17:37:14 +0200406quoting.
407
408Strong quoting is achieved by using single quotes (''). Inside single quotes,
409nothing is interpreted, it's the efficient way to quote regexes.
410
411Quoted and escaped strings are replaced in memory by their interpreted
412equivalent, it allows you to perform concatenation.
413
414 Example:
415 # those are equivalents:
416 log-format %{+Q}o\ %t\ %s\ %{-Q}r
417 log-format "%{+Q}o %t %s %{-Q}r"
418 log-format '%{+Q}o %t %s %{-Q}r'
419 log-format "%{+Q}o %t"' %s %{-Q}r'
420 log-format "%{+Q}o %t"' %s'\ %{-Q}r
421
422 # those are equivalents:
423 reqrep "^([^\ :]*)\ /static/(.*)" \1\ /\2
424 reqrep "^([^ :]*)\ /static/(.*)" '\1 /\2'
425 reqrep "^([^ :]*)\ /static/(.*)" "\1 /\2"
426 reqrep "^([^ :]*)\ /static/(.*)" "\1\ /\2"
427
428
William Lallemandb2f07452015-05-12 14:27:13 +02004292.3. Environment variables
430--------------------------
431
432HAProxy's configuration supports environment variables. Those variables are
433interpreted only within double quotes. Variables are expanded during the
434configuration parsing. Variable names must be preceded by a dollar ("$") and
435optionally enclosed with braces ("{}") similarly to what is done in Bourne
436shell. Variable names can contain alphanumerical characters or the character
437underscore ("_") but should not start with a digit.
438
439 Example:
440
441 bind "fd@${FD_APP1}"
442
443 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
444
445 user "$HAPROXY_USER"
446
447
4482.4. Time format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200449----------------
450
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100451Some parameters involve values representing time, such as timeouts. These
Willy Tarreau0ba27502007-12-24 16:55:16 +0100452values are generally expressed in milliseconds (unless explicitly stated
453otherwise) but may be expressed in any other unit by suffixing the unit to the
454numeric value. It is important to consider this because it will not be repeated
455for every keyword. Supported units are :
456
457 - us : microseconds. 1 microsecond = 1/1000000 second
458 - ms : milliseconds. 1 millisecond = 1/1000 second. This is the default.
459 - s : seconds. 1s = 1000ms
460 - m : minutes. 1m = 60s = 60000ms
461 - h : hours. 1h = 60m = 3600s = 3600000ms
462 - d : days. 1d = 24h = 1440m = 86400s = 86400000ms
463
464
William Lallemandf9873ba2015-05-05 17:37:14 +02004652.4. Examples
Patrick Mezard35da19c2010-06-12 17:02:47 +0200466-------------
467
468 # Simple configuration for an HTTP proxy listening on port 80 on all
469 # interfaces and forwarding requests to a single backend "servers" with a
470 # single server "server1" listening on 127.0.0.1:8000
471 global
472 daemon
473 maxconn 256
474
475 defaults
476 mode http
477 timeout connect 5000ms
478 timeout client 50000ms
479 timeout server 50000ms
480
481 frontend http-in
482 bind *:80
483 default_backend servers
484
485 backend servers
486 server server1 127.0.0.1:8000 maxconn 32
487
488
489 # The same configuration defined with a single listen block. Shorter but
490 # less expressive, especially in HTTP mode.
491 global
492 daemon
493 maxconn 256
494
495 defaults
496 mode http
497 timeout connect 5000ms
498 timeout client 50000ms
499 timeout server 50000ms
500
501 listen http-in
502 bind *:80
503 server server1 127.0.0.1:8000 maxconn 32
504
505
506Assuming haproxy is in $PATH, test these configurations in a shell with:
507
Willy Tarreauccb289d2010-12-11 20:19:38 +0100508 $ sudo haproxy -f configuration.conf -c
Patrick Mezard35da19c2010-06-12 17:02:47 +0200509
510
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005113. Global parameters
Willy Tarreau6a06a402007-07-15 20:15:28 +0200512--------------------
513
514Parameters in the "global" section are process-wide and often OS-specific. They
515are generally set once for all and do not need being changed once correct. Some
516of them have command-line equivalents.
517
518The following keywords are supported in the "global" section :
519
520 * Process management and security
Emeric Brunc8e8d122012-10-02 18:42:10 +0200521 - ca-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200522 - chroot
Emeric Brunc8e8d122012-10-02 18:42:10 +0200523 - crt-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200524 - daemon
Simon Horman98637e52014-06-20 12:30:16 +0900525 - external-check
Willy Tarreau6a06a402007-07-15 20:15:28 +0200526 - gid
527 - group
528 - log
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100529 - log-send-hostname
Willy Tarreau6a06a402007-07-15 20:15:28 +0200530 - nbproc
531 - pidfile
532 - uid
533 - ulimit-n
534 - user
Willy Tarreaufbee7132007-10-18 13:53:22 +0200535 - stats
Emeric Brun850efd52014-01-29 12:24:34 +0100536 - ssl-server-verify
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +0200537 - node
538 - description
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100539 - unix-bind
Willy Tarreaud72758d2010-01-12 10:42:19 +0100540
Willy Tarreau6a06a402007-07-15 20:15:28 +0200541 * Performance tuning
Willy Tarreau1746eec2014-04-25 10:46:47 +0200542 - max-spread-checks
Willy Tarreau6a06a402007-07-15 20:15:28 +0200543 - maxconn
Willy Tarreau81c25d02011-09-07 15:17:21 +0200544 - maxconnrate
William Lallemandd85f9172012-11-09 17:05:39 +0100545 - maxcomprate
William Lallemand072a2bf2012-11-20 17:01:01 +0100546 - maxcompcpuusage
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100547 - maxpipes
Willy Tarreau93e7c002013-10-07 18:51:07 +0200548 - maxsessrate
Willy Tarreau403edff2012-09-06 11:58:37 +0200549 - maxsslconn
Willy Tarreaue43d5322013-10-07 20:01:52 +0200550 - maxsslrate
Willy Tarreau6a06a402007-07-15 20:15:28 +0200551 - noepoll
552 - nokqueue
553 - nopoll
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100554 - nosplice
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300555 - nogetaddrinfo
Willy Tarreaufe255b72007-10-14 23:09:26 +0200556 - spread-checks
Willy Tarreau27a674e2009-08-17 07:23:33 +0200557 - tune.bufsize
Willy Tarreau43961d52010-10-04 20:39:20 +0200558 - tune.chksize
William Lallemandf3747832012-11-09 12:33:10 +0100559 - tune.comp.maxlevel
Willy Tarreau193b8c62012-11-22 00:17:38 +0100560 - tune.http.cookielen
Willy Tarreauac1932d2011-10-24 19:14:41 +0200561 - tune.http.maxhdr
Willy Tarreau7e312732014-02-12 16:35:14 +0100562 - tune.idletimer
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100563 - tune.lua.forced-yield
Willy Tarreau32f61e22015-03-18 17:54:59 +0100564 - tune.lua.maxmem
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100565 - tune.lua.session-timeout
566 - tune.lua.task-timeout
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100567 - tune.maxaccept
568 - tune.maxpollevents
Willy Tarreau27a674e2009-08-17 07:23:33 +0200569 - tune.maxrewrite
Willy Tarreauf3045d22015-04-29 16:24:50 +0200570 - tune.pattern.cache-size
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200571 - tune.pipesize
Willy Tarreaue803de22010-01-21 17:43:04 +0100572 - tune.rcvbuf.client
573 - tune.rcvbuf.server
574 - tune.sndbuf.client
575 - tune.sndbuf.server
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100576 - tune.ssl.cachesize
Willy Tarreaubfd59462013-02-21 07:46:09 +0100577 - tune.ssl.lifetime
Emeric Brun8dc60392014-05-09 13:52:00 +0200578 - tune.ssl.force-private-cache
Willy Tarreaubfd59462013-02-21 07:46:09 +0100579 - tune.ssl.maxrecord
Remi Gacognef46cd6e2014-06-12 14:58:40 +0200580 - tune.ssl.default-dh-param
William Lallemanda509e4c2012-11-07 16:54:34 +0100581 - tune.zlib.memlevel
582 - tune.zlib.windowsize
Willy Tarreaud72758d2010-01-12 10:42:19 +0100583
Willy Tarreau6a06a402007-07-15 20:15:28 +0200584 * Debugging
585 - debug
586 - quiet
Willy Tarreau6a06a402007-07-15 20:15:28 +0200587
588
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005893.1. Process management and security
Willy Tarreau6a06a402007-07-15 20:15:28 +0200590------------------------------------
591
Emeric Brunc8e8d122012-10-02 18:42:10 +0200592ca-base <dir>
593 Assigns a default directory to fetch SSL CA certificates and CRLs from when a
Emeric Brunfd33a262012-10-11 16:28:27 +0200594 relative path is used with "ca-file" or "crl-file" directives. Absolute
595 locations specified in "ca-file" and "crl-file" prevail and ignore "ca-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +0200596
Willy Tarreau6a06a402007-07-15 20:15:28 +0200597chroot <jail dir>
598 Changes current directory to <jail dir> and performs a chroot() there before
599 dropping privileges. This increases the security level in case an unknown
600 vulnerability would be exploited, since it would make it very hard for the
601 attacker to exploit the system. This only works when the process is started
602 with superuser privileges. It is important to ensure that <jail_dir> is both
603 empty and unwritable to anyone.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100604
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100605cpu-map <"all"|"odd"|"even"|process_num> <cpu-set>...
606 On Linux 2.6 and above, it is possible to bind a process to a specific CPU
607 set. This means that the process will never run on other CPUs. The "cpu-map"
608 directive specifies CPU sets for process sets. The first argument is the
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100609 process number to bind. This process must have a number between 1 and 32 or
610 64, depending on the machine's word size, and any process IDs above nbproc
611 are ignored. It is possible to specify all processes at once using "all",
612 only odd numbers using "odd" or even numbers using "even", just like with the
613 "bind-process" directive. The second and forthcoming arguments are CPU sets.
614 Each CPU set is either a unique number between 0 and 31 or 63 or a range with
615 two such numbers delimited by a dash ('-'). Multiple CPU numbers or ranges
616 may be specified, and the processes will be allowed to bind to all of them.
617 Obviously, multiple "cpu-map" directives may be specified. Each "cpu-map"
618 directive will replace the previous ones when they overlap.
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100619
Emeric Brunc8e8d122012-10-02 18:42:10 +0200620crt-base <dir>
621 Assigns a default directory to fetch SSL certificates from when a relative
622 path is used with "crtfile" directives. Absolute locations specified after
623 "crtfile" prevail and ignore "crt-base".
624
Willy Tarreau6a06a402007-07-15 20:15:28 +0200625daemon
626 Makes the process fork into background. This is the recommended mode of
627 operation. It is equivalent to the command line "-D" argument. It can be
628 disabled by the command line "-db" argument.
629
Simon Horman98637e52014-06-20 12:30:16 +0900630external-check
631 Allows the use of an external agent to perform health checks.
632 This is disabled by default as a security precaution.
633 See "option external-check".
634
Willy Tarreau6a06a402007-07-15 20:15:28 +0200635gid <number>
636 Changes the process' group ID to <number>. It is recommended that the group
637 ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
638 be started with a user belonging to this group, or with superuser privileges.
Michael Schererab012dd2013-01-12 18:35:19 +0100639 Note that if haproxy is started from a user having supplementary groups, it
640 will only be able to drop these groups if started with superuser privileges.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200641 See also "group" and "uid".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100642
Willy Tarreau6a06a402007-07-15 20:15:28 +0200643group <group name>
644 Similar to "gid" but uses the GID of group name <group name> from /etc/group.
645 See also "gid" and "user".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100646
Willy Tarreau18324f52014-06-27 18:10:07 +0200647log <address> [len <length>] <facility> [max level [min level]]
Willy Tarreau6a06a402007-07-15 20:15:28 +0200648 Adds a global syslog server. Up to two global servers can be defined. They
649 will receive logs for startups and exits, as well as all logs from proxies
Robert Tsai81ae1952007-12-05 10:47:29 +0100650 configured with "log global".
651
652 <address> can be one of:
653
Willy Tarreau2769aa02007-12-27 18:26:09 +0100654 - An IPv4 address optionally followed by a colon and a UDP port. If
Robert Tsai81ae1952007-12-05 10:47:29 +0100655 no port is specified, 514 is used by default (the standard syslog
656 port).
657
David du Colombier24bb5f52011-03-17 10:40:23 +0100658 - An IPv6 address followed by a colon and optionally a UDP port. If
659 no port is specified, 514 is used by default (the standard syslog
660 port).
661
Robert Tsai81ae1952007-12-05 10:47:29 +0100662 - A filesystem path to a UNIX domain socket, keeping in mind
663 considerations for chroot (be sure the path is accessible inside
664 the chroot) and uid/gid (be sure the path is appropriately
665 writeable).
666
William Lallemandb2f07452015-05-12 14:27:13 +0200667 You may want to reference some environment variables in the address
668 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +0100669
Willy Tarreau18324f52014-06-27 18:10:07 +0200670 <length> is an optional maximum line length. Log lines larger than this value
671 will be truncated before being sent. The reason is that syslog
672 servers act differently on log line length. All servers support the
673 default value of 1024, but some servers simply drop larger lines
674 while others do log them. If a server supports long lines, it may
675 make sense to set this value here in order to avoid truncating long
676 lines. Similarly, if a server drops long lines, it is preferable to
677 truncate them before sending them. Accepted values are 80 to 65535
678 inclusive. The default value of 1024 is generally fine for all
679 standard usages. Some specific cases of long captures or
680 JSON-formated logs may require larger values.
681
Robert Tsai81ae1952007-12-05 10:47:29 +0100682 <facility> must be one of the 24 standard syslog facilities :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200683
684 kern user mail daemon auth syslog lpr news
685 uucp cron auth2 ftp ntp audit alert cron2
686 local0 local1 local2 local3 local4 local5 local6 local7
687
688 An optional level can be specified to filter outgoing messages. By default,
Willy Tarreauf7edefa2009-05-10 17:20:05 +0200689 all messages are sent. If a maximum level is specified, only messages with a
690 severity at least as important as this level will be sent. An optional minimum
691 level can be specified. If it is set, logs emitted with a more severe level
692 than this one will be capped to this level. This is used to avoid sending
693 "emerg" messages on all terminals on some default syslog configurations.
694 Eight levels are known :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200695
Cyril Bontédc4d9032012-04-08 21:57:39 +0200696 emerg alert crit err warning notice info debug
Willy Tarreau6a06a402007-07-15 20:15:28 +0200697
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100698log-send-hostname [<string>]
699 Sets the hostname field in the syslog header. If optional "string" parameter
700 is set the header is set to the string contents, otherwise uses the hostname
701 of the system. Generally used if one is not relaying logs through an
702 intermediate syslog server or for simply customizing the hostname printed in
703 the logs.
704
Kevinm48936af2010-12-22 16:08:21 +0000705log-tag <string>
706 Sets the tag field in the syslog header to this string. It defaults to the
707 program name as launched from the command line, which usually is "haproxy".
708 Sometimes it can be useful to differentiate between multiple processes
Willy Tarreau094af4e2015-01-07 15:03:42 +0100709 running on the same host. See also the per-proxy "log-tag" directive.
Kevinm48936af2010-12-22 16:08:21 +0000710
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100711lua-load <file>
712 This global directive loads and executes a Lua file. This directive can be
713 used multiple times.
714
Willy Tarreau6a06a402007-07-15 20:15:28 +0200715nbproc <number>
716 Creates <number> processes when going daemon. This requires the "daemon"
717 mode. By default, only one process is created, which is the recommended mode
718 of operation. For systems limited to small sets of file descriptors per
719 process, it may be needed to fork multiple daemons. USING MULTIPLE PROCESSES
720 IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. See also "daemon".
721
722pidfile <pidfile>
723 Writes pids of all daemons into file <pidfile>. This option is equivalent to
724 the "-p" command line argument. The file must be accessible to the user
725 starting the process. See also "daemon".
726
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100727stats bind-process [ all | odd | even | <number 1-64>[-<number 1-64>] ] ...
Willy Tarreau35b7b162012-10-22 23:17:18 +0200728 Limits the stats socket to a certain set of processes numbers. By default the
729 stats socket is bound to all processes, causing a warning to be emitted when
730 nbproc is greater than 1 because there is no way to select the target process
731 when connecting. However, by using this setting, it becomes possible to pin
732 the stats socket to a specific set of processes, typically the first one. The
733 warning will automatically be disabled when this setting is used, whatever
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100734 the number of processes used. The maximum process ID depends on the machine's
Willy Tarreauae302532014-05-07 19:22:24 +0200735 word size (32 or 64). A better option consists in using the "process" setting
736 of the "stats socket" line to force the process on each line.
Willy Tarreau35b7b162012-10-22 23:17:18 +0200737
Willy Tarreau610f04b2014-02-13 11:36:41 +0100738ssl-default-bind-ciphers <ciphers>
739 This setting is only available when support for OpenSSL was built in. It sets
740 the default string describing the list of cipher algorithms ("cipher suite")
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300741 that are negotiated during the SSL/TLS handshake for all "bind" lines which
Willy Tarreau610f04b2014-02-13 11:36:41 +0100742 do not explicitly define theirs. The format of the string is defined in
743 "man 1 ciphers" from OpenSSL man pages, and can be for instance a string such
744 as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes). Please check the
745 "bind" keyword for more information.
746
Emeric Brun2c86cbf2014-10-30 15:56:50 +0100747ssl-default-bind-options [<option>]...
748 This setting is only available when support for OpenSSL was built in. It sets
749 default ssl-options to force on all "bind" lines. Please check the "bind"
750 keyword to see available options.
751
752 Example:
753 global
754 ssl-default-bind-options no-sslv3 no-tls-tickets
755
Willy Tarreau610f04b2014-02-13 11:36:41 +0100756ssl-default-server-ciphers <ciphers>
757 This setting is only available when support for OpenSSL was built in. It
758 sets the default string describing the list of cipher algorithms that are
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300759 negotiated during the SSL/TLS handshake with the server, for all "server"
Willy Tarreau610f04b2014-02-13 11:36:41 +0100760 lines which do not explicitly define theirs. The format of the string is
761 defined in "man 1 ciphers". Please check the "server" keyword for more
762 information.
763
Emeric Brun2c86cbf2014-10-30 15:56:50 +0100764ssl-default-server-options [<option>]...
765 This setting is only available when support for OpenSSL was built in. It sets
766 default ssl-options to force on all "server" lines. Please check the "server"
767 keyword to see available options.
768
Emeric Brun850efd52014-01-29 12:24:34 +0100769ssl-server-verify [none|required]
770 The default behavior for SSL verify on servers side. If specified to 'none',
771 servers certificates are not verified. The default is 'required' except if
772 forced using cmdline option '-dV'.
773
Willy Tarreauabb175f2012-09-24 12:43:26 +0200774stats socket [<address:port>|<path>] [param*]
775 Binds a UNIX socket to <path> or a TCPv4/v6 address to <address:port>.
776 Connections to this socket will return various statistics outputs and even
777 allow some commands to be issued to change some runtime settings. Please
778 consult section 9.2 "Unix Socket commands" for more details.
Willy Tarreau6162db22009-10-10 17:13:00 +0200779
Willy Tarreauabb175f2012-09-24 12:43:26 +0200780 All parameters supported by "bind" lines are supported, for instance to
781 restrict access to some users or their access rights. Please consult
782 section 5.1 for more information.
Willy Tarreaufbee7132007-10-18 13:53:22 +0200783
784stats timeout <timeout, in milliseconds>
785 The default timeout on the stats socket is set to 10 seconds. It is possible
786 to change this value with "stats timeout". The value must be passed in
Willy Tarreaubefdff12007-12-02 22:27:38 +0100787 milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.
Willy Tarreaufbee7132007-10-18 13:53:22 +0200788
789stats maxconn <connections>
790 By default, the stats socket is limited to 10 concurrent connections. It is
791 possible to change this value with "stats maxconn".
792
Willy Tarreau6a06a402007-07-15 20:15:28 +0200793uid <number>
794 Changes the process' user ID to <number>. It is recommended that the user ID
795 is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
796 be started with superuser privileges in order to be able to switch to another
797 one. See also "gid" and "user".
798
799ulimit-n <number>
800 Sets the maximum number of per-process file-descriptors to <number>. By
801 default, it is automatically computed, so it is recommended not to use this
802 option.
803
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100804unix-bind [ prefix <prefix> ] [ mode <mode> ] [ user <user> ] [ uid <uid> ]
805 [ group <group> ] [ gid <gid> ]
806
807 Fixes common settings to UNIX listening sockets declared in "bind" statements.
808 This is mainly used to simplify declaration of those UNIX sockets and reduce
809 the risk of errors, since those settings are most commonly required but are
810 also process-specific. The <prefix> setting can be used to force all socket
811 path to be relative to that directory. This might be needed to access another
812 component's chroot. Note that those paths are resolved before haproxy chroots
813 itself, so they are absolute. The <mode>, <user>, <uid>, <group> and <gid>
814 all have the same meaning as their homonyms used by the "bind" statement. If
815 both are specified, the "bind" statement has priority, meaning that the
816 "unix-bind" settings may be seen as process-wide default settings.
817
Willy Tarreau6a06a402007-07-15 20:15:28 +0200818user <user name>
819 Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.
820 See also "uid" and "group".
821
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +0200822node <name>
823 Only letters, digits, hyphen and underscore are allowed, like in DNS names.
824
825 This statement is useful in HA configurations where two or more processes or
826 servers share the same IP address. By setting a different node-name on all
827 nodes, it becomes easy to immediately spot what server is handling the
828 traffic.
829
830description <text>
831 Add a text that describes the instance.
832
833 Please note that it is required to escape certain characters (# for example)
834 and this text is inserted into a html page so you should avoid using
835 "<" and ">" characters.
836
Willy Tarreau6a06a402007-07-15 20:15:28 +0200837
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008383.2. Performance tuning
Willy Tarreau6a06a402007-07-15 20:15:28 +0200839-----------------------
840
Willy Tarreau1746eec2014-04-25 10:46:47 +0200841max-spread-checks <delay in milliseconds>
842 By default, haproxy tries to spread the start of health checks across the
843 smallest health check interval of all the servers in a farm. The principle is
844 to avoid hammering services running on the same server. But when using large
845 check intervals (10 seconds or more), the last servers in the farm take some
846 time before starting to be tested, which can be a problem. This parameter is
847 used to enforce an upper bound on delay between the first and the last check,
848 even if the servers' check intervals are larger. When servers run with
849 shorter intervals, their intervals will be respected though.
850
Willy Tarreau6a06a402007-07-15 20:15:28 +0200851maxconn <number>
852 Sets the maximum per-process number of concurrent connections to <number>. It
853 is equivalent to the command-line argument "-n". Proxies will stop accepting
854 connections when this limit is reached. The "ulimit-n" parameter is
Willy Tarreau8274e102014-06-19 15:31:25 +0200855 automatically adjusted according to this value. See also "ulimit-n". Note:
856 the "select" poller cannot reliably use more than 1024 file descriptors on
857 some platforms. If your platform only supports select and reports "select
858 FAILED" on startup, you need to reduce maxconn until it works (slightly
Willy Tarreaud0256482015-01-15 21:45:22 +0100859 below 500 in general). If this value is not set, it will default to the value
860 set in DEFAULT_MAXCONN at build time (reported in haproxy -vv) if no memory
861 limit is enforced, or will be computed based on the memory limit, the buffer
862 size, memory allocated to compression, SSL cache size, and use or not of SSL
863 and the associated maxsslconn (which can also be automatic).
Willy Tarreau6a06a402007-07-15 20:15:28 +0200864
Willy Tarreau81c25d02011-09-07 15:17:21 +0200865maxconnrate <number>
866 Sets the maximum per-process number of connections per second to <number>.
867 Proxies will stop accepting connections when this limit is reached. It can be
868 used to limit the global capacity regardless of each frontend capacity. It is
869 important to note that this can only be used as a service protection measure,
870 as there will not necessarily be a fair share between frontends when the
871 limit is reached, so it's a good idea to also limit each frontend to some
872 value close to its expected share. Also, lowering tune.maxaccept can improve
873 fairness.
874
William Lallemandd85f9172012-11-09 17:05:39 +0100875maxcomprate <number>
876 Sets the maximum per-process input compression rate to <number> kilobytes
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300877 per second. For each session, if the maximum is reached, the compression
William Lallemandd85f9172012-11-09 17:05:39 +0100878 level will be decreased during the session. If the maximum is reached at the
879 beginning of a session, the session will not compress at all. If the maximum
880 is not reached, the compression level will be increased up to
881 tune.comp.maxlevel. A value of zero means there is no limit, this is the
882 default value.
883
William Lallemand072a2bf2012-11-20 17:01:01 +0100884maxcompcpuusage <number>
885 Sets the maximum CPU usage HAProxy can reach before stopping the compression
886 for new requests or decreasing the compression level of current requests.
887 It works like 'maxcomprate' but measures CPU usage instead of incoming data
888 bandwidth. The value is expressed in percent of the CPU used by haproxy. In
889 case of multiple processes (nbproc > 1), each process manages its individual
890 usage. A value of 100 disable the limit. The default value is 100. Setting
891 a lower value will prevent the compression work from slowing the whole
892 process down and from introducing high latencies.
893
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100894maxpipes <number>
895 Sets the maximum per-process number of pipes to <number>. Currently, pipes
896 are only used by kernel-based tcp splicing. Since a pipe contains two file
897 descriptors, the "ulimit-n" value will be increased accordingly. The default
898 value is maxconn/4, which seems to be more than enough for most heavy usages.
899 The splice code dynamically allocates and releases pipes, and can fall back
900 to standard copy, so setting this value too low may only impact performance.
901
Willy Tarreau93e7c002013-10-07 18:51:07 +0200902maxsessrate <number>
903 Sets the maximum per-process number of sessions per second to <number>.
904 Proxies will stop accepting connections when this limit is reached. It can be
905 used to limit the global capacity regardless of each frontend capacity. It is
906 important to note that this can only be used as a service protection measure,
907 as there will not necessarily be a fair share between frontends when the
908 limit is reached, so it's a good idea to also limit each frontend to some
909 value close to its expected share. Also, lowering tune.maxaccept can improve
910 fairness.
911
Willy Tarreau403edff2012-09-06 11:58:37 +0200912maxsslconn <number>
913 Sets the maximum per-process number of concurrent SSL connections to
914 <number>. By default there is no SSL-specific limit, which means that the
915 global maxconn setting will apply to all connections. Setting this limit
916 avoids having openssl use too much memory and crash when malloc returns NULL
917 (since it unfortunately does not reliably check for such conditions). Note
918 that the limit applies both to incoming and outgoing connections, so one
919 connection which is deciphered then ciphered accounts for 2 SSL connections.
Willy Tarreaud0256482015-01-15 21:45:22 +0100920 If this value is not set, but a memory limit is enforced, this value will be
921 automatically computed based on the memory limit, maxconn, the buffer size,
922 memory allocated to compression, SSL cache size, and use of SSL in either
923 frontends, backends or both. If neither maxconn nor maxsslconn are specified
924 when there is a memory limit, haproxy will automatically adjust these values
925 so that 100% of the connections can be made over SSL with no risk, and will
926 consider the sides where it is enabled (frontend, backend, both).
Willy Tarreau403edff2012-09-06 11:58:37 +0200927
Willy Tarreaue43d5322013-10-07 20:01:52 +0200928maxsslrate <number>
929 Sets the maximum per-process number of SSL sessions per second to <number>.
930 SSL listeners will stop accepting connections when this limit is reached. It
931 can be used to limit the global SSL CPU usage regardless of each frontend
932 capacity. It is important to note that this can only be used as a service
933 protection measure, as there will not necessarily be a fair share between
934 frontends when the limit is reached, so it's a good idea to also limit each
935 frontend to some value close to its expected share. It is also important to
936 note that the sessions are accounted before they enter the SSL stack and not
937 after, which also protects the stack against bad handshakes. Also, lowering
938 tune.maxaccept can improve fairness.
939
William Lallemand9d5f5482012-11-07 16:12:57 +0100940maxzlibmem <number>
941 Sets the maximum amount of RAM in megabytes per process usable by the zlib.
942 When the maximum amount is reached, future sessions will not compress as long
943 as RAM is unavailable. When sets to 0, there is no limit.
William Lallemande3a7d992012-11-20 11:25:20 +0100944 The default value is 0. The value is available in bytes on the UNIX socket
945 with "show info" on the line "MaxZlibMemUsage", the memory used by zlib is
946 "ZlibMemUsage" in bytes.
947
Willy Tarreau6a06a402007-07-15 20:15:28 +0200948noepoll
949 Disables the use of the "epoll" event polling system on Linux. It is
950 equivalent to the command-line argument "-de". The next polling system
Willy Tarreaue9f49e72012-11-11 17:42:00 +0100951 used will generally be "poll". See also "nopoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +0200952
953nokqueue
954 Disables the use of the "kqueue" event polling system on BSD. It is
955 equivalent to the command-line argument "-dk". The next polling system
956 used will generally be "poll". See also "nopoll".
957
958nopoll
959 Disables the use of the "poll" event polling system. It is equivalent to the
960 command-line argument "-dp". The next polling system used will be "select".
Willy Tarreau0ba27502007-12-24 16:55:16 +0100961 It should never be needed to disable "poll" since it's available on all
Willy Tarreaue9f49e72012-11-11 17:42:00 +0100962 platforms supported by HAProxy. See also "nokqueue" and "noepoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +0200963
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100964nosplice
965 Disables the use of kernel tcp splicing between sockets on Linux. It is
966 equivalent to the command line argument "-dS". Data will then be copied
967 using conventional and more portable recv/send calls. Kernel tcp splicing is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100968 limited to some very recent instances of kernel 2.6. Most versions between
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100969 2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not
970 be used. This option makes it easier to globally disable kernel splicing in
971 case of doubt. See also "option splice-auto", "option splice-request" and
972 "option splice-response".
973
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300974nogetaddrinfo
975 Disables the use of getaddrinfo(3) for name resolving. It is equivalent to
976 the command line argument "-dG". Deprecated gethostbyname(3) will be used.
977
Willy Tarreaufe255b72007-10-14 23:09:26 +0200978spread-checks <0..50, in percent>
Simon Hormand60d6912013-11-25 10:46:36 +0900979 Sometimes it is desirable to avoid sending agent and health checks to
980 servers at exact intervals, for instance when many logical servers are
981 located on the same physical server. With the help of this parameter, it
982 becomes possible to add some randomness in the check interval between 0
983 and +/- 50%. A value between 2 and 5 seems to show good results. The
984 default value remains at 0.
Willy Tarreaufe255b72007-10-14 23:09:26 +0200985
Willy Tarreau33cb0652014-12-23 22:52:37 +0100986tune.buffers.limit <number>
987 Sets a hard limit on the number of buffers which may be allocated per process.
988 The default value is zero which means unlimited. The minimum non-zero value
989 will always be greater than "tune.buffers.reserve" and should ideally always
990 be about twice as large. Forcing this value can be particularly useful to
991 limit the amount of memory a process may take, while retaining a sane
992 behaviour. When this limit is reached, sessions which need a buffer wait for
993 another one to be released by another session. Since buffers are dynamically
994 allocated and released, the waiting time is very short and not perceptible
995 provided that limits remain reasonable. In fact sometimes reducing the limit
996 may even increase performance by increasing the CPU cache's efficiency. Tests
997 have shown good results on average HTTP traffic with a limit to 1/10 of the
998 expected global maxconn setting, which also significantly reduces memory
999 usage. The memory savings come from the fact that a number of connections
1000 will not allocate 2*tune.bufsize. It is best not to touch this value unless
1001 advised to do so by an haproxy core developer.
1002
Willy Tarreau1058ae72014-12-23 22:40:40 +01001003tune.buffers.reserve <number>
1004 Sets the number of buffers which are pre-allocated and reserved for use only
1005 during memory shortage conditions resulting in failed memory allocations. The
1006 minimum value is 2 and is also the default. There is no reason a user would
1007 want to change this value, it's mostly aimed at haproxy core developers.
1008
Willy Tarreau27a674e2009-08-17 07:23:33 +02001009tune.bufsize <number>
1010 Sets the buffer size to this size (in bytes). Lower values allow more
1011 sessions to coexist in the same amount of RAM, and higher values allow some
1012 applications with very large cookies to work. The default value is 16384 and
1013 can be changed at build time. It is strongly recommended not to change this
1014 from the default value, as very low values will break some services such as
1015 statistics, and values larger than default size will increase memory usage,
1016 possibly causing the system to run out of memory. At least the global maxconn
1017 parameter should be decreased by the same factor as this one is increased.
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04001018 If HTTP request is larger than (tune.bufsize - tune.maxrewrite), haproxy will
1019 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
1020 than this size, haproxy will return HTTP 502 (Bad Gateway).
Willy Tarreau27a674e2009-08-17 07:23:33 +02001021
Willy Tarreau43961d52010-10-04 20:39:20 +02001022tune.chksize <number>
1023 Sets the check buffer size to this size (in bytes). Higher values may help
1024 find string or regex patterns in very large pages, though doing so may imply
1025 more memory and CPU usage. The default value is 16384 and can be changed at
1026 build time. It is not recommended to change this value, but to use better
1027 checks whenever possible.
1028
William Lallemandf3747832012-11-09 12:33:10 +01001029tune.comp.maxlevel <number>
1030 Sets the maximum compression level. The compression level affects CPU
1031 usage during compression. This value affects CPU usage during compression.
1032 Each session using compression initializes the compression algorithm with
1033 this value. The default value is 1.
1034
Willy Tarreau193b8c62012-11-22 00:17:38 +01001035tune.http.cookielen <number>
1036 Sets the maximum length of captured cookies. This is the maximum value that
1037 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
1038 will automatically be truncated to this one. It is important not to set too
1039 high a value because all cookie captures still allocate this size whatever
1040 their configured value (they share a same pool). This value is per request
1041 per response, so the memory allocated is twice this value per connection.
1042 When not specified, the limit is set to 63 characters. It is recommended not
1043 to change this value.
1044
Willy Tarreauac1932d2011-10-24 19:14:41 +02001045tune.http.maxhdr <number>
1046 Sets the maximum number of headers in a request. When a request comes with a
1047 number of headers greater than this value (including the first line), it is
1048 rejected with a "400 Bad Request" status code. Similarly, too large responses
1049 are blocked with "502 Bad Gateway". The default value is 101, which is enough
1050 for all usages, considering that the widely deployed Apache server uses the
1051 same limit. It can be useful to push this limit further to temporarily allow
1052 a buggy application to work by the time it gets fixed. Keep in mind that each
1053 new header consumes 32bits of memory for each session, so don't push this
1054 limit too high.
1055
Willy Tarreau7e312732014-02-12 16:35:14 +01001056tune.idletimer <timeout>
1057 Sets the duration after which haproxy will consider that an empty buffer is
1058 probably associated with an idle stream. This is used to optimally adjust
1059 some packet sizes while forwarding large and small data alternatively. The
1060 decision to use splice() or to send large buffers in SSL is modulated by this
1061 parameter. The value is in milliseconds between 0 and 65535. A value of zero
1062 means that haproxy will not try to detect idle streams. The default is 1000,
1063 which seems to correctly detect end user pauses (eg: read a page before
1064 clicking). There should be not reason for changing this value. Please check
1065 tune.ssl.maxrecord below.
1066
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001067tune.lua.forced-yield <number>
1068 This directive forces the Lua engine to execute a yield each <number> of
1069 instructions executed. This permits interruptng a long script and allows the
1070 HAProxy scheduler to process other tasks like accepting connections or
1071 forwarding traffic. The default value is 10000 instructions. If HAProxy often
1072 executes some Lua code but more reactivity is required, this value can be
1073 lowered. If the Lua code is quite long and its result is absolutely required
1074 to process the data, the <number> can be increased.
1075
Willy Tarreau32f61e22015-03-18 17:54:59 +01001076tune.lua.maxmem
1077 Sets the maximum amount of RAM in megabytes per process usable by Lua. By
1078 default it is zero which means unlimited. It is important to set a limit to
1079 ensure that a bug in a script will not result in the system running out of
1080 memory.
1081
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001082tune.lua.session-timeout <timeout>
1083 This is the execution timeout for the Lua sessions. This is useful for
1084 preventing infinite loops or spending too much time in Lua. This timeout has a
1085 priority over other timeouts. For example, if this timeout is set to 4s and
1086 you run a 5s sleep, the code will be interrupted with an error after waiting
1087 4s.
1088
1089tune.lua.task-timeout <timeout>
1090 Purpose is the same as "tune.lua.session-timeout", but this timeout is
1091 dedicated to the tasks. By default, this timeout isn't set because a task may
1092 remain alive during of the lifetime of HAProxy. For example, a task used to
1093 check servers.
1094
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001095tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +01001096 Sets the maximum number of consecutive connections a process may accept in a
1097 row before switching to other work. In single process mode, higher numbers
1098 give better performance at high connection rates. However in multi-process
1099 modes, keeping a bit of fairness between processes generally is better to
1100 increase performance. This value applies individually to each listener, so
1101 that the number of processes a listener is bound to is taken into account.
1102 This value defaults to 64. In multi-process mode, it is divided by twice
1103 the number of processes the listener is bound to. Setting this value to -1
1104 completely disables the limitation. It should normally not be needed to tweak
1105 this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001106
1107tune.maxpollevents <number>
1108 Sets the maximum amount of events that can be processed at once in a call to
1109 the polling system. The default value is adapted to the operating system. It
1110 has been noticed that reducing it below 200 tends to slightly decrease
1111 latency at the expense of network bandwidth, and increasing it above 200
1112 tends to trade latency for slightly increased bandwidth.
1113
Willy Tarreau27a674e2009-08-17 07:23:33 +02001114tune.maxrewrite <number>
1115 Sets the reserved buffer space to this size in bytes. The reserved space is
1116 used for header rewriting or appending. The first reads on sockets will never
1117 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
1118 bufsize, though that does not make much sense since there are rarely large
1119 numbers of headers to add. Setting it too high prevents processing of large
1120 requests or responses. Setting it too low prevents addition of new headers
1121 to already large requests or to POST requests. It is generally wise to set it
1122 to about 1024. It is automatically readjusted to half of bufsize if it is
1123 larger than that. This means you don't have to worry about it when changing
1124 bufsize.
1125
Willy Tarreauf3045d22015-04-29 16:24:50 +02001126tune.pattern.cache-size <number>
1127 Sets the size of the pattern lookup cache to <number> entries. This is an LRU
1128 cache which reminds previous lookups and their results. It is used by ACLs
1129 and maps on slow pattern lookups, namely the ones using the "sub", "reg",
1130 "dir", "dom", "end", "bin" match methods as well as the case-insensitive
1131 strings. It applies to pattern expressions which means that it will be able
1132 to memorize the result of a lookup among all the patterns specified on a
1133 configuration line (including all those loaded from files). It automatically
1134 invalidates entries which are updated using HTTP actions or on the CLI. The
1135 default cache size is set to 10000 entries, which limits its footprint to
1136 about 5 MB on 32-bit systems and 8 MB on 64-bit systems. There is a very low
1137 risk of collision in this cache, which is in the order of the size of the
1138 cache divided by 2^64. Typically, at 10000 requests per second with the
1139 default cache size of 10000 entries, there's 1% chance that a brute force
1140 attack could cause a single collision after 60 years, or 0.1% after 6 years.
1141 This is considered much lower than the risk of a memory corruption caused by
1142 aging components. If this is not acceptable, the cache can be disabled by
1143 setting this parameter to 0.
1144
Willy Tarreaubd9a0a72011-10-23 21:14:29 +02001145tune.pipesize <number>
1146 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
1147 are the default size for the system. But sometimes when using TCP splicing,
1148 it can improve performance to increase pipe sizes, especially if it is
1149 suspected that pipes are not filled and that many calls to splice() are
1150 performed. This has an impact on the kernel's memory footprint, so this must
1151 not be changed if impacts are not understood.
1152
Willy Tarreaue803de22010-01-21 17:43:04 +01001153tune.rcvbuf.client <number>
1154tune.rcvbuf.server <number>
1155 Forces the kernel socket receive buffer size on the client or the server side
1156 to the specified value in bytes. This value applies to all TCP/HTTP frontends
1157 and backends. It should normally never be set, and the default size (0) lets
1158 the kernel autotune this value depending on the amount of available memory.
1159 However it can sometimes help to set it to very low values (eg: 4096) in
1160 order to save kernel memory by preventing it from buffering too large amounts
1161 of received data. Lower values will significantly increase CPU usage though.
1162
1163tune.sndbuf.client <number>
1164tune.sndbuf.server <number>
1165 Forces the kernel socket send buffer size on the client or the server side to
1166 the specified value in bytes. This value applies to all TCP/HTTP frontends
1167 and backends. It should normally never be set, and the default size (0) lets
1168 the kernel autotune this value depending on the amount of available memory.
1169 However it can sometimes help to set it to very low values (eg: 4096) in
1170 order to save kernel memory by preventing it from buffering too large amounts
1171 of received data. Lower values will significantly increase CPU usage though.
1172 Another use case is to prevent write timeouts with extremely slow clients due
1173 to the kernel waiting for a large part of the buffer to be read before
1174 notifying haproxy again.
1175
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001176tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +01001177 Sets the size of the global SSL session cache, in a number of blocks. A block
1178 is large enough to contain an encoded session without peer certificate.
1179 An encoded session with peer certificate is stored in multiple blocks
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001180 depending on the size of the peer certificate. A block uses approximately
Emeric Brunaf9619d2012-11-28 18:47:52 +01001181 200 bytes of memory. The default value may be forced at build time, otherwise
1182 defaults to 20000. When the cache is full, the most idle entries are purged
1183 and reassigned. Higher values reduce the occurrence of such a purge, hence
1184 the number of CPU-intensive SSL handshakes by ensuring that all users keep
1185 their session as long as possible. All entries are pre-allocated upon startup
Emeric Brun22890a12012-12-28 14:41:32 +01001186 and are shared between all processes if "nbproc" is greater than 1. Setting
1187 this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001188
Emeric Brun8dc60392014-05-09 13:52:00 +02001189tune.ssl.force-private-cache
1190 This boolean disables SSL session cache sharing between all processes. It
1191 should normally not be used since it will force many renegotiations due to
1192 clients hitting a random process. But it may be required on some operating
1193 systems where none of the SSL cache synchronization method may be used. In
1194 this case, adding a first layer of hash-based load balancing before the SSL
1195 layer might limit the impact of the lack of session sharing.
1196
Emeric Brun4f65bff2012-11-16 15:11:00 +01001197tune.ssl.lifetime <timeout>
1198 Sets how long a cached SSL session may remain valid. This time is expressed
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001199 in seconds and defaults to 300 (5 min). It is important to understand that it
Emeric Brun4f65bff2012-11-16 15:11:00 +01001200 does not guarantee that sessions will last that long, because if the cache is
1201 full, the longest idle sessions will be purged despite their configured
1202 lifetime. The real usefulness of this setting is to prevent sessions from
1203 being used for too long.
1204
Willy Tarreaubfd59462013-02-21 07:46:09 +01001205tune.ssl.maxrecord <number>
1206 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
1207 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
1208 data only once it has received a full record. With large records, it means
1209 that clients might have to download up to 16kB of data before starting to
1210 process them. Limiting the value can improve page load times on browsers
1211 located over high latency or low bandwidth networks. It is suggested to find
1212 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
1213 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
1214 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
1215 2859 gave good results during tests. Use "strace -e trace=write" to find the
Willy Tarreau7e312732014-02-12 16:35:14 +01001216 best value. Haproxy will automatically switch to this setting after an idle
1217 stream has been detected (see tune.idletimer above).
Willy Tarreaubfd59462013-02-21 07:46:09 +01001218
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001219tune.ssl.default-dh-param <number>
1220 Sets the maximum size of the Diffie-Hellman parameters used for generating
1221 the ephemeral/temporary Diffie-Hellman key in case of DHE key exchange. The
1222 final size will try to match the size of the server's RSA (or DSA) key (e.g,
1223 a 2048 bits temporary DH key for a 2048 bits RSA key), but will not exceed
1224 this maximum value. Default value if 1024. Only 1024 or higher values are
1225 allowed. Higher values will increase the CPU load, and values greater than
1226 1024 bits are not supported by Java 7 and earlier clients. This value is not
1227 used if static Diffie-Hellman parameters are supplied via the certificate file.
1228
William Lallemanda509e4c2012-11-07 16:54:34 +01001229tune.zlib.memlevel <number>
1230 Sets the memLevel parameter in zlib initialization for each session. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001231 defines how much memory should be allocated for the internal compression
William Lallemanda509e4c2012-11-07 16:54:34 +01001232 state. A value of 1 uses minimum memory but is slow and reduces compression
1233 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
1234 between 1 and 9. The default value is 8.
1235
1236tune.zlib.windowsize <number>
1237 Sets the window size (the size of the history buffer) as a parameter of the
1238 zlib initialization for each session. Larger values of this parameter result
1239 in better compression at the expense of memory usage. Can be a value between
1240 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001241
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012423.3. Debugging
1243--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02001244
1245debug
1246 Enables debug mode which dumps to stdout all exchanges, and disables forking
1247 into background. It is the equivalent of the command-line argument "-d". It
1248 should never be used in a production configuration since it may prevent full
1249 system startup.
1250
1251quiet
1252 Do not display any message during startup. It is equivalent to the command-
1253 line argument "-q".
1254
Emeric Brunf099e792010-09-27 12:05:28 +02001255
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010012563.4. Userlists
1257--------------
1258It is possible to control access to frontend/backend/listen sections or to
1259http stats by allowing only authenticated and authorized users. To do this,
1260it is required to create at least one userlist and to define users.
1261
1262userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +01001263 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001264 used to store authentication & authorization data for independent customers.
1265
1266group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +01001267 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001268 attach users to this group by using a comma separated list of names
1269 proceeded by "users" keyword.
1270
Cyril Bontéf0c60612010-02-06 14:44:47 +01001271user <username> [password|insecure-password <password>]
1272 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001273 Adds user <username> to the current userlist. Both secure (encrypted) and
1274 insecure (unencrypted) passwords can be used. Encrypted passwords are
Cyril Bonté78caf842010-03-10 22:41:43 +01001275 evaluated using the crypt(3) function so depending of the system's
1276 capabilities, different algorithms are supported. For example modern Glibc
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001277 based Linux system supports MD5, SHA-256, SHA-512 and of course classic,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001278 DES-based method of encrypting passwords.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001279
1280
1281 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +01001282 userlist L1
1283 group G1 users tiger,scott
1284 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001285
Cyril Bontéf0c60612010-02-06 14:44:47 +01001286 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
1287 user scott insecure-password elgato
1288 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001289
Cyril Bontéf0c60612010-02-06 14:44:47 +01001290 userlist L2
1291 group G1
1292 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001293
Cyril Bontéf0c60612010-02-06 14:44:47 +01001294 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
1295 user scott insecure-password elgato groups G1,G2
1296 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001297
1298 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001299
Emeric Brunf099e792010-09-27 12:05:28 +02001300
13013.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02001302----------
Emeric Brunf099e792010-09-27 12:05:28 +02001303It is possible to synchronize server entries in stick tables between several
1304haproxy instances over TCP connections in a multi-master fashion. Each instance
1305pushes its local updates and insertions to remote peers. Server IDs are used to
1306identify servers remotely, so it is important that configurations look similar
1307or at least that the same IDs are forced on each server on all participants.
1308Interrupted exchanges are automatically detected and recovered from the last
1309known point. In addition, during a soft restart, the old process connects to
1310the new one using such a TCP connection to push all its entries before the new
1311process tries to connect to other peers. That ensures very fast replication
1312during a reload, it typically takes a fraction of a second even for large
1313tables.
1314
1315peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001316 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02001317 which is referenced by one or more stick-tables.
1318
Willy Tarreau77e4bd12015-05-01 20:02:17 +02001319disabled
1320 Disables a peers section. It disables both listening and any synchronization
1321 related to this section. This is provided to disable synchronization of stick
1322 tables without having to comment out all "peers" references.
1323
1324enable
1325 This re-enables a disabled peers section which was previously disabled.
1326
Emeric Brunf099e792010-09-27 12:05:28 +02001327peer <peername> <ip>:<port>
1328 Defines a peer inside a peers section.
1329 If <peername> is set to the local peer name (by default hostname, or forced
1330 using "-L" command line option), haproxy will listen for incoming remote peer
1331 connection on <ip>:<port>. Otherwise, <ip>:<port> defines where to connect to
1332 to join the remote peer, and <peername> is used at the protocol level to
1333 identify and validate the remote peer on the server side.
1334
1335 During a soft restart, local peer <ip>:<port> is used by the old instance to
1336 connect the new one and initiate a complete replication (teaching process).
1337
1338 It is strongly recommended to have the exact same peers declaration on all
1339 peers and to only rely on the "-L" command line argument to change the local
1340 peer name. This makes it easier to maintain coherent configuration files
1341 across all peers.
1342
William Lallemandb2f07452015-05-12 14:27:13 +02001343 You may want to reference some environment variables in the address
1344 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01001345
Cyril Bontédc4d9032012-04-08 21:57:39 +02001346 Example:
Emeric Brunf099e792010-09-27 12:05:28 +02001347 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001348 peer haproxy1 192.168.0.1:1024
1349 peer haproxy2 192.168.0.2:1024
1350 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02001351
1352 backend mybackend
1353 mode tcp
1354 balance roundrobin
1355 stick-table type ip size 20k peers mypeers
1356 stick on src
1357
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001358 server srv1 192.168.0.30:80
1359 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02001360
1361
Simon Horman51a1cf62015-02-03 13:00:44 +090013623.6. Mailers
1363------------
1364It is possible to send email alerts when the state of servers changes.
1365If configured email alerts are sent to each mailer that is configured
1366in a mailers section. Email is sent to mailers using SMTP.
1367
1368mailer <mailersect>
1369 Creates a new mailer list with the name <mailersect>. It is an
1370 independent section which is referenced by one or more proxies.
1371
1372mailer <mailername> <ip>:<port>
1373 Defines a mailer inside a mailers section.
1374
1375 Example:
1376 mailers mymailers
1377 mailer smtp1 192.168.0.1:587
1378 mailer smtp2 192.168.0.2:587
1379
1380 backend mybackend
1381 mode tcp
1382 balance roundrobin
1383
1384 email-alert mailers mymailers
1385 email-alert from test1@horms.org
1386 email-alert to test2@horms.org
1387
1388 server srv1 192.168.0.30:80
1389 server srv2 192.168.0.31:80
1390
1391
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013924. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02001393----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001394
Willy Tarreau6a06a402007-07-15 20:15:28 +02001395Proxy configuration can be located in a set of sections :
1396 - defaults <name>
1397 - frontend <name>
1398 - backend <name>
1399 - listen <name>
1400
1401A "defaults" section sets default parameters for all other sections following
1402its declaration. Those default parameters are reset by the next "defaults"
1403section. See below for the list of parameters which can be set in a "defaults"
Willy Tarreau0ba27502007-12-24 16:55:16 +01001404section. The name is optional but its use is encouraged for better readability.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001405
1406A "frontend" section describes a set of listening sockets accepting client
1407connections.
1408
1409A "backend" section describes a set of servers to which the proxy will connect
1410to forward incoming connections.
1411
1412A "listen" section defines a complete proxy with its frontend and backend
1413parts combined in one section. It is generally useful for TCP-only traffic.
1414
Willy Tarreau0ba27502007-12-24 16:55:16 +01001415All proxy names must be formed from upper and lower case letters, digits,
1416'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
1417case-sensitive, which means that "www" and "WWW" are two different proxies.
1418
1419Historically, all proxy names could overlap, it just caused troubles in the
1420logs. Since the introduction of content switching, it is mandatory that two
1421proxies with overlapping capabilities (frontend/backend) have different names.
1422However, it is still permitted that a frontend and a backend share the same
1423name, as this configuration seems to be commonly encountered.
1424
1425Right now, two major proxy modes are supported : "tcp", also known as layer 4,
1426and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001427bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01001428protocol, and can interact with it by allowing, blocking, switching, adding,
1429modifying, or removing arbitrary contents in requests or responses, based on
1430arbitrary criteria.
1431
Willy Tarreau70dffda2014-01-30 03:07:23 +01001432In HTTP mode, the processing applied to requests and responses flowing over
1433a connection depends in the combination of the frontend's HTTP options and
1434the backend's. HAProxy supports 5 connection modes :
1435
1436 - KAL : keep alive ("option http-keep-alive") which is the default mode : all
1437 requests and responses are processed, and connections remain open but idle
1438 between responses and new requests.
1439
1440 - TUN: tunnel ("option http-tunnel") : this was the default mode for versions
1441 1.0 to 1.5-dev21 : only the first request and response are processed, and
1442 everything else is forwarded with no analysis at all. This mode should not
1443 be used as it creates lots of trouble with logging and HTTP processing.
1444
1445 - PCL: passive close ("option httpclose") : exactly the same as tunnel mode,
1446 but with "Connection: close" appended in both directions to try to make
1447 both ends close after the first request/response exchange.
1448
1449 - SCL: server close ("option http-server-close") : the server-facing
1450 connection is closed after the end of the response is received, but the
1451 client-facing connection remains open.
1452
1453 - FCL: forced close ("option forceclose") : the connection is actively closed
1454 after the end of the response.
1455
1456The effective mode that will be applied to a connection passing through a
1457frontend and a backend can be determined by both proxy modes according to the
1458following matrix, but in short, the modes are symmetric, keep-alive is the
1459weakest option and force close is the strongest.
1460
1461 Backend mode
1462
1463 | KAL | TUN | PCL | SCL | FCL
1464 ----+-----+-----+-----+-----+----
1465 KAL | KAL | TUN | PCL | SCL | FCL
1466 ----+-----+-----+-----+-----+----
1467 TUN | TUN | TUN | PCL | SCL | FCL
1468 Frontend ----+-----+-----+-----+-----+----
1469 mode PCL | PCL | PCL | PCL | FCL | FCL
1470 ----+-----+-----+-----+-----+----
1471 SCL | SCL | SCL | FCL | SCL | FCL
1472 ----+-----+-----+-----+-----+----
1473 FCL | FCL | FCL | FCL | FCL | FCL
1474
Willy Tarreau0ba27502007-12-24 16:55:16 +01001475
Willy Tarreau70dffda2014-01-30 03:07:23 +01001476
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014774.1. Proxy keywords matrix
1478--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001479
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001480The following list of keywords is supported. Most of them may only be used in a
1481limited set of section types. Some of them are marked as "deprecated" because
1482they are inherited from an old syntax which may be confusing or functionally
1483limited, and there are new recommended keywords to replace them. Keywords
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001484marked with "(*)" can be optionally inverted using the "no" prefix, eg. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001485option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02001486and must be disabled for a specific instance. Such options may also be prefixed
1487with "default" in order to restore default settings regardless of what has been
1488specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001489
Willy Tarreau6a06a402007-07-15 20:15:28 +02001490
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001491 keyword defaults frontend listen backend
1492------------------------------------+----------+----------+---------+---------
1493acl - X X X
1494appsession - - X X
1495backlog X X X -
1496balance X - X X
1497bind - X X -
1498bind-process X X X X
1499block - X X X
1500capture cookie - X X -
1501capture request header - X X -
1502capture response header - X X -
1503clitimeout (deprecated) X X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02001504compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001505contimeout (deprecated) X - X X
1506cookie X - X X
1507default-server X - X X
1508default_backend X X X -
1509description - X X X
1510disabled X X X X
1511dispatch - - X X
Simon Horman51a1cf62015-02-03 13:00:44 +09001512email-alert from X X X X
Simon Horman64e34162015-02-06 11:11:57 +09001513email-alert level X X X X
Simon Horman51a1cf62015-02-03 13:00:44 +09001514email-alert mailers X X X X
1515email-alert myhostname X X X X
1516email-alert to X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001517enabled X X X X
1518errorfile X X X X
1519errorloc X X X X
1520errorloc302 X X X X
1521-- keyword -------------------------- defaults - frontend - listen -- backend -
1522errorloc303 X X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001523force-persist - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001524fullconn X - X X
1525grace X X X X
1526hash-type X - X X
1527http-check disable-on-404 X - X X
Willy Tarreaubd741542010-03-16 18:46:54 +01001528http-check expect - - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02001529http-check send-state X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001530http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02001531http-response - X X X
Baptiste Assmann2c42ef52013-10-09 21:57:02 +02001532http-send-name-header - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001533id - X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001534ignore-persist - X X X
William Lallemand0f99e342011-10-12 17:50:54 +02001535log (*) X X X X
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01001536log-format X X X -
Willy Tarreau094af4e2015-01-07 15:03:42 +01001537log-tag X X X X
Willy Tarreauc35362a2014-04-25 13:58:37 +02001538max-keep-alive-queue X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001539maxconn X X X -
1540mode X X X X
1541monitor fail - X X -
1542monitor-net X X X -
1543monitor-uri X X X -
1544option abortonclose (*) X - X X
1545option accept-invalid-http-request (*) X X X -
1546option accept-invalid-http-response (*) X - X X
1547option allbackups (*) X - X X
1548option checkcache (*) X - X X
1549option clitcpka (*) X X X -
1550option contstats (*) X X X -
1551option dontlog-normal (*) X X X -
1552option dontlognull (*) X X X -
1553option forceclose (*) X X X X
1554-- keyword -------------------------- defaults - frontend - listen -- backend -
1555option forwardfor X X X X
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02001556option http-buffer-request (*) X X X X
Willy Tarreau82649f92015-05-01 22:40:51 +02001557option http-ignore-probes (*) X X X -
Willy Tarreau16bfb022010-01-16 19:48:41 +01001558option http-keep-alive (*) X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02001559option http-no-delay (*) X X X X
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02001560option http-pretend-keepalive (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001561option http-server-close (*) X X X X
Willy Tarreau02bce8b2014-01-30 00:15:28 +01001562option http-tunnel (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001563option http-use-proxy-header (*) X X X -
1564option httpchk X - X X
1565option httpclose (*) X X X X
1566option httplog X X X X
1567option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001568option independent-streams (*) X X X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02001569option ldap-check X - X X
Simon Horman98637e52014-06-20 12:30:16 +09001570option external-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001571option log-health-checks (*) X - X X
1572option log-separate-errors (*) X X X -
1573option logasap (*) X X X -
1574option mysql-check X - X X
Rauf Kuliyev38b41562011-01-04 15:14:13 +01001575option pgsql-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001576option nolinger (*) X X X X
1577option originalto X X X X
1578option persist (*) X - X X
1579option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02001580option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001581option smtpchk X - X X
1582option socket-stats (*) X X X -
1583option splice-auto (*) X X X X
1584option splice-request (*) X X X X
1585option splice-response (*) X X X X
1586option srvtcpka (*) X - X X
1587option ssl-hello-chk X - X X
1588-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreaued179852013-12-16 01:07:00 +01001589option tcp-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001590option tcp-smart-accept (*) X X X -
1591option tcp-smart-connect (*) X - X X
1592option tcpka X X X X
1593option tcplog X X X X
1594option transparent (*) X - X X
Simon Horman98637e52014-06-20 12:30:16 +09001595external-check command X - X X
1596external-check path X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001597persist rdp-cookie X - X X
1598rate-limit sessions X X X -
1599redirect - X X X
1600redisp (deprecated) X - X X
1601redispatch (deprecated) X - X X
1602reqadd - X X X
1603reqallow - X X X
1604reqdel - X X X
1605reqdeny - X X X
1606reqiallow - X X X
1607reqidel - X X X
1608reqideny - X X X
1609reqipass - X X X
1610reqirep - X X X
1611reqisetbe - X X X
1612reqitarpit - X X X
1613reqpass - X X X
1614reqrep - X X X
1615-- keyword -------------------------- defaults - frontend - listen -- backend -
1616reqsetbe - X X X
1617reqtarpit - X X X
1618retries X - X X
1619rspadd - X X X
1620rspdel - X X X
1621rspdeny - X X X
1622rspidel - X X X
1623rspideny - X X X
1624rspirep - X X X
1625rsprep - X X X
1626server - - X X
1627source X - X X
1628srvtimeout (deprecated) X - X X
Cyril Bonté66c327d2010-10-12 00:14:37 +02001629stats admin - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001630stats auth X - X X
1631stats enable X - X X
1632stats hide-version X - X X
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02001633stats http-request - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001634stats realm X - X X
1635stats refresh X - X X
1636stats scope X - X X
1637stats show-desc X - X X
1638stats show-legends X - X X
1639stats show-node X - X X
1640stats uri X - X X
1641-- keyword -------------------------- defaults - frontend - listen -- backend -
1642stick match - - X X
1643stick on - - X X
1644stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02001645stick store-response - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001646stick-table - - X X
Willy Tarreau938c7fe2014-04-25 14:21:39 +02001647tcp-check connect - - X X
1648tcp-check expect - - X X
1649tcp-check send - - X X
1650tcp-check send-binary - - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02001651tcp-request connection - X X -
1652tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02001653tcp-request inspect-delay - X X X
Emeric Brun0a3b67f2010-09-24 15:34:53 +02001654tcp-response content - - X X
1655tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001656timeout check X - X X
1657timeout client X X X -
Willy Tarreau05cdd962014-05-10 14:30:07 +02001658timeout client-fin X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001659timeout clitimeout (deprecated) X X X -
1660timeout connect X - X X
1661timeout contimeout (deprecated) X - X X
1662timeout http-keep-alive X X X X
1663timeout http-request X X X X
1664timeout queue X - X X
1665timeout server X - X X
Willy Tarreau05cdd962014-05-10 14:30:07 +02001666timeout server-fin X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001667timeout srvtimeout (deprecated) X - X X
1668timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02001669timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001670transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01001671unique-id-format X X X -
1672unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001673use_backend - X X -
Willy Tarreau4a5cade2012-04-05 21:09:48 +02001674use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001675------------------------------------+----------+----------+---------+---------
1676 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02001677
Willy Tarreau0ba27502007-12-24 16:55:16 +01001678
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016794.2. Alphabetically sorted keywords reference
1680---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001681
1682This section provides a description of each keyword and its usage.
1683
1684
1685acl <aclname> <criterion> [flags] [operator] <value> ...
1686 Declare or complete an access list.
1687 May be used in sections : defaults | frontend | listen | backend
1688 no | yes | yes | yes
1689 Example:
1690 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
1691 acl invalid_src src_port 0:1023
1692 acl local_dst hdr(host) -i localhost
1693
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001694 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001695
1696
Cyril Bontéb21570a2009-11-29 20:04:48 +01001697appsession <cookie> len <length> timeout <holdtime>
1698 [request-learn] [prefix] [mode <path-parameters|query-string>]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001699 Define session stickiness on an existing application cookie.
1700 May be used in sections : defaults | frontend | listen | backend
1701 no | no | yes | yes
1702 Arguments :
1703 <cookie> this is the name of the cookie used by the application and which
1704 HAProxy will have to learn for each new session.
1705
Cyril Bontéb21570a2009-11-29 20:04:48 +01001706 <length> this is the max number of characters that will be memorized and
Willy Tarreau0ba27502007-12-24 16:55:16 +01001707 checked in each cookie value.
1708
1709 <holdtime> this is the time after which the cookie will be removed from
1710 memory if unused. If no unit is specified, this time is in
1711 milliseconds.
1712
Cyril Bontébf47aeb2009-10-15 00:15:40 +02001713 request-learn
1714 If this option is specified, then haproxy will be able to learn
1715 the cookie found in the request in case the server does not
1716 specify any in response. This is typically what happens with
1717 PHPSESSID cookies, or when haproxy's session expires before
1718 the application's session and the correct server is selected.
1719 It is recommended to specify this option to improve reliability.
1720
Cyril Bontéb21570a2009-11-29 20:04:48 +01001721 prefix When this option is specified, haproxy will match on the cookie
1722 prefix (or URL parameter prefix). The appsession value is the
1723 data following this prefix.
1724
1725 Example :
1726 appsession ASPSESSIONID len 64 timeout 3h prefix
1727
1728 This will match the cookie ASPSESSIONIDXXXX=XXXXX,
1729 the appsession value will be XXXX=XXXXX.
1730
1731 mode This option allows to change the URL parser mode.
1732 2 modes are currently supported :
1733 - path-parameters :
1734 The parser looks for the appsession in the path parameters
1735 part (each parameter is separated by a semi-colon), which is
1736 convenient for JSESSIONID for example.
1737 This is the default mode if the option is not set.
1738 - query-string :
1739 In this mode, the parser will look for the appsession in the
1740 query string.
1741
Willy Tarreau0ba27502007-12-24 16:55:16 +01001742 When an application cookie is defined in a backend, HAProxy will check when
1743 the server sets such a cookie, and will store its value in a table, and
1744 associate it with the server's identifier. Up to <length> characters from
1745 the value will be retained. On each connection, haproxy will look for this
Cyril Bontéb21570a2009-11-29 20:04:48 +01001746 cookie both in the "Cookie:" headers, and as a URL parameter (depending on
1747 the mode used). If a known value is found, the client will be directed to the
1748 server associated with this value. Otherwise, the load balancing algorithm is
Willy Tarreau0ba27502007-12-24 16:55:16 +01001749 applied. Cookies are automatically removed from memory when they have been
1750 unused for a duration longer than <holdtime>.
1751
1752 The definition of an application cookie is limited to one per backend.
1753
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01001754 Note : Consider not using this feature in multi-process mode (nbproc > 1)
1755 unless you know what you do : memory is not shared between the
1756 processes, which can result in random behaviours.
1757
Willy Tarreau0ba27502007-12-24 16:55:16 +01001758 Example :
1759 appsession JSESSIONID len 52 timeout 3h
1760
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01001761 See also : "cookie", "capture cookie", "balance", "stick", "stick-table",
1762 "ignore-persist", "nbproc" and "bind-process".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001763
1764
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01001765backlog <conns>
1766 Give hints to the system about the approximate listen backlog desired size
1767 May be used in sections : defaults | frontend | listen | backend
1768 yes | yes | yes | no
1769 Arguments :
1770 <conns> is the number of pending connections. Depending on the operating
1771 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02001772 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01001773
1774 In order to protect against SYN flood attacks, one solution is to increase
1775 the system's SYN backlog size. Depending on the system, sometimes it is just
1776 tunable via a system parameter, sometimes it is not adjustable at all, and
1777 sometimes the system relies on hints given by the application at the time of
1778 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
1779 to the listen() syscall. On systems which can make use of this value, it can
1780 sometimes be useful to be able to specify a different value, hence this
1781 backlog parameter.
1782
1783 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
1784 used as a hint and the system accepts up to the smallest greater power of
1785 two, and never more than some limits (usually 32768).
1786
1787 See also : "maxconn" and the target operating system's tuning guide.
1788
1789
Willy Tarreau0ba27502007-12-24 16:55:16 +01001790balance <algorithm> [ <arguments> ]
Willy Tarreau226071e2014-04-10 11:55:45 +02001791balance url_param <param> [check_post]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001792 Define the load balancing algorithm to be used in a backend.
1793 May be used in sections : defaults | frontend | listen | backend
1794 yes | no | yes | yes
1795 Arguments :
1796 <algorithm> is the algorithm used to select a server when doing load
1797 balancing. This only applies when no persistence information
1798 is available, or when a connection is redispatched to another
1799 server. <algorithm> may be one of the following :
1800
1801 roundrobin Each server is used in turns, according to their weights.
1802 This is the smoothest and fairest algorithm when the server's
1803 processing time remains equally distributed. This algorithm
1804 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02001805 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08001806 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02001807 large farms, when a server becomes up after having been down
1808 for a very short time, it may sometimes take a few hundreds
1809 requests for it to be re-integrated into the farm and start
1810 receiving traffic. This is normal, though very rare. It is
1811 indicated here in case you would have the chance to observe
1812 it, so that you don't worry.
1813
1814 static-rr Each server is used in turns, according to their weights.
1815 This algorithm is as similar to roundrobin except that it is
1816 static, which means that changing a server's weight on the
1817 fly will have no effect. On the other hand, it has no design
1818 limitation on the number of servers, and when a server goes
1819 up, it is always immediately reintroduced into the farm, once
1820 the full map is recomputed. It also uses slightly less CPU to
1821 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01001822
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01001823 leastconn The server with the lowest number of connections receives the
1824 connection. Round-robin is performed within groups of servers
1825 of the same load to ensure that all servers will be used. Use
1826 of this algorithm is recommended where very long sessions are
1827 expected, such as LDAP, SQL, TSE, etc... but is not very well
1828 suited for protocols using short sessions such as HTTP. This
1829 algorithm is dynamic, which means that server weights may be
1830 adjusted on the fly for slow starts for instance.
1831
Willy Tarreauf09c6602012-02-13 17:12:08 +01001832 first The first server with available connection slots receives the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001833 connection. The servers are chosen from the lowest numeric
Willy Tarreauf09c6602012-02-13 17:12:08 +01001834 identifier to the highest (see server parameter "id"), which
1835 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02001836 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01001837 not make sense to use this algorithm without setting maxconn.
1838 The purpose of this algorithm is to always use the smallest
1839 number of servers so that extra servers can be powered off
1840 during non-intensive hours. This algorithm ignores the server
1841 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02001842 or IMAP than HTTP, though it can be useful there too. In
1843 order to use this algorithm efficiently, it is recommended
1844 that a cloud controller regularly checks server usage to turn
1845 them off when unused, and regularly checks backend queue to
1846 turn new servers on when the queue inflates. Alternatively,
1847 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01001848
Willy Tarreau0ba27502007-12-24 16:55:16 +01001849 source The source IP address is hashed and divided by the total
1850 weight of the running servers to designate which server will
1851 receive the request. This ensures that the same client IP
1852 address will always reach the same server as long as no
1853 server goes down or up. If the hash result changes due to the
1854 number of running servers changing, many clients will be
1855 directed to a different server. This algorithm is generally
1856 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001857 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01001858 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001859 static by default, which means that changing a server's
1860 weight on the fly will have no effect, but this can be
1861 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001862
Oskar Stolc8dc41842012-05-19 10:19:54 +01001863 uri This algorithm hashes either the left part of the URI (before
1864 the question mark) or the whole URI (if the "whole" parameter
1865 is present) and divides the hash value by the total weight of
1866 the running servers. The result designates which server will
1867 receive the request. This ensures that the same URI will
1868 always be directed to the same server as long as no server
1869 goes up or down. This is used with proxy caches and
1870 anti-virus proxies in order to maximize the cache hit rate.
1871 Note that this algorithm may only be used in an HTTP backend.
1872 This algorithm is static by default, which means that
1873 changing a server's weight on the fly will have no effect,
1874 but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001875
Oskar Stolc8dc41842012-05-19 10:19:54 +01001876 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02001877 "depth", both followed by a positive integer number. These
1878 options may be helpful when it is needed to balance servers
1879 based on the beginning of the URI only. The "len" parameter
1880 indicates that the algorithm should only consider that many
1881 characters at the beginning of the URI to compute the hash.
1882 Note that having "len" set to 1 rarely makes sense since most
1883 URIs start with a leading "/".
1884
1885 The "depth" parameter indicates the maximum directory depth
1886 to be used to compute the hash. One level is counted for each
1887 slash in the request. If both parameters are specified, the
1888 evaluation stops when either is reached.
1889
Willy Tarreau0ba27502007-12-24 16:55:16 +01001890 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001891 the query string of each HTTP GET request.
1892
1893 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02001894 request entity will be searched for the parameter argument,
1895 when it is not found in a query string after a question mark
Willy Tarreau226071e2014-04-10 11:55:45 +02001896 ('?') in the URL. The message body will only start to be
1897 analyzed once either the advertised amount of data has been
1898 received or the request buffer is full. In the unlikely event
1899 that chunked encoding is used, only the first chunk is
Cyril Bontédc4d9032012-04-08 21:57:39 +02001900 scanned. Parameter values separated by a chunk boundary, may
Willy Tarreau226071e2014-04-10 11:55:45 +02001901 be randomly balanced if at all. This keyword used to support
1902 an optional <max_wait> parameter which is now ignored.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001903
1904 If the parameter is found followed by an equal sign ('=') and
1905 a value, then the value is hashed and divided by the total
1906 weight of the running servers. The result designates which
1907 server will receive the request.
1908
1909 This is used to track user identifiers in requests and ensure
1910 that a same user ID will always be sent to the same server as
1911 long as no server goes up or down. If no value is found or if
1912 the parameter is not found, then a round robin algorithm is
1913 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001914 backend. This algorithm is static by default, which means
1915 that changing a server's weight on the fly will have no
1916 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001917
Cyril Bontédc4d9032012-04-08 21:57:39 +02001918 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
1919 request. Just as with the equivalent ACL 'hdr()' function,
1920 the header name in parenthesis is not case sensitive. If the
1921 header is absent or if it does not contain any value, the
1922 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01001923
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001924 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01001925 reducing the hash algorithm to the main domain part with some
1926 specific headers such as 'Host'. For instance, in the Host
1927 value "haproxy.1wt.eu", only "1wt" will be considered.
1928
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001929 This algorithm is static by default, which means that
1930 changing a server's weight on the fly will have no effect,
1931 but this can be changed using "hash-type".
1932
Emeric Brun736aa232009-06-30 17:56:00 +02001933 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02001934 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02001935 The RDP cookie <name> (or "mstshash" if omitted) will be
1936 looked up and hashed for each incoming TCP request. Just as
1937 with the equivalent ACL 'req_rdp_cookie()' function, the name
1938 is not case-sensitive. This mechanism is useful as a degraded
1939 persistence mode, as it makes it possible to always send the
1940 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001941 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02001942 used instead.
1943
1944 Note that for this to work, the frontend must ensure that an
1945 RDP cookie is already present in the request buffer. For this
1946 you must use 'tcp-request content accept' rule combined with
1947 a 'req_rdp_cookie_cnt' ACL.
1948
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001949 This algorithm is static by default, which means that
1950 changing a server's weight on the fly will have no effect,
1951 but this can be changed using "hash-type".
1952
Cyril Bontédc4d9032012-04-08 21:57:39 +02001953 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09001954
Willy Tarreau0ba27502007-12-24 16:55:16 +01001955 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02001956 algorithms. Right now, only "url_param" and "uri" support an
1957 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001958
Willy Tarreau3cd9af22009-03-15 14:06:41 +01001959 The load balancing algorithm of a backend is set to roundrobin when no other
1960 algorithm, mode nor option have been set. The algorithm may only be set once
1961 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001962
1963 Examples :
1964 balance roundrobin
1965 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001966 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01001967 balance hdr(User-Agent)
1968 balance hdr(host)
1969 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001970
1971 Note: the following caveats and limitations on using the "check_post"
1972 extension with "url_param" must be considered :
1973
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001974 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001975 to determine if the parameters will be found in the body or entity which
1976 may contain binary data. Therefore another method may be required to
1977 restrict consideration of POST requests that have no URL parameters in
1978 the body. (see acl reqideny http_end)
1979
1980 - using a <max_wait> value larger than the request buffer size does not
1981 make sense and is useless. The buffer size is set at build time, and
1982 defaults to 16 kB.
1983
1984 - Content-Encoding is not supported, the parameter search will probably
1985 fail; and load balancing will fall back to Round Robin.
1986
1987 - Expect: 100-continue is not supported, load balancing will fall back to
1988 Round Robin.
1989
1990 - Transfer-Encoding (RFC2616 3.6.1) is only supported in the first chunk.
1991 If the entire parameter value is not present in the first chunk, the
1992 selection of server is undefined (actually, defined by how little
1993 actually appeared in the first chunk).
1994
1995 - This feature does not support generation of a 100, 411 or 501 response.
1996
1997 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001998 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001999 white space or control characters are found, indicating the end of what
2000 might be a URL parameter list. This is probably not a concern with SGML
2001 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002002
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002003 See also : "dispatch", "cookie", "appsession", "transparent", "hash-type" and
2004 "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002005
2006
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002007bind [<address>]:<port_range> [, ...] [param*]
2008bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002009 Define one or several listening addresses and/or ports in a frontend.
2010 May be used in sections : defaults | frontend | listen | backend
2011 no | yes | yes | no
2012 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002013 <address> is optional and can be a host name, an IPv4 address, an IPv6
2014 address, or '*'. It designates the address the frontend will
2015 listen on. If unset, all IPv4 addresses of the system will be
2016 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01002017 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01002018 Optionally, an address family prefix may be used before the
2019 address to force the family regardless of the address format,
2020 which can be useful to specify a path to a unix socket with
2021 no slash ('/'). Currently supported prefixes are :
2022 - 'ipv4@' -> address is always IPv4
2023 - 'ipv6@' -> address is always IPv6
2024 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau70f72e02014-07-08 00:37:50 +02002025 - 'abns@' -> address is in abstract namespace (Linux only).
2026 Note: since abstract sockets are not "rebindable", they
2027 do not cope well with multi-process mode during
2028 soft-restart, so it is better to avoid them if
2029 nbproc is greater than 1. The effect is that if the
2030 new process fails to start, only one of the old ones
2031 will be able to rebind to the socket.
Willy Tarreau40aa0702013-03-10 23:51:38 +01002032 - 'fd@<n>' -> use file descriptor <n> inherited from the
2033 parent. The fd must be bound and may or may not already
2034 be listening.
William Lallemandb2f07452015-05-12 14:27:13 +02002035 You may want to reference some environment variables in the
2036 address parameter, see section 2.3 about environment
2037 variables.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002038
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002039 <port_range> is either a unique TCP port, or a port range for which the
2040 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002041 above. The port is mandatory for TCP listeners. Note that in
2042 the case of an IPv6 address, the port is always the number
2043 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002044 - a numerical port (ex: '80')
2045 - a dash-delimited ports range explicitly stating the lower
2046 and upper bounds (ex: '2000-2100') which are included in
2047 the range.
2048
2049 Particular care must be taken against port ranges, because
2050 every <address:port> couple consumes one socket (= a file
2051 descriptor), so it's easy to consume lots of descriptors
2052 with a simple range, and to run out of sockets. Also, each
2053 <address:port> couple must be used only once among all
2054 instances running on a same system. Please note that binding
2055 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002056 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002057 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002058
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002059 <path> is a UNIX socket path beginning with a slash ('/'). This is
2060 alternative to the TCP listening port. Haproxy will then
2061 receive UNIX connections on the socket located at this place.
2062 The path must begin with a slash and by default is absolute.
2063 It can be relative to the prefix defined by "unix-bind" in
2064 the global section. Note that the total length of the prefix
2065 followed by the socket path cannot exceed some system limits
2066 for UNIX sockets, which commonly are set to 107 characters.
2067
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002068 <param*> is a list of parameters common to all sockets declared on the
2069 same line. These numerous parameters depend on OS and build
2070 options and have a complete section dedicated to them. Please
2071 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002072
Willy Tarreau0ba27502007-12-24 16:55:16 +01002073 It is possible to specify a list of address:port combinations delimited by
2074 commas. The frontend will then listen on all of these addresses. There is no
2075 fixed limit to the number of addresses and ports which can be listened on in
2076 a frontend, as well as there is no limit to the number of "bind" statements
2077 in a frontend.
2078
2079 Example :
2080 listen http_proxy
2081 bind :80,:443
2082 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002083 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01002084
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002085 listen http_https_proxy
2086 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02002087 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002088
Willy Tarreau24709282013-03-10 21:32:12 +01002089 listen http_https_proxy_explicit
2090 bind ipv6@:80
2091 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
2092 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
2093
Willy Tarreaudad36a32013-03-11 01:20:04 +01002094 listen external_bind_app1
William Lallemandb2f07452015-05-12 14:27:13 +02002095 bind "fd@${FD_APP1}"
Willy Tarreaudad36a32013-03-11 01:20:04 +01002096
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002097 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002098 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002099
2100
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002101bind-process [ all | odd | even | <number 1-64>[-<number 1-64>] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002102 Limit visibility of an instance to a certain set of processes numbers.
2103 May be used in sections : defaults | frontend | listen | backend
2104 yes | yes | yes | yes
2105 Arguments :
2106 all All process will see this instance. This is the default. It
2107 may be used to override a default value.
2108
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002109 odd This instance will be enabled on processes 1,3,5,...63. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002110 option may be combined with other numbers.
2111
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002112 even This instance will be enabled on processes 2,4,6,...64. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002113 option may be combined with other numbers. Do not use it
2114 with less than 2 processes otherwise some instances might be
2115 missing from all processes.
2116
Willy Tarreau110ecc12012-11-15 17:50:01 +01002117 number The instance will be enabled on this process number or range,
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002118 whose values must all be between 1 and 32 or 64 depending on
Willy Tarreau102df612014-05-07 23:56:38 +02002119 the machine's word size. If a proxy is bound to process
2120 numbers greater than the configured global.nbproc, it will
2121 either be forced to process #1 if a single process was
2122 specified, or to all processes otherwise.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002123
2124 This keyword limits binding of certain instances to certain processes. This
2125 is useful in order not to have too many processes listening to the same
2126 ports. For instance, on a dual-core machine, it might make sense to set
2127 'nbproc 2' in the global section, then distributes the listeners among 'odd'
2128 and 'even' instances.
2129
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002130 At the moment, it is not possible to reference more than 32 or 64 processes
2131 using this keyword, but this should be more than enough for most setups.
2132 Please note that 'all' really means all processes regardless of the machine's
2133 word size, and is not limited to the first 32 or 64.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002134
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002135 Each "bind" line may further be limited to a subset of the proxy's processes,
2136 please consult the "process" bind keyword in section 5.1.
2137
Willy Tarreaub369a042014-09-16 13:21:03 +02002138 When a frontend has no explicit "bind-process" line, it tries to bind to all
2139 the processes referenced by its "bind" lines. That means that frontends can
2140 easily adapt to their listeners' processes.
2141
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002142 If some backends are referenced by frontends bound to other processes, the
2143 backend automatically inherits the frontend's processes.
2144
2145 Example :
2146 listen app_ip1
2147 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002148 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002149
2150 listen app_ip2
2151 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002152 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002153
2154 listen management
2155 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002156 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002157
Willy Tarreau110ecc12012-11-15 17:50:01 +01002158 listen management
2159 bind 10.0.0.4:80
2160 bind-process 1-4
2161
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002162 See also : "nbproc" in global section, and "process" in section 5.1.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002163
2164
Willy Tarreau0ba27502007-12-24 16:55:16 +01002165block { if | unless } <condition>
2166 Block a layer 7 request if/unless a condition is matched
2167 May be used in sections : defaults | frontend | listen | backend
2168 no | yes | yes | yes
2169
2170 The HTTP request will be blocked very early in the layer 7 processing
2171 if/unless <condition> is matched. A 403 error will be returned if the request
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002172 is blocked. The condition has to reference ACLs (see section 7). This is
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02002173 typically used to deny access to certain sensitive resources if some
Willy Tarreau0ba27502007-12-24 16:55:16 +01002174 conditions are met or not met. There is no fixed limit to the number of
2175 "block" statements per instance.
2176
2177 Example:
2178 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2179 acl invalid_src src_port 0:1023
2180 acl local_dst hdr(host) -i localhost
2181 block if invalid_src || local_dst
2182
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002183 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002184
2185
2186capture cookie <name> len <length>
2187 Capture and log a cookie in the request and in the response.
2188 May be used in sections : defaults | frontend | listen | backend
2189 no | yes | yes | no
2190 Arguments :
2191 <name> is the beginning of the name of the cookie to capture. In order
2192 to match the exact name, simply suffix the name with an equal
2193 sign ('='). The full name will appear in the logs, which is
2194 useful with application servers which adjust both the cookie name
2195 and value (eg: ASPSESSIONXXXXX).
2196
2197 <length> is the maximum number of characters to report in the logs, which
2198 include the cookie name, the equal sign and the value, all in the
2199 standard "name=value" form. The string will be truncated on the
2200 right if it exceeds <length>.
2201
2202 Only the first cookie is captured. Both the "cookie" request headers and the
2203 "set-cookie" response headers are monitored. This is particularly useful to
2204 check for application bugs causing session crossing or stealing between
2205 users, because generally the user's cookies can only change on a login page.
2206
2207 When the cookie was not presented by the client, the associated log column
2208 will report "-". When a request does not cause a cookie to be assigned by the
2209 server, a "-" is reported in the response column.
2210
2211 The capture is performed in the frontend only because it is necessary that
2212 the log format does not change for a given frontend depending on the
2213 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01002214 "capture cookie" statement in a frontend. The maximum capture length is set
2215 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
2216 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002217
2218 Example:
2219 capture cookie ASPSESSION len 32
2220
2221 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002222 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002223
2224
2225capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01002226 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002227 May be used in sections : defaults | frontend | listen | backend
2228 no | yes | yes | no
2229 Arguments :
2230 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002231 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01002232 appear in the requests, with the first letter of each word in
2233 upper case. The header name will not appear in the logs, only the
2234 value is reported, but the position in the logs is respected.
2235
2236 <length> is the maximum number of characters to extract from the value and
2237 report in the logs. The string will be truncated on the right if
2238 it exceeds <length>.
2239
Willy Tarreau4460d032012-11-21 23:37:37 +01002240 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01002241 value will be added to the logs between braces ('{}'). If multiple headers
2242 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002243 in the same order they were declared in the configuration. Non-existent
2244 headers will be logged just as an empty string. Common uses for request
2245 header captures include the "Host" field in virtual hosting environments, the
2246 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002247 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002248 environments to find where the request came from.
2249
2250 Note that when capturing headers such as "User-agent", some spaces may be
2251 logged, making the log analysis more difficult. Thus be careful about what
2252 you log if you know your log parser is not smart enough to rely on the
2253 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002254
Willy Tarreau0900abb2012-11-22 00:21:46 +01002255 There is no limit to the number of captured request headers nor to their
2256 length, though it is wise to keep them low to limit memory usage per session.
2257 In order to keep log format consistent for a same frontend, header captures
2258 can only be declared in a frontend. It is not possible to specify a capture
2259 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002260
2261 Example:
2262 capture request header Host len 15
2263 capture request header X-Forwarded-For len 15
2264 capture request header Referrer len 15
2265
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002266 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01002267 about logging.
2268
2269
2270capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01002271 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002272 May be used in sections : defaults | frontend | listen | backend
2273 no | yes | yes | no
2274 Arguments :
2275 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002276 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01002277 appear in the response, with the first letter of each word in
2278 upper case. The header name will not appear in the logs, only the
2279 value is reported, but the position in the logs is respected.
2280
2281 <length> is the maximum number of characters to extract from the value and
2282 report in the logs. The string will be truncated on the right if
2283 it exceeds <length>.
2284
Willy Tarreau4460d032012-11-21 23:37:37 +01002285 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01002286 result will be added to the logs between braces ('{}') after the captured
2287 request headers. If multiple headers are captured, they will be delimited by
2288 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002289 the configuration. Non-existent headers will be logged just as an empty
2290 string. Common uses for response header captures include the "Content-length"
2291 header which indicates how many bytes are expected to be returned, the
2292 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002293
Willy Tarreau0900abb2012-11-22 00:21:46 +01002294 There is no limit to the number of captured response headers nor to their
2295 length, though it is wise to keep them low to limit memory usage per session.
2296 In order to keep log format consistent for a same frontend, header captures
2297 can only be declared in a frontend. It is not possible to specify a capture
2298 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002299
2300 Example:
2301 capture response header Content-length len 9
2302 capture response header Location len 15
2303
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002304 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01002305 about logging.
2306
2307
Cyril Bontéf0c60612010-02-06 14:44:47 +01002308clitimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002309 Set the maximum inactivity time on the client side.
2310 May be used in sections : defaults | frontend | listen | backend
2311 yes | yes | yes | no
2312 Arguments :
2313 <timeout> is the timeout value is specified in milliseconds by default, but
2314 can be in any other unit if the number is suffixed by the unit,
2315 as explained at the top of this document.
2316
2317 The inactivity timeout applies when the client is expected to acknowledge or
2318 send data. In HTTP mode, this timeout is particularly important to consider
2319 during the first phase, when the client sends the request, and during the
2320 response while it is reading data sent by the server. The value is specified
2321 in milliseconds by default, but can be in any other unit if the number is
2322 suffixed by the unit, as specified at the top of this document. In TCP mode
2323 (and to a lesser extent, in HTTP mode), it is highly recommended that the
2324 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002325 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01002326 losses by specifying timeouts that are slightly above multiples of 3 seconds
2327 (eg: 4 or 5 seconds).
2328
2329 This parameter is specific to frontends, but can be specified once for all in
2330 "defaults" sections. This is in fact one of the easiest solutions not to
2331 forget about it. An unspecified timeout results in an infinite timeout, which
2332 is not recommended. Such a usage is accepted and works but reports a warning
2333 during startup because it may results in accumulation of expired sessions in
2334 the system if the system's timeouts are not configured either.
2335
2336 This parameter is provided for compatibility but is currently deprecated.
2337 Please use "timeout client" instead.
2338
Willy Tarreau036fae02008-01-06 13:24:40 +01002339 See also : "timeout client", "timeout http-request", "timeout server", and
2340 "srvtimeout".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002341
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002342compression algo <algorithm> ...
2343compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02002344compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02002345 Enable HTTP compression.
2346 May be used in sections : defaults | frontend | listen | backend
2347 yes | yes | yes | yes
2348 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002349 algo is followed by the list of supported compression algorithms.
2350 type is followed by the list of MIME types that will be compressed.
2351 offload makes haproxy work as a compression offloader only (see notes).
2352
2353 The currently supported algorithms are :
Willy Tarreauc91840a2015-03-28 17:00:39 +01002354 identity this is mostly for debugging, and it was useful for developing
2355 the compression feature. Identity does not apply any change on
2356 data.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002357
Willy Tarreauc91840a2015-03-28 17:00:39 +01002358 gzip applies gzip compression. This setting is only available when
2359 support for zlib was built in.
2360
2361 deflate same as "gzip", but with deflate algorithm and zlib format.
2362 Note that this algorithm has ambiguous support on many
2363 browsers and no support at all from recent ones. It is
2364 strongly recommended not to use it for anything else than
2365 experimentation. This setting is only available when support
2366 for zlib was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002367
Willy Tarreauc91840a2015-03-28 17:00:39 +01002368 raw-deflate same as "deflate" without the zlib wrapper, and used as an
2369 alternative when the browser wants "deflate". All major
2370 browsers understand it and despite violating the standards,
2371 it is known to work better than "deflate", at least on MSIE
2372 and some versions of Safari. Do not use it in conjunction
2373 with "deflate", use either one or the other since both react
2374 to the same Accept-Encoding token. This setting is only
2375 available when support for zlib was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002376
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04002377 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002378 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04002379 If backend servers support HTTP compression, these directives
2380 will be no-op: haproxy will see the compressed response and will not
2381 compress again. If backend servers do not support HTTP compression and
2382 there is Accept-Encoding header in request, haproxy will compress the
2383 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02002384
2385 The "offload" setting makes haproxy remove the Accept-Encoding header to
2386 prevent backend servers from compressing responses. It is strongly
2387 recommended not to do this because this means that all the compression work
2388 will be done on the single point where haproxy is located. However in some
2389 deployment scenarios, haproxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04002390 with broken HTTP compression implementation which can't be turned off.
2391 In that case haproxy can be used to prevent that gateway from emitting
2392 invalid payloads. In this case, simply removing the header in the
2393 configuration does not work because it applies before the header is parsed,
2394 so that prevents haproxy from compressing. The "offload" setting should
Willy Tarreauffea9fd2014-07-12 16:37:02 +02002395 then be used for such scenarios. Note: for now, the "offload" setting is
2396 ignored when set in a defaults section.
William Lallemand82fe75c2012-10-23 10:25:10 +02002397
William Lallemand05097442012-11-20 12:14:28 +01002398 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002399 * the request does not advertise a supported compression algorithm in the
2400 "Accept-Encoding" header
2401 * the response message is not HTTP/1.1
William Lallemandd3002612012-11-26 14:34:47 +01002402 * HTTP status code is not 200
William Lallemand8bb4e342013-12-10 17:28:48 +01002403 * response header "Transfer-Encoding" contains "chunked" (Temporary
2404 Workaround)
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002405 * response contain neither a "Content-Length" header nor a
2406 "Transfer-Encoding" whose last value is "chunked"
2407 * response contains a "Content-Type" header whose first value starts with
2408 "multipart"
2409 * the response contains the "no-transform" value in the "Cache-control"
2410 header
2411 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
2412 and later
2413 * The response contains a "Content-Encoding" header, indicating that the
2414 response is already compressed (see compression offload)
William Lallemand05097442012-11-20 12:14:28 +01002415
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002416 Note: The compression does not rewrite Etag headers, and does not emit the
2417 Warning header.
William Lallemand05097442012-11-20 12:14:28 +01002418
William Lallemand82fe75c2012-10-23 10:25:10 +02002419 Examples :
2420 compression algo gzip
2421 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01002422
Cyril Bontéf0c60612010-02-06 14:44:47 +01002423contimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002424 Set the maximum time to wait for a connection attempt to a server to succeed.
2425 May be used in sections : defaults | frontend | listen | backend
2426 yes | no | yes | yes
2427 Arguments :
2428 <timeout> is the timeout value is specified in milliseconds by default, but
2429 can be in any other unit if the number is suffixed by the unit,
2430 as explained at the top of this document.
2431
2432 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002433 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01002434 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01002435 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
2436 connect timeout also presets the queue timeout to the same value if this one
2437 has not been specified. Historically, the contimeout was also used to set the
2438 tarpit timeout in a listen section, which is not possible in a pure frontend.
2439
2440 This parameter is specific to backends, but can be specified once for all in
2441 "defaults" sections. This is in fact one of the easiest solutions not to
2442 forget about it. An unspecified timeout results in an infinite timeout, which
2443 is not recommended. Such a usage is accepted and works but reports a warning
2444 during startup because it may results in accumulation of failed sessions in
2445 the system if the system's timeouts are not configured either.
2446
2447 This parameter is provided for backwards compatibility but is currently
2448 deprecated. Please use "timeout connect", "timeout queue" or "timeout tarpit"
2449 instead.
2450
2451 See also : "timeout connect", "timeout queue", "timeout tarpit",
2452 "timeout server", "contimeout".
2453
2454
Willy Tarreau55165fe2009-05-10 12:02:55 +02002455cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02002456 [ postonly ] [ preserve ] [ httponly ] [ secure ]
2457 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002458 Enable cookie-based persistence in a backend.
2459 May be used in sections : defaults | frontend | listen | backend
2460 yes | no | yes | yes
2461 Arguments :
2462 <name> is the name of the cookie which will be monitored, modified or
2463 inserted in order to bring persistence. This cookie is sent to
2464 the client via a "Set-Cookie" header in the response, and is
2465 brought back by the client in a "Cookie" header in all requests.
2466 Special care should be taken to choose a name which does not
2467 conflict with any likely application cookie. Also, if the same
2468 backends are subject to be used by the same clients (eg:
2469 HTTP/HTTPS), care should be taken to use different cookie names
2470 between all backends if persistence between them is not desired.
2471
2472 rewrite This keyword indicates that the cookie will be provided by the
2473 server and that haproxy will have to modify its value to set the
2474 server's identifier in it. This mode is handy when the management
2475 of complex combinations of "Set-cookie" and "Cache-control"
2476 headers is left to the application. The application can then
2477 decide whether or not it is appropriate to emit a persistence
2478 cookie. Since all responses should be monitored, this mode only
2479 works in HTTP close mode. Unless the application behaviour is
2480 very complex and/or broken, it is advised not to start with this
2481 mode for new deployments. This keyword is incompatible with
2482 "insert" and "prefix".
2483
2484 insert This keyword indicates that the persistence cookie will have to
Willy Tarreaua79094d2010-08-31 22:54:15 +02002485 be inserted by haproxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002486
Willy Tarreaua79094d2010-08-31 22:54:15 +02002487 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002488 server. When used without the "preserve" option, if the server
2489 emits a cookie with the same name, it will be remove before
2490 processing. For this reason, this mode can be used to upgrade
2491 existing configurations running in the "rewrite" mode. The cookie
2492 will only be a session cookie and will not be stored on the
2493 client's disk. By default, unless the "indirect" option is added,
2494 the server will see the cookies emitted by the client. Due to
2495 caching effects, it is generally wise to add the "nocache" or
2496 "postonly" keywords (see below). The "insert" keyword is not
2497 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002498
2499 prefix This keyword indicates that instead of relying on a dedicated
2500 cookie for the persistence, an existing one will be completed.
2501 This may be needed in some specific environments where the client
2502 does not support more than one single cookie and the application
2503 already needs it. In this case, whenever the server sets a cookie
2504 named <name>, it will be prefixed with the server's identifier
2505 and a delimiter. The prefix will be removed from all client
2506 requests so that the server still finds the cookie it emitted.
2507 Since all requests and responses are subject to being modified,
2508 this mode requires the HTTP close mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02002509 not compatible with "rewrite" and "insert". Note: it is highly
2510 recommended not to use "indirect" with "prefix", otherwise server
2511 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002512
Willy Tarreaua79094d2010-08-31 22:54:15 +02002513 indirect When this option is specified, no cookie will be emitted to a
2514 client which already has a valid one for the server which has
2515 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002516 it will be removed, unless the "preserve" option is also set. In
2517 "insert" mode, this will additionally remove cookies from the
2518 requests transmitted to the server, making the persistence
2519 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02002520 Note: it is highly recommended not to use "indirect" with
2521 "prefix", otherwise server cookie updates would not be sent to
2522 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002523
2524 nocache This option is recommended in conjunction with the insert mode
2525 when there is a cache between the client and HAProxy, as it
2526 ensures that a cacheable response will be tagged non-cacheable if
2527 a cookie needs to be inserted. This is important because if all
2528 persistence cookies are added on a cacheable home page for
2529 instance, then all customers will then fetch the page from an
2530 outer cache and will all share the same persistence cookie,
2531 leading to one server receiving much more traffic than others.
2532 See also the "insert" and "postonly" options.
2533
2534 postonly This option ensures that cookie insertion will only be performed
2535 on responses to POST requests. It is an alternative to the
2536 "nocache" option, because POST responses are not cacheable, so
2537 this ensures that the persistence cookie will never get cached.
2538 Since most sites do not need any sort of persistence before the
2539 first POST which generally is a login request, this is a very
2540 efficient method to optimize caching without risking to find a
2541 persistence cookie in the cache.
2542 See also the "insert" and "nocache" options.
2543
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002544 preserve This option may only be used with "insert" and/or "indirect". It
2545 allows the server to emit the persistence cookie itself. In this
2546 case, if a cookie is found in the response, haproxy will leave it
2547 untouched. This is useful in order to end persistence after a
2548 logout request for instance. For this, the server just has to
2549 emit a cookie with an invalid value (eg: empty) or with a date in
2550 the past. By combining this mechanism with the "disable-on-404"
2551 check option, it is possible to perform a completely graceful
2552 shutdown because users will definitely leave the server after
2553 they logout.
2554
Willy Tarreau4992dd22012-05-31 21:02:17 +02002555 httponly This option tells haproxy to add an "HttpOnly" cookie attribute
2556 when a cookie is inserted. This attribute is used so that a
2557 user agent doesn't share the cookie with non-HTTP components.
2558 Please check RFC6265 for more information on this attribute.
2559
2560 secure This option tells haproxy to add a "Secure" cookie attribute when
2561 a cookie is inserted. This attribute is used so that a user agent
2562 never emits this cookie over non-secure channels, which means
2563 that a cookie learned with this flag will be presented only over
2564 SSL/TLS connections. Please check RFC6265 for more information on
2565 this attribute.
2566
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002567 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002568 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01002569 name. If the domain begins with a dot, the browser is allowed to
2570 use it for any host ending with that name. It is also possible to
2571 specify several domain names by invoking this option multiple
2572 times. Some browsers might have small limits on the number of
2573 domains, so be careful when doing that. For the record, sending
2574 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002575
Willy Tarreau996a92c2010-10-13 19:30:47 +02002576 maxidle This option allows inserted cookies to be ignored after some idle
2577 time. It only works with insert-mode cookies. When a cookie is
2578 sent to the client, the date this cookie was emitted is sent too.
2579 Upon further presentations of this cookie, if the date is older
2580 than the delay indicated by the parameter (in seconds), it will
2581 be ignored. Otherwise, it will be refreshed if needed when the
2582 response is sent to the client. This is particularly useful to
2583 prevent users who never close their browsers from remaining for
2584 too long on the same server (eg: after a farm size change). When
2585 this option is set and a cookie has no date, it is always
2586 accepted, but gets refreshed in the response. This maintains the
2587 ability for admins to access their sites. Cookies that have a
2588 date in the future further than 24 hours are ignored. Doing so
2589 lets admins fix timezone issues without risking kicking users off
2590 the site.
2591
2592 maxlife This option allows inserted cookies to be ignored after some life
2593 time, whether they're in use or not. It only works with insert
2594 mode cookies. When a cookie is first sent to the client, the date
2595 this cookie was emitted is sent too. Upon further presentations
2596 of this cookie, if the date is older than the delay indicated by
2597 the parameter (in seconds), it will be ignored. If the cookie in
2598 the request has no date, it is accepted and a date will be set.
2599 Cookies that have a date in the future further than 24 hours are
2600 ignored. Doing so lets admins fix timezone issues without risking
2601 kicking users off the site. Contrary to maxidle, this value is
2602 not refreshed, only the first visit date counts. Both maxidle and
2603 maxlife may be used at the time. This is particularly useful to
2604 prevent users who never close their browsers from remaining for
2605 too long on the same server (eg: after a farm size change). This
2606 is stronger than the maxidle method in that it forces a
2607 redispatch after some absolute delay.
2608
Willy Tarreau0ba27502007-12-24 16:55:16 +01002609 There can be only one persistence cookie per HTTP backend, and it can be
2610 declared in a defaults section. The value of the cookie will be the value
2611 indicated after the "cookie" keyword in a "server" statement. If no cookie
2612 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002613
Willy Tarreau0ba27502007-12-24 16:55:16 +01002614 Examples :
2615 cookie JSESSIONID prefix
2616 cookie SRV insert indirect nocache
2617 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02002618 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01002619
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02002620 See also : "appsession", "balance source", "capture cookie", "server"
Cyril Bonté0d4bf012010-04-25 23:21:46 +02002621 and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002622
Willy Tarreau983e01e2010-01-11 18:42:06 +01002623
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002624default-server [param*]
2625 Change default options for a server in a backend
2626 May be used in sections : defaults | frontend | listen | backend
2627 yes | no | yes | yes
2628 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01002629 <param*> is a list of parameters for this server. The "default-server"
2630 keyword accepts an important number of options and has a complete
2631 section dedicated to it. Please refer to section 5 for more
2632 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002633
Willy Tarreau983e01e2010-01-11 18:42:06 +01002634 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002635 default-server inter 1000 weight 13
2636
2637 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01002638
Willy Tarreau983e01e2010-01-11 18:42:06 +01002639
Willy Tarreau0ba27502007-12-24 16:55:16 +01002640default_backend <backend>
2641 Specify the backend to use when no "use_backend" rule has been matched.
2642 May be used in sections : defaults | frontend | listen | backend
2643 yes | yes | yes | no
2644 Arguments :
2645 <backend> is the name of the backend to use.
2646
2647 When doing content-switching between frontend and backends using the
2648 "use_backend" keyword, it is often useful to indicate which backend will be
2649 used when no rule has matched. It generally is the dynamic backend which
2650 will catch all undetermined requests.
2651
Willy Tarreau0ba27502007-12-24 16:55:16 +01002652 Example :
2653
2654 use_backend dynamic if url_dyn
2655 use_backend static if url_css url_img extension_img
2656 default_backend dynamic
2657
Willy Tarreau2769aa02007-12-27 18:26:09 +01002658 See also : "use_backend", "reqsetbe", "reqisetbe"
2659
Willy Tarreau0ba27502007-12-24 16:55:16 +01002660
Baptiste Assmann27f51342013-10-09 06:51:49 +02002661description <string>
2662 Describe a listen, frontend or backend.
2663 May be used in sections : defaults | frontend | listen | backend
2664 no | yes | yes | yes
2665 Arguments : string
2666
2667 Allows to add a sentence to describe the related object in the HAProxy HTML
2668 stats page. The description will be printed on the right of the object name
2669 it describes.
2670 No need to backslash spaces in the <string> arguments.
2671
2672
Willy Tarreau0ba27502007-12-24 16:55:16 +01002673disabled
2674 Disable a proxy, frontend or backend.
2675 May be used in sections : defaults | frontend | listen | backend
2676 yes | yes | yes | yes
2677 Arguments : none
2678
2679 The "disabled" keyword is used to disable an instance, mainly in order to
2680 liberate a listening port or to temporarily disable a service. The instance
2681 will still be created and its configuration will be checked, but it will be
2682 created in the "stopped" state and will appear as such in the statistics. It
2683 will not receive any traffic nor will it send any health-checks or logs. It
2684 is possible to disable many instances at once by adding the "disabled"
2685 keyword in a "defaults" section.
2686
2687 See also : "enabled"
2688
2689
Willy Tarreau5ce94572010-06-07 14:35:41 +02002690dispatch <address>:<port>
2691 Set a default server address
2692 May be used in sections : defaults | frontend | listen | backend
2693 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02002694 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02002695
2696 <address> is the IPv4 address of the default server. Alternatively, a
2697 resolvable hostname is supported, but this name will be resolved
2698 during start-up.
2699
2700 <ports> is a mandatory port specification. All connections will be sent
2701 to this port, and it is not permitted to use port offsets as is
2702 possible with normal servers.
2703
Willy Tarreau787aed52011-04-15 06:45:37 +02002704 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02002705 server can take the connection. In the past it was used to forward non
2706 persistent connections to an auxiliary load balancer. Due to its simple
2707 syntax, it has also been used for simple TCP relays. It is recommended not to
2708 use it for more clarity, and to use the "server" directive instead.
2709
2710 See also : "server"
2711
2712
Willy Tarreau0ba27502007-12-24 16:55:16 +01002713enabled
2714 Enable a proxy, frontend or backend.
2715 May be used in sections : defaults | frontend | listen | backend
2716 yes | yes | yes | yes
2717 Arguments : none
2718
2719 The "enabled" keyword is used to explicitly enable an instance, when the
2720 defaults has been set to "disabled". This is very rarely used.
2721
2722 See also : "disabled"
2723
2724
2725errorfile <code> <file>
2726 Return a file contents instead of errors generated by HAProxy
2727 May be used in sections : defaults | frontend | listen | backend
2728 yes | yes | yes | yes
2729 Arguments :
2730 <code> is the HTTP status code. Currently, HAProxy is capable of
CJ Ess108b1dd2015-04-07 12:03:37 -04002731 generating codes 200, 400, 403, 405, 408, 429, 500, 502, 503, and
2732 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002733
2734 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002735 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01002736 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01002737 error pages, and to use absolute paths, since files are read
2738 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002739
2740 It is important to understand that this keyword is not meant to rewrite
2741 errors returned by the server, but errors detected and returned by HAProxy.
2742 This is why the list of supported errors is limited to a small set.
2743
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002744 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2745
Willy Tarreau0ba27502007-12-24 16:55:16 +01002746 The files are returned verbatim on the TCP socket. This allows any trick such
2747 as redirections to another URL or site, as well as tricks to clean cookies,
2748 force enable or disable caching, etc... The package provides default error
2749 files returning the same contents as default errors.
2750
Willy Tarreau59140a22009-02-22 12:02:30 +01002751 The files should not exceed the configured buffer size (BUFSIZE), which
2752 generally is 8 or 16 kB, otherwise they will be truncated. It is also wise
2753 not to put any reference to local contents (eg: images) in order to avoid
2754 loops between the client and HAProxy when all servers are down, causing an
2755 error to be returned instead of an image. For better HTTP compliance, it is
2756 recommended that all header lines end with CR-LF and not LF alone.
2757
Willy Tarreau0ba27502007-12-24 16:55:16 +01002758 The files are read at the same time as the configuration and kept in memory.
2759 For this reason, the errors continue to be returned even when the process is
2760 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01002761 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01002762 403 status code and interrogating a blocked URL.
2763
2764 See also : "errorloc", "errorloc302", "errorloc303"
2765
Willy Tarreau59140a22009-02-22 12:02:30 +01002766 Example :
2767 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
Willy Tarreau2705a612014-05-23 17:38:34 +02002768 errorfile 408 /dev/null # workaround Chrome pre-connect bug
Willy Tarreau59140a22009-02-22 12:02:30 +01002769 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
2770 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
2771
Willy Tarreau2769aa02007-12-27 18:26:09 +01002772
2773errorloc <code> <url>
2774errorloc302 <code> <url>
2775 Return an HTTP redirection to a URL instead of errors generated by HAProxy
2776 May be used in sections : defaults | frontend | listen | backend
2777 yes | yes | yes | yes
2778 Arguments :
2779 <code> is the HTTP status code. Currently, HAProxy is capable of
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002780 generating codes 200, 400, 403, 408, 500, 502, 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002781
2782 <url> it is the exact contents of the "Location" header. It may contain
2783 either a relative URI to an error page hosted on the same site,
2784 or an absolute URI designating an error page on another site.
2785 Special care should be given to relative URIs to avoid redirect
2786 loops if the URI itself may generate the same error (eg: 500).
2787
2788 It is important to understand that this keyword is not meant to rewrite
2789 errors returned by the server, but errors detected and returned by HAProxy.
2790 This is why the list of supported errors is limited to a small set.
2791
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002792 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2793
Willy Tarreau2769aa02007-12-27 18:26:09 +01002794 Note that both keyword return the HTTP 302 status code, which tells the
2795 client to fetch the designated URL using the same HTTP method. This can be
2796 quite problematic in case of non-GET methods such as POST, because the URL
2797 sent to the client might not be allowed for something other than GET. To
2798 workaround this problem, please use "errorloc303" which send the HTTP 303
2799 status code, indicating to the client that the URL must be fetched with a GET
2800 request.
2801
2802 See also : "errorfile", "errorloc303"
2803
2804
2805errorloc303 <code> <url>
2806 Return an HTTP redirection to a URL instead of errors generated by HAProxy
2807 May be used in sections : defaults | frontend | listen | backend
2808 yes | yes | yes | yes
2809 Arguments :
2810 <code> is the HTTP status code. Currently, HAProxy is capable of
2811 generating codes 400, 403, 408, 500, 502, 503, and 504.
2812
2813 <url> it is the exact contents of the "Location" header. It may contain
2814 either a relative URI to an error page hosted on the same site,
2815 or an absolute URI designating an error page on another site.
2816 Special care should be given to relative URIs to avoid redirect
2817 loops if the URI itself may generate the same error (eg: 500).
2818
2819 It is important to understand that this keyword is not meant to rewrite
2820 errors returned by the server, but errors detected and returned by HAProxy.
2821 This is why the list of supported errors is limited to a small set.
2822
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002823 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2824
Willy Tarreau2769aa02007-12-27 18:26:09 +01002825 Note that both keyword return the HTTP 303 status code, which tells the
2826 client to fetch the designated URL using the same HTTP GET method. This
2827 solves the usual problems associated with "errorloc" and the 302 code. It is
2828 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002829 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002830
2831 See also : "errorfile", "errorloc", "errorloc302"
2832
2833
Simon Horman51a1cf62015-02-03 13:00:44 +09002834email-alert from <emailaddr>
2835 Declare the from email address to be used in both the envelope and header
2836 of email alerts. This is the address that email alerts are sent from.
2837 May be used in sections: defaults | frontend | listen | backend
2838 yes | yes | yes | yes
2839
2840 Arguments :
2841
2842 <emailaddr> is the from email address to use when sending email alerts
2843
2844 Also requires "email-alert mailers" and "email-alert to" to be set
2845 and if so sending email alerts is enabled for the proxy.
2846
Simon Horman64e34162015-02-06 11:11:57 +09002847 See also : "email-alert level", "email-alert mailers",
2848 "email-alert myhostname", "email-alert to", section 3.6 about mailers.
2849
2850
2851email-alert level <level>
2852 Declare the maximum log level of messages for which email alerts will be
2853 sent. This acts as a filter on the sending of email alerts.
2854 May be used in sections: defaults | frontend | listen | backend
2855 yes | yes | yes | yes
2856
2857 Arguments :
2858
2859 <level> One of the 8 syslog levels:
2860 emerg alert crit err warning notice info debug
2861 The above syslog levels are ordered from lowest to highest.
2862
2863 By default level is alert
2864
2865 Also requires "email-alert from", "email-alert mailers" and
2866 "email-alert to" to be set and if so sending email alerts is enabled
2867 for the proxy.
2868
Simon Horman1421e212015-04-30 13:10:35 +09002869 Alerts are sent when :
2870
2871 * An un-paused server is marked as down and <level> is alert or lower
2872 * A paused server is marked as down and <level> is notice or lower
2873 * A server is marked as up or enters the drain state and <level>
2874 is notice or lower
2875 * "option log-health-checks" is enabled, <level> is info or lower,
2876 and a health check status update occurs
2877
Simon Horman64e34162015-02-06 11:11:57 +09002878 See also : "email-alert from", "email-alert mailers",
2879 "email-alert myhostname", "email-alert to",
Simon Horman51a1cf62015-02-03 13:00:44 +09002880 section 3.6 about mailers.
2881
2882
2883email-alert mailers <mailersect>
2884 Declare the mailers to be used when sending email alerts
2885 May be used in sections: defaults | frontend | listen | backend
2886 yes | yes | yes | yes
2887
2888 Arguments :
2889
2890 <mailersect> is the name of the mailers section to send email alerts.
2891
2892 Also requires "email-alert from" and "email-alert to" to be set
2893 and if so sending email alerts is enabled for the proxy.
2894
Simon Horman64e34162015-02-06 11:11:57 +09002895 See also : "email-alert from", "email-alert level", "email-alert myhostname",
2896 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09002897
2898
2899email-alert myhostname <hostname>
2900 Declare the to hostname address to be used when communicating with
2901 mailers.
2902 May be used in sections: defaults | frontend | listen | backend
2903 yes | yes | yes | yes
2904
2905 Arguments :
2906
2907 <emailaddr> is the to email address to use when sending email alerts
2908
2909 By default the systems hostname is used.
2910
2911 Also requires "email-alert from", "email-alert mailers" and
2912 "email-alert to" to be set and if so sending email alerts is enabled
2913 for the proxy.
2914
Simon Horman64e34162015-02-06 11:11:57 +09002915 See also : "email-alert from", "email-alert level", "email-alert mailers",
2916 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09002917
2918
2919email-alert to <emailaddr>
2920 Declare both the recipent address in the envelope and to address in the
2921 header of email alerts. This is the address that email alerts are sent to.
2922 May be used in sections: defaults | frontend | listen | backend
2923 yes | yes | yes | yes
2924
2925 Arguments :
2926
2927 <emailaddr> is the to email address to use when sending email alerts
2928
2929 Also requires "email-alert mailers" and "email-alert to" to be set
2930 and if so sending email alerts is enabled for the proxy.
2931
Simon Horman64e34162015-02-06 11:11:57 +09002932 See also : "email-alert from", "email-alert level", "email-alert mailers",
Simon Horman51a1cf62015-02-03 13:00:44 +09002933 "email-alert myhostname", section 3.6 about mailers.
2934
2935
Willy Tarreau4de91492010-01-22 19:10:05 +01002936force-persist { if | unless } <condition>
2937 Declare a condition to force persistence on down servers
2938 May be used in sections: defaults | frontend | listen | backend
2939 no | yes | yes | yes
2940
2941 By default, requests are not dispatched to down servers. It is possible to
2942 force this using "option persist", but it is unconditional and redispatches
2943 to a valid server if "option redispatch" is set. That leaves with very little
2944 possibilities to force some requests to reach a server which is artificially
2945 marked down for maintenance operations.
2946
2947 The "force-persist" statement allows one to declare various ACL-based
2948 conditions which, when met, will cause a request to ignore the down status of
2949 a server and still try to connect to it. That makes it possible to start a
2950 server, still replying an error to the health checks, and run a specially
2951 configured browser to test the service. Among the handy methods, one could
2952 use a specific source IP address, or a specific cookie. The cookie also has
2953 the advantage that it can easily be added/removed on the browser from a test
2954 page. Once the service is validated, it is then possible to open the service
2955 to the world by returning a valid response to health checks.
2956
2957 The forced persistence is enabled when an "if" condition is met, or unless an
2958 "unless" condition is met. The final redispatch is always disabled when this
2959 is used.
2960
Cyril Bonté0d4bf012010-04-25 23:21:46 +02002961 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02002962 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01002963
2964
Willy Tarreau2769aa02007-12-27 18:26:09 +01002965fullconn <conns>
2966 Specify at what backend load the servers will reach their maxconn
2967 May be used in sections : defaults | frontend | listen | backend
2968 yes | no | yes | yes
2969 Arguments :
2970 <conns> is the number of connections on the backend which will make the
2971 servers use the maximal number of connections.
2972
Willy Tarreau198a7442008-01-17 12:05:32 +01002973 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01002974 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01002975 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01002976 load. The server will then always accept at least <minconn> connections,
2977 never more than <maxconn>, and the limit will be on the ramp between both
2978 values when the backend has less than <conns> concurrent connections. This
2979 makes it possible to limit the load on the servers during normal loads, but
2980 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002981 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002982
Willy Tarreaufbb78422011-06-05 15:38:35 +02002983 Since it's hard to get this value right, haproxy automatically sets it to
2984 10% of the sum of the maxconns of all frontends that may branch to this
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01002985 backend (based on "use_backend" and "default_backend" rules). That way it's
2986 safe to leave it unset. However, "use_backend" involving dynamic names are
2987 not counted since there is no way to know if they could match or not.
Willy Tarreaufbb78422011-06-05 15:38:35 +02002988
Willy Tarreau2769aa02007-12-27 18:26:09 +01002989 Example :
2990 # The servers will accept between 100 and 1000 concurrent connections each
2991 # and the maximum of 1000 will be reached when the backend reaches 10000
2992 # connections.
2993 backend dynamic
2994 fullconn 10000
2995 server srv1 dyn1:80 minconn 100 maxconn 1000
2996 server srv2 dyn2:80 minconn 100 maxconn 1000
2997
2998 See also : "maxconn", "server"
2999
3000
3001grace <time>
3002 Maintain a proxy operational for some time after a soft stop
3003 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01003004 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01003005 Arguments :
3006 <time> is the time (by default in milliseconds) for which the instance
3007 will remain operational with the frontend sockets still listening
3008 when a soft-stop is received via the SIGUSR1 signal.
3009
3010 This may be used to ensure that the services disappear in a certain order.
3011 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003012 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01003013 needed by the equipment to detect the failure.
3014
3015 Note that currently, there is very little benefit in using this parameter,
3016 and it may in fact complicate the soft-reconfiguration process more than
3017 simplify it.
3018
Willy Tarreau0ba27502007-12-24 16:55:16 +01003019
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003020hash-type <method> <function> <modifier>
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003021 Specify a method to use for mapping hashes to servers
3022 May be used in sections : defaults | frontend | listen | backend
3023 yes | no | yes | yes
3024 Arguments :
Bhaskar98634f02013-10-29 23:30:51 -04003025 <method> is the method used to select a server from the hash computed by
3026 the <function> :
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003027
Bhaskar98634f02013-10-29 23:30:51 -04003028 map-based the hash table is a static array containing all alive servers.
3029 The hashes will be very smooth, will consider weights, but
3030 will be static in that weight changes while a server is up
3031 will be ignored. This means that there will be no slow start.
3032 Also, since a server is selected by its position in the array,
3033 most mappings are changed when the server count changes. This
3034 means that when a server goes up or down, or when a server is
3035 added to a farm, most connections will be redistributed to
3036 different servers. This can be inconvenient with caches for
3037 instance.
Willy Tarreau798a39c2010-11-24 15:04:29 +01003038
Bhaskar98634f02013-10-29 23:30:51 -04003039 consistent the hash table is a tree filled with many occurrences of each
3040 server. The hash key is looked up in the tree and the closest
3041 server is chosen. This hash is dynamic, it supports changing
3042 weights while the servers are up, so it is compatible with the
3043 slow start feature. It has the advantage that when a server
3044 goes up or down, only its associations are moved. When a
3045 server is added to the farm, only a few part of the mappings
3046 are redistributed, making it an ideal method for caches.
3047 However, due to its principle, the distribution will never be
3048 very smooth and it may sometimes be necessary to adjust a
3049 server's weight or its ID to get a more balanced distribution.
3050 In order to get the same distribution on multiple load
3051 balancers, it is important that all servers have the exact
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003052 same IDs. Note: consistent hash uses sdbm and avalanche if no
3053 hash function is specified.
Bhaskar98634f02013-10-29 23:30:51 -04003054
3055 <function> is the hash function to be used :
3056
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003057 sdbm this function was created initially for sdbm (a public-domain
Bhaskar98634f02013-10-29 23:30:51 -04003058 reimplementation of ndbm) database library. It was found to do
3059 well in scrambling bits, causing better distribution of the keys
3060 and fewer splits. It also happens to be a good general hashing
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003061 function with good distribution, unless the total server weight
3062 is a multiple of 64, in which case applying the avalanche
3063 modifier may help.
Bhaskar98634f02013-10-29 23:30:51 -04003064
3065 djb2 this function was first proposed by Dan Bernstein many years ago
3066 on comp.lang.c. Studies have shown that for certain workload this
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003067 function provides a better distribution than sdbm. It generally
3068 works well with text-based inputs though it can perform extremely
3069 poorly with numeric-only input or when the total server weight is
3070 a multiple of 33, unless the avalanche modifier is also used.
3071
Willy Tarreaua0f42712013-11-14 14:30:35 +01003072 wt6 this function was designed for haproxy while testing other
3073 functions in the past. It is not as smooth as the other ones, but
3074 is much less sensible to the input data set or to the number of
3075 servers. It can make sense as an alternative to sdbm+avalanche or
3076 djb2+avalanche for consistent hashing or when hashing on numeric
3077 data such as a source IP address or a visitor identifier in a URL
3078 parameter.
3079
Willy Tarreau324f07f2015-01-20 19:44:50 +01003080 crc32 this is the most common CRC32 implementation as used in Ethernet,
3081 gzip, PNG, etc. It is slower than the other ones but may provide
3082 a better distribution or less predictable results especially when
3083 used on strings.
3084
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003085 <modifier> indicates an optional method applied after hashing the key :
3086
3087 avalanche This directive indicates that the result from the hash
3088 function above should not be used in its raw form but that
3089 a 4-byte full avalanche hash must be applied first. The
3090 purpose of this step is to mix the resulting bits from the
3091 previous hash in order to avoid any undesired effect when
3092 the input contains some limited values or when the number of
3093 servers is a multiple of one of the hash's components (64
3094 for SDBM, 33 for DJB2). Enabling avalanche tends to make the
3095 result less predictable, but it's also not as smooth as when
3096 using the original function. Some testing might be needed
3097 with some workloads. This hash is one of the many proposed
3098 by Bob Jenkins.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003099
Bhaskar98634f02013-10-29 23:30:51 -04003100 The default hash type is "map-based" and is recommended for most usages. The
3101 default function is "sdbm", the selection of a function should be based on
3102 the range of the values being hashed.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003103
3104 See also : "balance", "server"
3105
3106
Willy Tarreau0ba27502007-12-24 16:55:16 +01003107http-check disable-on-404
3108 Enable a maintenance mode upon HTTP/404 response to health-checks
3109 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01003110 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01003111 Arguments : none
3112
3113 When this option is set, a server which returns an HTTP code 404 will be
3114 excluded from further load-balancing, but will still receive persistent
3115 connections. This provides a very convenient method for Web administrators
3116 to perform a graceful shutdown of their servers. It is also important to note
3117 that a server which is detected as failed while it was in this mode will not
3118 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
3119 will immediately be reinserted into the farm. The status on the stats page
3120 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01003121 option only works in conjunction with the "httpchk" option. If this option
3122 is used with "http-check expect", then it has precedence over it so that 404
3123 responses will still be considered as soft-stop.
3124
3125 See also : "option httpchk", "http-check expect"
3126
3127
3128http-check expect [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003129 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01003130 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02003131 yes | no | yes | yes
Willy Tarreaubd741542010-03-16 18:46:54 +01003132 Arguments :
3133 <match> is a keyword indicating how to look for a specific pattern in the
3134 response. The keyword may be one of "status", "rstatus",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003135 "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01003136 exclamation mark ("!") to negate the match. Spaces are allowed
3137 between the exclamation mark and the keyword. See below for more
3138 details on the supported keywords.
3139
3140 <pattern> is the pattern to look for. It may be a string or a regular
3141 expression. If the pattern contains spaces, they must be escaped
3142 with the usual backslash ('\').
3143
3144 By default, "option httpchk" considers that response statuses 2xx and 3xx
3145 are valid, and that others are invalid. When "http-check expect" is used,
3146 it defines what is considered valid or invalid. Only one "http-check"
3147 statement is supported in a backend. If a server fails to respond or times
3148 out, the check obviously fails. The available matches are :
3149
3150 status <string> : test the exact string match for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003151 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003152 response's status code is exactly this string. If the
3153 "status" keyword is prefixed with "!", then the response
3154 will be considered invalid if the status code matches.
3155
3156 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003157 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003158 response's status code matches the expression. If the
3159 "rstatus" keyword is prefixed with "!", then the response
3160 will be considered invalid if the status code matches.
3161 This is mostly used to check for multiple codes.
3162
3163 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003164 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003165 response's body contains this exact string. If the
3166 "string" keyword is prefixed with "!", then the response
3167 will be considered invalid if the body contains this
3168 string. This can be used to look for a mandatory word at
3169 the end of a dynamic page, or to detect a failure when a
3170 specific error appears on the check page (eg: a stack
3171 trace).
3172
3173 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003174 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003175 response's body matches this expression. If the "rstring"
3176 keyword is prefixed with "!", then the response will be
3177 considered invalid if the body matches the expression.
3178 This can be used to look for a mandatory word at the end
3179 of a dynamic page, or to detect a failure when a specific
3180 error appears on the check page (eg: a stack trace).
3181
3182 It is important to note that the responses will be limited to a certain size
3183 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
3184 Thus, too large responses may not contain the mandatory pattern when using
3185 "string" or "rstring". If a large response is absolutely required, it is
3186 possible to change the default max size by setting the global variable.
3187 However, it is worth keeping in mind that parsing very large responses can
3188 waste some CPU cycles, especially when regular expressions are used, and that
3189 it is always better to focus the checks on smaller resources.
3190
Cyril Bonté32602d22015-01-30 00:07:07 +01003191 Also "http-check expect" doesn't support HTTP keep-alive. Keep in mind that it
3192 will automatically append a "Connection: close" header, meaning that this
3193 header should not be present in the request provided by "option httpchk".
3194
Willy Tarreaubd741542010-03-16 18:46:54 +01003195 Last, if "http-check expect" is combined with "http-check disable-on-404",
3196 then this last one has precedence when the server responds with 404.
3197
3198 Examples :
3199 # only accept status 200 as valid
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003200 http-check expect status 200
Willy Tarreaubd741542010-03-16 18:46:54 +01003201
3202 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003203 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01003204
3205 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003206 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01003207
3208 # check that we have a correct hexadecimal tag before /html
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003209 http-check expect rstring <!--tag:[0-9a-f]*</html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01003210
Willy Tarreaubd741542010-03-16 18:46:54 +01003211 See also : "option httpchk", "http-check disable-on-404"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003212
3213
Willy Tarreauef781042010-01-27 11:53:01 +01003214http-check send-state
3215 Enable emission of a state header with HTTP health checks
3216 May be used in sections : defaults | frontend | listen | backend
3217 yes | no | yes | yes
3218 Arguments : none
3219
3220 When this option is set, haproxy will systematically send a special header
3221 "X-Haproxy-Server-State" with a list of parameters indicating to each server
3222 how they are seen by haproxy. This can be used for instance when a server is
3223 manipulated without access to haproxy and the operator needs to know whether
3224 haproxy still sees it up or not, or if the server is the last one in a farm.
3225
3226 The header is composed of fields delimited by semi-colons, the first of which
3227 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
3228 checks on the total number before transition, just as appears in the stats
3229 interface. Next headers are in the form "<variable>=<value>", indicating in
3230 no specific order some values available in the stats interface :
Joseph Lynch514061c2015-01-15 17:52:59 -08003231 - a variable "address", containing the address of the backend server.
3232 This corresponds to the <address> field in the server declaration. For
3233 unix domain sockets, it will read "unix".
3234
3235 - a variable "port", containing the port of the backend server. This
3236 corresponds to the <port> field in the server declaration. For unix
3237 domain sockets, it will read "unix".
3238
Willy Tarreauef781042010-01-27 11:53:01 +01003239 - a variable "name", containing the name of the backend followed by a slash
3240 ("/") then the name of the server. This can be used when a server is
3241 checked in multiple backends.
3242
3243 - a variable "node" containing the name of the haproxy node, as set in the
3244 global "node" variable, otherwise the system's hostname if unspecified.
3245
3246 - a variable "weight" indicating the weight of the server, a slash ("/")
3247 and the total weight of the farm (just counting usable servers). This
3248 helps to know if other servers are available to handle the load when this
3249 one fails.
3250
3251 - a variable "scur" indicating the current number of concurrent connections
3252 on the server, followed by a slash ("/") then the total number of
3253 connections on all servers of the same backend.
3254
3255 - a variable "qcur" indicating the current number of requests in the
3256 server's queue.
3257
3258 Example of a header received by the application server :
3259 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
3260 scur=13/22; qcur=0
3261
3262 See also : "option httpchk", "http-check disable-on-404"
3263
Willy Tarreauccbcc372012-12-27 12:37:57 +01003264http-request { allow | deny | tarpit | auth [realm <realm>] | redirect <rule> |
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003265 add-header <name> <fmt> | set-header <name> <fmt> |
Willy Tarreaua9083d02015-05-08 15:27:59 +02003266 capture <sample> len <length> |
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003267 del-header <name> | set-nice <nice> | set-log-level <level> |
Sasha Pachev218f0642014-06-16 12:05:59 -06003268 replace-header <name> <match-regex> <replace-fmt> |
3269 replace-value <name> <match-regex> <replace-fmt> |
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01003270 set-method <fmt> | set-path <fmt> | set-query <fmt> |
3271 set-uri <fmt> | set-tos <tos> | set-mark <mark> |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003272 add-acl(<file name>) <key fmt> |
3273 del-acl(<file name>) <key fmt> |
3274 del-map(<file name>) <key fmt> |
Baptiste Assmannbb7e86a2014-09-03 18:29:47 +02003275 set-map(<file name>) <key fmt> <value fmt> |
Thierry FOURNIER90da1912015-03-05 11:17:06 +01003276 { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] |
3277 lua <function name>
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003278 }
Cyril Bontéf0c60612010-02-06 14:44:47 +01003279 [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003280 Access control for Layer 7 requests
3281
3282 May be used in sections: defaults | frontend | listen | backend
3283 no | yes | yes | yes
3284
Willy Tarreau20b0de52012-12-24 15:45:22 +01003285 The http-request statement defines a set of rules which apply to layer 7
3286 processing. The rules are evaluated in their declaration order when they are
3287 met in a frontend, listen or backend section. Any rule may optionally be
3288 followed by an ACL-based condition, in which case it will only be evaluated
3289 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003290
Willy Tarreau20b0de52012-12-24 15:45:22 +01003291 The first keyword is the rule's action. Currently supported actions include :
3292 - "allow" : this stops the evaluation of the rules and lets the request
3293 pass the check. No further "http-request" rules are evaluated.
3294
3295 - "deny" : this stops the evaluation of the rules and immediately rejects
3296 the request and emits an HTTP 403 error. No further "http-request" rules
3297 are evaluated.
3298
Willy Tarreauccbcc372012-12-27 12:37:57 +01003299 - "tarpit" : this stops the evaluation of the rules and immediately blocks
3300 the request without responding for a delay specified by "timeout tarpit"
3301 or "timeout connect" if the former is not set. After that delay, if the
3302 client is still connected, an HTTP error 500 is returned so that the
3303 client does not suspect it has been tarpitted. Logs will report the flags
3304 "PT". The goal of the tarpit rule is to slow down robots during an attack
3305 when they're limited on the number of concurrent requests. It can be very
3306 efficient against very dumb robots, and will significantly reduce the
3307 load on firewalls compared to a "deny" rule. But when facing "correctly"
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003308 developed robots, it can make things worse by forcing haproxy and the
Willy Tarreauccbcc372012-12-27 12:37:57 +01003309 front firewall to support insane number of concurrent connections.
3310
Willy Tarreau20b0de52012-12-24 15:45:22 +01003311 - "auth" : this stops the evaluation of the rules and immediately responds
3312 with an HTTP 401 or 407 error code to invite the user to present a valid
3313 user name and password. No further "http-request" rules are evaluated. An
3314 optional "realm" parameter is supported, it sets the authentication realm
3315 that is returned with the response (typically the application's name).
3316
Willy Tarreau81499eb2012-12-27 12:19:02 +01003317 - "redirect" : this performs an HTTP redirection based on a redirect rule.
3318 This is exactly the same as the "redirect" statement except that it
3319 inserts a redirect rule which can be processed in the middle of other
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01003320 "http-request" rules and that these rules use the "log-format" strings.
3321 See the "redirect" keyword for the rule's syntax.
Willy Tarreau81499eb2012-12-27 12:19:02 +01003322
Willy Tarreau20b0de52012-12-24 15:45:22 +01003323 - "add-header" appends an HTTP header field whose name is specified in
3324 <name> and whose value is defined by <fmt> which follows the log-format
3325 rules (see Custom Log Format in section 8.2.4). This is particularly
3326 useful to pass connection-specific information to the server (eg: the
3327 client's SSL certificate), or to combine several headers into one. This
3328 rule is not final, so it is possible to add other similar rules. Note
3329 that header addition is performed immediately, so one rule might reuse
3330 the resulting header from a previous rule.
3331
3332 - "set-header" does the same as "add-header" except that the header name
3333 is first removed if it existed. This is useful when passing security
3334 information to the server, where the header must not be manipulated by
Willy Tarreau85603282015-01-21 20:39:27 +01003335 external users. Note that the new value is computed before the removal so
3336 it is possible to concatenate a value to an existing header.
Willy Tarreau20b0de52012-12-24 15:45:22 +01003337
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003338 - "del-header" removes all HTTP header fields whose name is specified in
3339 <name>.
3340
Sasha Pachev218f0642014-06-16 12:05:59 -06003341 - "replace-header" matches the regular expression in all occurrences of
3342 header field <name> according to <match-regex>, and replaces them with
3343 the <replace-fmt> argument. Format characters are allowed in replace-fmt
3344 and work like in <fmt> arguments in "add-header". The match is only
3345 case-sensitive. It is important to understand that this action only
3346 considers whole header lines, regardless of the number of values they
3347 may contain. This usage is suited to headers naturally containing commas
3348 in their value, such as If-Modified-Since and so on.
3349
3350 Example:
3351
3352 http-request replace-header Cookie foo=([^;]*);(.*) foo=\1;ip=%bi;\2
3353
3354 applied to:
3355
3356 Cookie: foo=foobar; expires=Tue, 14-Jun-2016 01:40:45 GMT;
3357
3358 outputs:
3359
3360 Cookie: foo=foobar;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT;
3361
3362 assuming the backend IP is 192.168.1.20
3363
3364 - "replace-value" works like "replace-header" except that it matches the
3365 regex against every comma-delimited value of the header field <name>
3366 instead of the entire header. This is suited for all headers which are
3367 allowed to carry more than one value. An example could be the Accept
3368 header.
3369
3370 Example:
3371
3372 http-request replace-value X-Forwarded-For ^192\.168\.(.*)$ 172.16.\1
3373
3374 applied to:
3375
3376 X-Forwarded-For: 192.168.10.1, 192.168.13.24, 10.0.0.37
3377
3378 outputs:
3379
3380 X-Forwarded-For: 172.16.10.1, 172.16.13.24, 10.0.0.37
3381
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01003382 - "set-method" rewrites the request method with the result of the
3383 evaluation of format string <fmt>. There should be very few valid reasons
3384 for having to do so as this is more likely to break something than to fix
3385 it.
3386
3387 - "set-path" rewrites the request path with the result of the evaluation of
3388 format string <fmt>. The query string, if any, is left intact. If a
3389 scheme and authority is found before the path, they are left intact as
3390 well. If the request doesn't have a path ("*"), this one is replaced with
3391 the format. This can be used to prepend a directory component in front of
3392 a path for example. See also "set-query" and "set-uri".
3393
3394 Example :
3395 # prepend the host name before the path
3396 http-request set-path /%[hdr(host)]%[path]
3397
3398 - "set-query" rewrites the request's query string which appears after the
3399 first question mark ("?") with the result of the evaluation of format
3400 string <fmt>. The part prior to the question mark is left intact. If the
3401 request doesn't contain a question mark and the new value is not empty,
3402 then one is added at the end of the URI, followed by the new value. If
3403 a question mark was present, it will never be removed even if the value
3404 is empty. This can be used to add or remove parameters from the query
3405 string. See also "set-query" and "set-uri".
3406
3407 Example :
3408 # replace "%3D" with "=" in the query string
3409 http-request set-query %[query,regsub(%3D,=,g)]
3410
3411 - "set-uri" rewrites the request URI with the result of the evaluation of
3412 format string <fmt>. The scheme, authority, path and query string are all
3413 replaced at once. This can be used to rewrite hosts in front of proxies,
3414 or to perform complex modifications to the URI such as moving parts
3415 between the path and the query string. See also "set-path" and
3416 "set-query".
3417
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003418 - "set-nice" sets the "nice" factor of the current request being processed.
3419 It only has effect against the other requests being processed at the same
3420 time. The default value is 0, unless altered by the "nice" setting on the
3421 "bind" line. The accepted range is -1024..1024. The higher the value, the
3422 nicest the request will be. Lower values will make the request more
3423 important than other ones. This can be useful to improve the speed of
3424 some requests, or lower the priority of non-important requests. Using
3425 this setting without prior experimentation can cause some major slowdown.
3426
Willy Tarreau9a355ec2013-06-11 17:45:46 +02003427 - "set-log-level" is used to change the log level of the current request
3428 when a certain condition is met. Valid levels are the 8 syslog levels
3429 (see the "log" keyword) plus the special level "silent" which disables
3430 logging for this request. This rule is not final so the last matching
3431 rule wins. This rule can be useful to disable health checks coming from
3432 another equipment.
3433
Willy Tarreau42cf39e2013-06-11 18:51:32 +02003434 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
3435 the client to the value passed in <tos> on platforms which support this.
3436 This value represents the whole 8 bits of the IP TOS field, and can be
3437 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
3438 that only the 6 higher bits are used in DSCP or TOS, and the two lower
3439 bits are always 0. This can be used to adjust some routing behaviour on
3440 border routers based on some information from the request. See RFC 2474,
3441 2597, 3260 and 4594 for more information.
3442
Willy Tarreau51347ed2013-06-11 19:34:13 +02003443 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
3444 client to the value passed in <mark> on platforms which support it. This
3445 value is an unsigned 32 bit value which can be matched by netfilter and
3446 by the routing table. It can be expressed both in decimal or hexadecimal
3447 format (prefixed by "0x"). This can be useful to force certain packets to
3448 take a different route (for example a cheaper network path for bulk
3449 downloads). This works on Linux kernels 2.6.32 and above and requires
3450 admin privileges.
3451
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003452 - "add-acl" is used to add a new entry into an ACL. The ACL must be loaded
3453 from a file (even a dummy empty file). The file name of the ACL to be
3454 updated is passed between parentheses. It takes one argument: <key fmt>,
3455 which follows log-format rules, to collect content of the new entry. It
3456 performs a lookup in the ACL before insertion, to avoid duplicated (or
3457 more) values. This lookup is done by a linear search and can be expensive
3458 with large lists! It is the equivalent of the "add acl" command from the
3459 stats socket, but can be triggered by an HTTP request.
3460
3461 - "del-acl" is used to delete an entry from an ACL. The ACL must be loaded
3462 from a file (even a dummy empty file). The file name of the ACL to be
3463 updated is passed between parentheses. It takes one argument: <key fmt>,
3464 which follows log-format rules, to collect content of the entry to delete.
3465 It is the equivalent of the "del acl" command from the stats socket, but
3466 can be triggered by an HTTP request.
3467
3468 - "del-map" is used to delete an entry from a MAP. The MAP must be loaded
3469 from a file (even a dummy empty file). The file name of the MAP to be
3470 updated is passed between parentheses. It takes one argument: <key fmt>,
3471 which follows log-format rules, to collect content of the entry to delete.
3472 It takes one argument: "file name" It is the equivalent of the "del map"
3473 command from the stats socket, but can be triggered by an HTTP request.
3474
3475 - "set-map" is used to add a new entry into a MAP. The MAP must be loaded
3476 from a file (even a dummy empty file). The file name of the MAP to be
3477 updated is passed between parentheses. It takes 2 arguments: <key fmt>,
3478 which follows log-format rules, used to collect MAP key, and <value fmt>,
3479 which follows log-format rules, used to collect content for the new entry.
3480 It performs a lookup in the MAP before insertion, to avoid duplicated (or
3481 more) values. This lookup is done by a linear search and can be expensive
3482 with large lists! It is the equivalent of the "set map" command from the
3483 stats socket, but can be triggered by an HTTP request.
3484
Willy Tarreaua9083d02015-05-08 15:27:59 +02003485 - capture <sample> len <length> :
3486 captures sample expression <sample> from the request buffer, and converts
3487 it to a string of at most <len> characters. The resulting string is
3488 stored into the next request "capture" slot, so it will possibly appear
3489 next to some captured HTTP headers. It will then automatically appear in
3490 the logs, and it will be possible to extract it using sample fetch rules
3491 to feed it into headers or anything. The length should be limited given
3492 that this size will be allocated for each capture during the whole
3493 session life. Please check section 7.3 (Fetching samples) and "capture
3494 request header" for more information.
3495
Willy Tarreau09448f72014-06-25 18:12:15 +02003496 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
3497 enables tracking of sticky counters from current request. These rules
3498 do not stop evaluation and do not change default action. Three sets of
3499 counters may be simultaneously tracked by the same connection. The first
3500 "track-sc0" rule executed enables tracking of the counters of the
3501 specified table as the first set. The first "track-sc1" rule executed
3502 enables tracking of the counters of the specified table as the second
3503 set. The first "track-sc2" rule executed enables tracking of the
3504 counters of the specified table as the third set. It is a recommended
3505 practice to use the first set of counters for the per-frontend counters
3506 and the second set for the per-backend ones. But this is just a
3507 guideline, all may be used everywhere.
3508
3509 These actions take one or two arguments :
3510 <key> is mandatory, and is a sample expression rule as described
3511 in section 7.3. It describes what elements of the incoming
3512 request or connection will be analysed, extracted, combined,
3513 and used to select which table entry to update the counters.
3514
3515 <table> is an optional table to be used instead of the default one,
3516 which is the stick-table declared in the current proxy. All
3517 the counters for the matches and updates for the key will
3518 then be performed in that table until the session ends.
3519
3520 Once a "track-sc*" rule is executed, the key is looked up in the table
3521 and if it is not found, an entry is allocated for it. Then a pointer to
3522 that entry is kept during all the session's life, and this entry's
3523 counters are updated as often as possible, every time the session's
3524 counters are updated, and also systematically when the session ends.
3525 Counters are only updated for events that happen after the tracking has
3526 been started. As an exception, connection counters and request counters
3527 are systematically updated so that they reflect useful information.
3528
3529 If the entry tracks concurrent connection counters, one connection is
3530 counted for as long as the entry is tracked, and the entry will not
3531 expire during that time. Tracking counters also provides a performance
3532 advantage over just checking the keys, because only one table lookup is
3533 performed for all ACL checks that make use of it.
3534
Thierry FOURNIER90da1912015-03-05 11:17:06 +01003535 - "lua" is used to run a Lua function if the action is executed. The single
3536 parameter is the name of the function to run. The prototype of the
3537 function is documented in the API documentation.
3538
Willy Tarreau20b0de52012-12-24 15:45:22 +01003539 There is no limit to the number of http-request statements per instance.
3540
3541 It is important to know that http-request rules are processed very early in
3542 the HTTP processing, just after "block" rules and before "reqdel" or "reqrep"
3543 rules. That way, headers added by "add-header"/"set-header" are visible by
3544 almost all further ACL rules.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003545
3546 Example:
Cyril Bonté78caf842010-03-10 22:41:43 +01003547 acl nagios src 192.168.129.3
3548 acl local_net src 192.168.0.0/16
3549 acl auth_ok http_auth(L1)
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003550
Cyril Bonté78caf842010-03-10 22:41:43 +01003551 http-request allow if nagios
3552 http-request allow if local_net auth_ok
3553 http-request auth realm Gimme if local_net auth_ok
3554 http-request deny
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003555
Cyril Bonté78caf842010-03-10 22:41:43 +01003556 Example:
3557 acl auth_ok http_auth_group(L1) G1
Cyril Bonté78caf842010-03-10 22:41:43 +01003558 http-request auth unless auth_ok
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003559
Willy Tarreau20b0de52012-12-24 15:45:22 +01003560 Example:
3561 http-request set-header X-Haproxy-Current-Date %T
3562 http-request set-header X-SSL %[ssl_fc]
3563 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id]
3564 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
3565 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
3566 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
3567 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
3568 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
3569 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
3570
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003571 Example:
3572 acl key req.hdr(X-Add-Acl-Key) -m found
3573 acl add path /addacl
3574 acl del path /delacl
3575
3576 acl myhost hdr(Host) -f myhost.lst
3577
3578 http-request add-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key add
3579 http-request del-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key del
3580
3581 Example:
3582 acl value req.hdr(X-Value) -m found
3583 acl setmap path /setmap
3584 acl delmap path /delmap
3585
3586 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
3587
3588 http-request set-map(map.lst) %[src] %[req.hdr(X-Value)] if setmap value
3589 http-request del-map(map.lst) %[src] if delmap
3590
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02003591 See also : "stats http-request", section 3.4 about userlists and section 7
3592 about ACL usage.
Willy Tarreauef781042010-01-27 11:53:01 +01003593
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003594http-response { allow | deny | add-header <name> <fmt> | set-nice <nice> |
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003595 set-header <name> <fmt> | del-header <name> |
Sasha Pachev218f0642014-06-16 12:05:59 -06003596 replace-header <name> <regex-match> <replace-fmt> |
3597 replace-value <name> <regex-match> <replace-fmt> |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003598 set-log-level <level> | set-mark <mark> | set-tos <tos> |
3599 add-acl(<file name>) <key fmt> |
3600 del-acl(<file name>) <key fmt> |
3601 del-map(<file name>) <key fmt> |
Thierry FOURNIER90da1912015-03-05 11:17:06 +01003602 set-map(<file name>) <key fmt> <value fmt> |
3603 lua <function name>
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003604 }
Lukas Tribus2dd1d1a2013-06-19 23:34:41 +02003605 [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02003606 Access control for Layer 7 responses
3607
3608 May be used in sections: defaults | frontend | listen | backend
3609 no | yes | yes | yes
3610
3611 The http-response statement defines a set of rules which apply to layer 7
3612 processing. The rules are evaluated in their declaration order when they are
3613 met in a frontend, listen or backend section. Any rule may optionally be
3614 followed by an ACL-based condition, in which case it will only be evaluated
3615 if the condition is true. Since these rules apply on responses, the backend
3616 rules are applied first, followed by the frontend's rules.
3617
3618 The first keyword is the rule's action. Currently supported actions include :
3619 - "allow" : this stops the evaluation of the rules and lets the response
3620 pass the check. No further "http-response" rules are evaluated for the
3621 current section.
3622
3623 - "deny" : this stops the evaluation of the rules and immediately rejects
3624 the response and emits an HTTP 502 error. No further "http-response"
3625 rules are evaluated.
3626
3627 - "add-header" appends an HTTP header field whose name is specified in
3628 <name> and whose value is defined by <fmt> which follows the log-format
3629 rules (see Custom Log Format in section 8.2.4). This may be used to send
3630 a cookie to a client for example, or to pass some internal information.
3631 This rule is not final, so it is possible to add other similar rules.
3632 Note that header addition is performed immediately, so one rule might
3633 reuse the resulting header from a previous rule.
3634
3635 - "set-header" does the same as "add-header" except that the header name
3636 is first removed if it existed. This is useful when passing security
3637 information to the server, where the header must not be manipulated by
3638 external users.
3639
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003640 - "del-header" removes all HTTP header fields whose name is specified in
3641 <name>.
3642
Sasha Pachev218f0642014-06-16 12:05:59 -06003643 - "replace-header" matches the regular expression in all occurrences of
3644 header field <name> according to <match-regex>, and replaces them with
3645 the <replace-fmt> argument. Format characters are allowed in replace-fmt
3646 and work like in <fmt> arguments in "add-header". The match is only
3647 case-sensitive. It is important to understand that this action only
3648 considers whole header lines, regardless of the number of values they
3649 may contain. This usage is suited to headers naturally containing commas
3650 in their value, such as Set-Cookie, Expires and so on.
3651
3652 Example:
3653
3654 http-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
3655
3656 applied to:
3657
3658 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
3659
3660 outputs:
3661
3662 Set-Cookie: C=1;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT
3663
3664 assuming the backend IP is 192.168.1.20.
3665
3666 - "replace-value" works like "replace-header" except that it matches the
3667 regex against every comma-delimited value of the header field <name>
3668 instead of the entire header. This is suited for all headers which are
3669 allowed to carry more than one value. An example could be the Accept
3670 header.
3671
3672 Example:
3673
3674 http-response replace-value Cache-control ^public$ private
3675
3676 applied to:
3677
3678 Cache-Control: max-age=3600, public
3679
3680 outputs:
3681
3682 Cache-Control: max-age=3600, private
3683
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003684 - "set-nice" sets the "nice" factor of the current request being processed.
3685 It only has effect against the other requests being processed at the same
3686 time. The default value is 0, unless altered by the "nice" setting on the
3687 "bind" line. The accepted range is -1024..1024. The higher the value, the
3688 nicest the request will be. Lower values will make the request more
3689 important than other ones. This can be useful to improve the speed of
3690 some requests, or lower the priority of non-important requests. Using
3691 this setting without prior experimentation can cause some major slowdown.
3692
Willy Tarreau9a355ec2013-06-11 17:45:46 +02003693 - "set-log-level" is used to change the log level of the current request
3694 when a certain condition is met. Valid levels are the 8 syslog levels
3695 (see the "log" keyword) plus the special level "silent" which disables
3696 logging for this request. This rule is not final so the last matching
3697 rule wins. This rule can be useful to disable health checks coming from
3698 another equipment.
3699
Willy Tarreau42cf39e2013-06-11 18:51:32 +02003700 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
3701 the client to the value passed in <tos> on platforms which support this.
3702 This value represents the whole 8 bits of the IP TOS field, and can be
3703 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
3704 that only the 6 higher bits are used in DSCP or TOS, and the two lower
3705 bits are always 0. This can be used to adjust some routing behaviour on
3706 border routers based on some information from the request. See RFC 2474,
3707 2597, 3260 and 4594 for more information.
3708
Willy Tarreau51347ed2013-06-11 19:34:13 +02003709 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
3710 client to the value passed in <mark> on platforms which support it. This
3711 value is an unsigned 32 bit value which can be matched by netfilter and
3712 by the routing table. It can be expressed both in decimal or hexadecimal
3713 format (prefixed by "0x"). This can be useful to force certain packets to
3714 take a different route (for example a cheaper network path for bulk
3715 downloads). This works on Linux kernels 2.6.32 and above and requires
3716 admin privileges.
3717
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003718 - "add-acl" is used to add a new entry into an ACL. The ACL must be loaded
3719 from a file (even a dummy empty file). The file name of the ACL to be
3720 updated is passed between parentheses. It takes one argument: <key fmt>,
3721 which follows log-format rules, to collect content of the new entry. It
3722 performs a lookup in the ACL before insertion, to avoid duplicated (or
3723 more) values. This lookup is done by a linear search and can be expensive
3724 with large lists! It is the equivalent of the "add acl" command from the
3725 stats socket, but can be triggered by an HTTP response.
3726
3727 - "del-acl" is used to delete an entry from an ACL. The ACL must be loaded
3728 from a file (even a dummy empty file). The file name of the ACL to be
3729 updated is passed between parentheses. It takes one argument: <key fmt>,
3730 which follows log-format rules, to collect content of the entry to delete.
3731 It is the equivalent of the "del acl" command from the stats socket, but
3732 can be triggered by an HTTP response.
3733
3734 - "del-map" is used to delete an entry from a MAP. The MAP must be loaded
3735 from a file (even a dummy empty file). The file name of the MAP to be
3736 updated is passed between parentheses. It takes one argument: <key fmt>,
3737 which follows log-format rules, to collect content of the entry to delete.
3738 It takes one argument: "file name" It is the equivalent of the "del map"
3739 command from the stats socket, but can be triggered by an HTTP response.
3740
3741 - "set-map" is used to add a new entry into a MAP. The MAP must be loaded
3742 from a file (even a dummy empty file). The file name of the MAP to be
3743 updated is passed between parentheses. It takes 2 arguments: <key fmt>,
3744 which follows log-format rules, used to collect MAP key, and <value fmt>,
3745 which follows log-format rules, used to collect content for the new entry.
3746 It performs a lookup in the MAP before insertion, to avoid duplicated (or
3747 more) values. This lookup is done by a linear search and can be expensive
3748 with large lists! It is the equivalent of the "set map" command from the
3749 stats socket, but can be triggered by an HTTP response.
3750
Thierry FOURNIER90da1912015-03-05 11:17:06 +01003751 - "lua" is used to run a Lua function if the action is executed. The single
3752 parameter is the name of the function to run. The prototype of the
3753 function is documented in the API documentation.
3754
Willy Tarreaue365c0b2013-06-11 16:06:12 +02003755 There is no limit to the number of http-response statements per instance.
3756
Godbach09250262013-07-02 01:19:15 +08003757 It is important to know that http-response rules are processed very early in
Willy Tarreaue365c0b2013-06-11 16:06:12 +02003758 the HTTP processing, before "reqdel" or "reqrep" rules. That way, headers
3759 added by "add-header"/"set-header" are visible by almost all further ACL
3760 rules.
3761
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003762 Example:
3763 acl key_acl res.hdr(X-Acl-Key) -m found
3764
3765 acl myhost hdr(Host) -f myhost.lst
3766
3767 http-response add-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
3768 http-response del-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
3769
3770 Example:
3771 acl value res.hdr(X-Value) -m found
3772
3773 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
3774
3775 http-response set-map(map.lst) %[src] %[res.hdr(X-Value)] if value
3776 http-response del-map(map.lst) %[src] if ! value
3777
Willy Tarreaue365c0b2013-06-11 16:06:12 +02003778 See also : "http-request", section 3.4 about userlists and section 7 about
3779 ACL usage.
3780
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02003781
Mark Lamourinec2247f02012-01-04 13:02:01 -05003782http-send-name-header [<header>]
3783 Add the server name to a request. Use the header string given by <header>
3784
3785 May be used in sections: defaults | frontend | listen | backend
3786 yes | no | yes | yes
3787
3788 Arguments :
3789
3790 <header> The header string to use to send the server name
3791
3792 The "http-send-name-header" statement causes the name of the target
3793 server to be added to the headers of an HTTP request. The name
3794 is added with the header string proved.
3795
3796 See also : "server"
3797
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01003798id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02003799 Set a persistent ID to a proxy.
3800 May be used in sections : defaults | frontend | listen | backend
3801 no | yes | yes | yes
3802 Arguments : none
3803
3804 Set a persistent ID for the proxy. This ID must be unique and positive.
3805 An unused ID will automatically be assigned if unset. The first assigned
3806 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01003807
3808
Cyril Bonté0d4bf012010-04-25 23:21:46 +02003809ignore-persist { if | unless } <condition>
3810 Declare a condition to ignore persistence
3811 May be used in sections: defaults | frontend | listen | backend
3812 no | yes | yes | yes
3813
3814 By default, when cookie persistence is enabled, every requests containing
3815 the cookie are unconditionally persistent (assuming the target server is up
3816 and running).
3817
3818 The "ignore-persist" statement allows one to declare various ACL-based
3819 conditions which, when met, will cause a request to ignore persistence.
3820 This is sometimes useful to load balance requests for static files, which
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003821 often don't require persistence. This can also be used to fully disable
Cyril Bonté0d4bf012010-04-25 23:21:46 +02003822 persistence for a specific User-Agent (for example, some web crawler bots).
3823
3824 Combined with "appsession", it can also help reduce HAProxy memory usage, as
3825 the appsession table won't grow if persistence is ignored.
3826
3827 The persistence is ignored when an "if" condition is met, or unless an
3828 "unless" condition is met.
3829
3830 See also : "force-persist", "cookie", and section 7 about ACL usage.
3831
3832
Willy Tarreau2769aa02007-12-27 18:26:09 +01003833log global
Willy Tarreau18324f52014-06-27 18:10:07 +02003834log <address> [len <length>] <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02003835no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01003836 Enable per-instance logging of events and traffic.
3837 May be used in sections : defaults | frontend | listen | backend
3838 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02003839
3840 Prefix :
3841 no should be used when the logger list must be flushed. For example,
3842 if you don't want to inherit from the default logger list. This
3843 prefix does not allow arguments.
3844
Willy Tarreau2769aa02007-12-27 18:26:09 +01003845 Arguments :
3846 global should be used when the instance's logging parameters are the
3847 same as the global ones. This is the most common usage. "global"
3848 replaces <address>, <facility> and <level> with those of the log
3849 entries found in the "global" section. Only one "log global"
3850 statement may be used per instance, and this form takes no other
3851 parameter.
3852
3853 <address> indicates where to send the logs. It takes the same format as
3854 for the "global" section's logs, and can be one of :
3855
3856 - An IPv4 address optionally followed by a colon (':') and a UDP
3857 port. If no port is specified, 514 is used by default (the
3858 standard syslog port).
3859
David du Colombier24bb5f52011-03-17 10:40:23 +01003860 - An IPv6 address followed by a colon (':') and optionally a UDP
3861 port. If no port is specified, 514 is used by default (the
3862 standard syslog port).
3863
Willy Tarreau2769aa02007-12-27 18:26:09 +01003864 - A filesystem path to a UNIX domain socket, keeping in mind
3865 considerations for chroot (be sure the path is accessible
3866 inside the chroot) and uid/gid (be sure the path is
3867 appropriately writeable).
3868
William Lallemandb2f07452015-05-12 14:27:13 +02003869 You may want to reference some environment variables in the
3870 address parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01003871
Willy Tarreau18324f52014-06-27 18:10:07 +02003872 <length> is an optional maximum line length. Log lines larger than this
3873 value will be truncated before being sent. The reason is that
3874 syslog servers act differently on log line length. All servers
3875 support the default value of 1024, but some servers simply drop
3876 larger lines while others do log them. If a server supports long
3877 lines, it may make sense to set this value here in order to avoid
3878 truncating long lines. Similarly, if a server drops long lines,
3879 it is preferable to truncate them before sending them. Accepted
3880 values are 80 to 65535 inclusive. The default value of 1024 is
3881 generally fine for all standard usages. Some specific cases of
3882 long captures or JSON-formated logs may require larger values.
3883
Willy Tarreau2769aa02007-12-27 18:26:09 +01003884 <facility> must be one of the 24 standard syslog facilities :
3885
3886 kern user mail daemon auth syslog lpr news
3887 uucp cron auth2 ftp ntp audit alert cron2
3888 local0 local1 local2 local3 local4 local5 local6 local7
3889
3890 <level> is optional and can be specified to filter outgoing messages. By
3891 default, all messages are sent. If a level is specified, only
3892 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02003893 will be sent. An optional minimum level can be specified. If it
3894 is set, logs emitted with a more severe level than this one will
3895 be capped to this level. This is used to avoid sending "emerg"
3896 messages on all terminals on some default syslog configurations.
3897 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01003898
3899 emerg alert crit err warning notice info debug
3900
William Lallemand0f99e342011-10-12 17:50:54 +02003901 It is important to keep in mind that it is the frontend which decides what to
3902 log from a connection, and that in case of content switching, the log entries
3903 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003904
3905 However, backend log declaration define how and where servers status changes
3906 will be logged. Level "notice" will be used to indicate a server going up,
3907 "warning" will be used for termination signals and definitive service
3908 termination, and "alert" will be used for when a server goes down.
3909
3910 Note : According to RFC3164, messages are truncated to 1024 bytes before
3911 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003912
3913 Example :
3914 log global
Willy Tarreauf7edefa2009-05-10 17:20:05 +02003915 log 127.0.0.1:514 local0 notice # only send important events
3916 log 127.0.0.1:514 local0 notice notice # same but limit output level
William Lallemandb2f07452015-05-12 14:27:13 +02003917 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
Willy Tarreaudad36a32013-03-11 01:20:04 +01003918
Willy Tarreau2769aa02007-12-27 18:26:09 +01003919
William Lallemand48940402012-01-30 16:47:22 +01003920log-format <string>
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01003921 Specifies the log format string to use for traffic logs
3922 May be used in sections: defaults | frontend | listen | backend
3923 yes | yes | yes | no
William Lallemand48940402012-01-30 16:47:22 +01003924
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01003925 This directive specifies the log format string that will be used for all logs
3926 resulting from traffic passing through the frontend using this line. If the
3927 directive is used in a defaults section, all subsequent frontends will use
3928 the same log format. Please see section 8.2.4 which covers the log format
3929 string in depth.
William Lallemand48940402012-01-30 16:47:22 +01003930
Willy Tarreau094af4e2015-01-07 15:03:42 +01003931log-tag <string>
3932 Specifies the log tag to use for all outgoing logs
3933 May be used in sections: defaults | frontend | listen | backend
3934 yes | yes | yes | yes
3935
3936 Sets the tag field in the syslog header to this string. It defaults to the
3937 log-tag set in the global section, otherwise the program name as launched
3938 from the command line, which usually is "haproxy". Sometimes it can be useful
3939 to differentiate between multiple processes running on the same host, or to
3940 differentiate customer instances running in the same process. In the backend,
3941 logs about servers up/down will use this tag. As a hint, it can be convenient
3942 to set a log-tag related to a hosted customer in a defaults section then put
3943 all the frontends and backends for that customer, then start another customer
3944 in a new defaults section. See also the global "log-tag" directive.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003945
Willy Tarreauc35362a2014-04-25 13:58:37 +02003946max-keep-alive-queue <value>
3947 Set the maximum server queue size for maintaining keep-alive connections
3948 May be used in sections: defaults | frontend | listen | backend
3949 yes | no | yes | yes
3950
3951 HTTP keep-alive tries to reuse the same server connection whenever possible,
3952 but sometimes it can be counter-productive, for example if a server has a lot
3953 of connections while other ones are idle. This is especially true for static
3954 servers.
3955
3956 The purpose of this setting is to set a threshold on the number of queued
3957 connections at which haproxy stops trying to reuse the same server and prefers
3958 to find another one. The default value, -1, means there is no limit. A value
3959 of zero means that keep-alive requests will never be queued. For very close
3960 servers which can be reached with a low latency and which are not sensible to
3961 breaking keep-alive, a low value is recommended (eg: local static server can
3962 use a value of 10 or less). For remote servers suffering from a high latency,
3963 higher values might be needed to cover for the latency and/or the cost of
3964 picking a different server.
3965
3966 Note that this has no impact on responses which are maintained to the same
3967 server consecutively to a 401 response. They will still go to the same server
3968 even if they have to be queued.
3969
3970 See also : "option http-server-close", "option prefer-last-server", server
3971 "maxconn" and cookie persistence.
3972
3973
Willy Tarreau2769aa02007-12-27 18:26:09 +01003974maxconn <conns>
3975 Fix the maximum number of concurrent connections on a frontend
3976 May be used in sections : defaults | frontend | listen | backend
3977 yes | yes | yes | no
3978 Arguments :
3979 <conns> is the maximum number of concurrent connections the frontend will
3980 accept to serve. Excess connections will be queued by the system
3981 in the socket's listen queue and will be served once a connection
3982 closes.
3983
3984 If the system supports it, it can be useful on big sites to raise this limit
3985 very high so that haproxy manages connection queues, instead of leaving the
3986 clients with unanswered connection attempts. This value should not exceed the
3987 global maxconn. Also, keep in mind that a connection contains two buffers
3988 of 8kB each, as well as some other data resulting in about 17 kB of RAM being
3989 consumed per established connection. That means that a medium system equipped
3990 with 1GB of RAM can withstand around 40000-50000 concurrent connections if
3991 properly tuned.
3992
3993 Also, when <conns> is set to large values, it is possible that the servers
3994 are not sized to accept such loads, and for this reason it is generally wise
3995 to assign them some reasonable connection limits.
3996
Vincent Bernat6341be52012-06-27 17:18:30 +02003997 By default, this value is set to 2000.
3998
Willy Tarreau2769aa02007-12-27 18:26:09 +01003999 See also : "server", global section's "maxconn", "fullconn"
4000
4001
4002mode { tcp|http|health }
4003 Set the running mode or protocol of the instance
4004 May be used in sections : defaults | frontend | listen | backend
4005 yes | yes | yes | yes
4006 Arguments :
4007 tcp The instance will work in pure TCP mode. A full-duplex connection
4008 will be established between clients and servers, and no layer 7
4009 examination will be performed. This is the default mode. It
4010 should be used for SSL, SSH, SMTP, ...
4011
4012 http The instance will work in HTTP mode. The client request will be
4013 analyzed in depth before connecting to any server. Any request
4014 which is not RFC-compliant will be rejected. Layer 7 filtering,
4015 processing and switching will be possible. This is the mode which
4016 brings HAProxy most of its value.
4017
4018 health The instance will work in "health" mode. It will just reply "OK"
Willy Tarreau82569f92012-09-27 23:48:56 +02004019 to incoming connections and close the connection. Alternatively,
4020 If the "httpchk" option is set, "HTTP/1.0 200 OK" will be sent
4021 instead. Nothing will be logged in either case. This mode is used
4022 to reply to external components health checks. This mode is
4023 deprecated and should not be used anymore as it is possible to do
4024 the same and even better by combining TCP or HTTP modes with the
4025 "monitor" keyword.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004026
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004027 When doing content switching, it is mandatory that the frontend and the
4028 backend are in the same mode (generally HTTP), otherwise the configuration
4029 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004030
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004031 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01004032 defaults http_instances
4033 mode http
4034
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004035 See also : "monitor", "monitor-net"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004036
Willy Tarreau0ba27502007-12-24 16:55:16 +01004037
Cyril Bontéf0c60612010-02-06 14:44:47 +01004038monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01004039 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004040 May be used in sections : defaults | frontend | listen | backend
4041 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01004042 Arguments :
4043 if <cond> the monitor request will fail if the condition is satisfied,
4044 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004045 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01004046 are met, for instance a low number of servers both in a
4047 backend and its backup.
4048
4049 unless <cond> the monitor request will succeed only if the condition is
4050 satisfied, and will fail otherwise. Such a condition may be
4051 based on a test on the presence of a minimum number of active
4052 servers in a list of backends.
4053
4054 This statement adds a condition which can force the response to a monitor
4055 request to report a failure. By default, when an external component queries
4056 the URI dedicated to monitoring, a 200 response is returned. When one of the
4057 conditions above is met, haproxy will return 503 instead of 200. This is
4058 very useful to report a site failure to an external component which may base
4059 routing advertisements between multiple sites on the availability reported by
4060 haproxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02004061 criterion. Note that "monitor fail" only works in HTTP mode. Both status
4062 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004063
4064 Example:
4065 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01004066 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01004067 acl site_dead nbsrv(dynamic) lt 2
4068 acl site_dead nbsrv(static) lt 2
4069 monitor-uri /site_alive
4070 monitor fail if site_dead
4071
Willy Tarreauae94d4d2011-05-11 16:28:49 +02004072 See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004073
4074
4075monitor-net <source>
4076 Declare a source network which is limited to monitor requests
4077 May be used in sections : defaults | frontend | listen | backend
4078 yes | yes | yes | no
4079 Arguments :
4080 <source> is the source IPv4 address or network which will only be able to
4081 get monitor responses to any request. It can be either an IPv4
4082 address, a host name, or an address followed by a slash ('/')
4083 followed by a mask.
4084
4085 In TCP mode, any connection coming from a source matching <source> will cause
4086 the connection to be immediately closed without any log. This allows another
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004087 equipment to probe the port and verify that it is still listening, without
Willy Tarreau2769aa02007-12-27 18:26:09 +01004088 forwarding the connection to a remote server.
4089
4090 In HTTP mode, a connection coming from a source matching <source> will be
4091 accepted, the following response will be sent without waiting for a request,
4092 then the connection will be closed : "HTTP/1.0 200 OK". This is normally
4093 enough for any front-end HTTP probe to detect that the service is UP and
Willy Tarreau82569f92012-09-27 23:48:56 +02004094 running without forwarding the request to a backend server. Note that this
4095 response is sent in raw format, without any transformation. This is important
4096 as it means that it will not be SSL-encrypted on SSL listeners.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004097
Willy Tarreau82569f92012-09-27 23:48:56 +02004098 Monitor requests are processed very early, just after tcp-request connection
4099 ACLs which are the only ones able to block them. These connections are short
4100 lived and never wait for any data from the client. They cannot be logged, and
4101 it is the intended purpose. They are only used to report HAProxy's health to
4102 an upper component, nothing more. Please note that "monitor fail" rules do
4103 not apply to connections intercepted by "monitor-net".
Willy Tarreau2769aa02007-12-27 18:26:09 +01004104
Willy Tarreau95cd2832010-03-04 23:36:33 +01004105 Last, please note that only one "monitor-net" statement can be specified in
4106 a frontend. If more than one is found, only the last one will be considered.
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004107
Willy Tarreau2769aa02007-12-27 18:26:09 +01004108 Example :
4109 # addresses .252 and .253 are just probing us.
4110 frontend www
4111 monitor-net 192.168.0.252/31
4112
4113 See also : "monitor fail", "monitor-uri"
4114
4115
4116monitor-uri <uri>
4117 Intercept a URI used by external components' monitor requests
4118 May be used in sections : defaults | frontend | listen | backend
4119 yes | yes | yes | no
4120 Arguments :
4121 <uri> is the exact URI which we want to intercept to return HAProxy's
4122 health status instead of forwarding the request.
4123
4124 When an HTTP request referencing <uri> will be received on a frontend,
4125 HAProxy will not forward it nor log it, but instead will return either
4126 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
4127 conditions defined with "monitor fail". This is normally enough for any
4128 front-end HTTP probe to detect that the service is UP and running without
4129 forwarding the request to a backend server. Note that the HTTP method, the
4130 version and all headers are ignored, but the request must at least be valid
4131 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
4132
4133 Monitor requests are processed very early. It is not possible to block nor
4134 divert them using ACLs. They cannot be logged either, and it is the intended
4135 purpose. They are only used to report HAProxy's health to an upper component,
4136 nothing more. However, it is possible to add any number of conditions using
4137 "monitor fail" and ACLs so that the result can be adjusted to whatever check
4138 can be imagined (most often the number of available servers in a backend).
4139
4140 Example :
4141 # Use /haproxy_test to report haproxy's status
4142 frontend www
4143 mode http
4144 monitor-uri /haproxy_test
4145
4146 See also : "monitor fail", "monitor-net"
4147
Willy Tarreau0ba27502007-12-24 16:55:16 +01004148
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004149option abortonclose
4150no option abortonclose
4151 Enable or disable early dropping of aborted requests pending in queues.
4152 May be used in sections : defaults | frontend | listen | backend
4153 yes | no | yes | yes
4154 Arguments : none
4155
4156 In presence of very high loads, the servers will take some time to respond.
4157 The per-instance connection queue will inflate, and the response time will
4158 increase respective to the size of the queue times the average per-session
4159 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01004160 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004161 the queue, and slowing down other users, and the servers as well, because the
4162 request will eventually be served, then aborted at the first error
4163 encountered while delivering the response.
4164
4165 As there is no way to distinguish between a full STOP and a simple output
4166 close on the client side, HTTP agents should be conservative and consider
4167 that the client might only have closed its output channel while waiting for
4168 the response. However, this introduces risks of congestion when lots of users
4169 do the same, and is completely useless nowadays because probably no client at
4170 all will close the session while waiting for the response. Some HTTP agents
Willy Tarreaud72758d2010-01-12 10:42:19 +01004171 support this behaviour (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004172 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01004173 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004174 of being the single component to break rare but valid traffic is extremely
4175 low, which adds to the temptation to be able to abort a session early while
4176 still not served and not pollute the servers.
4177
4178 In HAProxy, the user can choose the desired behaviour using the option
4179 "abortonclose". By default (without the option) the behaviour is HTTP
4180 compliant and aborted requests will be served. But when the option is
4181 specified, a session with an incoming channel closed will be aborted while
4182 it is still possible, either pending in the queue for a connection slot, or
4183 during the connection establishment if the server has not yet acknowledged
4184 the connection request. This considerably reduces the queue size and the load
4185 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01004186 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004187
4188 If this option has been enabled in a "defaults" section, it can be disabled
4189 in a specific instance by prepending the "no" keyword before it.
4190
4191 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
4192
4193
Willy Tarreau4076a152009-04-02 15:18:36 +02004194option accept-invalid-http-request
4195no option accept-invalid-http-request
4196 Enable or disable relaxing of HTTP request parsing
4197 May be used in sections : defaults | frontend | listen | backend
4198 yes | yes | yes | no
4199 Arguments : none
4200
Willy Tarreau91852eb2015-05-01 13:26:00 +02004201 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02004202 means that invalid characters in header names are not permitted and cause an
4203 error to be returned to the client. This is the desired behaviour as such
4204 forbidden characters are essentially used to build attacks exploiting server
4205 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
4206 server will emit invalid header names for whatever reason (configuration,
4207 implementation) and the issue will not be immediately fixed. In such a case,
4208 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01004209 even if that does not make sense, by specifying this option. Similarly, the
4210 list of characters allowed to appear in a URI is well defined by RFC3986, and
4211 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
4212 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
4213 not allowed at all. Haproxy always blocks a number of them (0..32, 127). The
Willy Tarreau91852eb2015-05-01 13:26:00 +02004214 remaining ones are blocked by default unless this option is enabled. This
Willy Tarreau13317662015-05-01 13:47:08 +02004215 option also relaxes the test on the HTTP version, it allows HTTP/0.9 requests
4216 to pass through (no version specified) and multiple digits for both the major
4217 and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02004218
4219 This option should never be enabled by default as it hides application bugs
4220 and open security breaches. It should only be deployed after a problem has
4221 been confirmed.
4222
4223 When this option is enabled, erroneous header names will still be accepted in
4224 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01004225 analysis using the "show errors" request on the UNIX stats socket. Similarly,
4226 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02004227 also helps confirming that the issue has been solved.
4228
4229 If this option has been enabled in a "defaults" section, it can be disabled
4230 in a specific instance by prepending the "no" keyword before it.
4231
4232 See also : "option accept-invalid-http-response" and "show errors" on the
4233 stats socket.
4234
4235
4236option accept-invalid-http-response
4237no option accept-invalid-http-response
4238 Enable or disable relaxing of HTTP response parsing
4239 May be used in sections : defaults | frontend | listen | backend
4240 yes | no | yes | yes
4241 Arguments : none
4242
Willy Tarreau91852eb2015-05-01 13:26:00 +02004243 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02004244 means that invalid characters in header names are not permitted and cause an
4245 error to be returned to the client. This is the desired behaviour as such
4246 forbidden characters are essentially used to build attacks exploiting server
4247 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
4248 server will emit invalid header names for whatever reason (configuration,
4249 implementation) and the issue will not be immediately fixed. In such a case,
4250 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau91852eb2015-05-01 13:26:00 +02004251 even if that does not make sense, by specifying this option. This option also
4252 relaxes the test on the HTTP version format, it allows multiple digits for
4253 both the major and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02004254
4255 This option should never be enabled by default as it hides application bugs
4256 and open security breaches. It should only be deployed after a problem has
4257 been confirmed.
4258
4259 When this option is enabled, erroneous header names will still be accepted in
4260 responses, but the complete response will be captured in order to permit
4261 later analysis using the "show errors" request on the UNIX stats socket.
4262 Doing this also helps confirming that the issue has been solved.
4263
4264 If this option has been enabled in a "defaults" section, it can be disabled
4265 in a specific instance by prepending the "no" keyword before it.
4266
4267 See also : "option accept-invalid-http-request" and "show errors" on the
4268 stats socket.
4269
4270
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004271option allbackups
4272no option allbackups
4273 Use either all backup servers at a time or only the first one
4274 May be used in sections : defaults | frontend | listen | backend
4275 yes | no | yes | yes
4276 Arguments : none
4277
4278 By default, the first operational backup server gets all traffic when normal
4279 servers are all down. Sometimes, it may be preferred to use multiple backups
4280 at once, because one will not be enough. When "option allbackups" is enabled,
4281 the load balancing will be performed among all backup servers when all normal
4282 ones are unavailable. The same load balancing algorithm will be used and the
4283 servers' weights will be respected. Thus, there will not be any priority
4284 order between the backup servers anymore.
4285
4286 This option is mostly used with static server farms dedicated to return a
4287 "sorry" page when an application is completely offline.
4288
4289 If this option has been enabled in a "defaults" section, it can be disabled
4290 in a specific instance by prepending the "no" keyword before it.
4291
4292
4293option checkcache
4294no option checkcache
Godbach7056a352013-12-11 20:01:07 +08004295 Analyze all server responses and block responses with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004296 May be used in sections : defaults | frontend | listen | backend
4297 yes | no | yes | yes
4298 Arguments : none
4299
4300 Some high-level frameworks set application cookies everywhere and do not
4301 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004302 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004303 high risk of session crossing or stealing between users traversing the same
4304 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02004305 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004306
4307 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004308 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01004309 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004310 response to check if there's a risk of caching a cookie on a client-side
4311 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01004312 to the client are :
4313 - all those without "Set-Cookie" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004314 - all those with a return code other than 200, 203, 206, 300, 301, 410,
Willy Tarreau198a7442008-01-17 12:05:32 +01004315 provided that the server has not set a "Cache-control: public" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004316 - all those that come from a POST request, provided that the server has not
4317 set a 'Cache-Control: public' header ;
4318 - those with a 'Pragma: no-cache' header
4319 - those with a 'Cache-control: private' header
4320 - those with a 'Cache-control: no-store' header
4321 - those with a 'Cache-control: max-age=0' header
4322 - those with a 'Cache-control: s-maxage=0' header
4323 - those with a 'Cache-control: no-cache' header
4324 - those with a 'Cache-control: no-cache="set-cookie"' header
4325 - those with a 'Cache-control: no-cache="set-cookie,' header
4326 (allowing other fields after set-cookie)
4327
4328 If a response doesn't respect these requirements, then it will be blocked
Willy Tarreau198a7442008-01-17 12:05:32 +01004329 just as if it was from an "rspdeny" filter, with an "HTTP 502 bad gateway".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004330 The session state shows "PH--" meaning that the proxy blocked the response
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004331 during headers processing. Additionally, an alert will be sent in the logs so
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004332 that admins are informed that there's something to be fixed.
4333
4334 Due to the high impact on the application, the application should be tested
4335 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01004336 good practice to always activate it during tests, even if it is not used in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004337 production, as it will report potentially dangerous application behaviours.
4338
4339 If this option has been enabled in a "defaults" section, it can be disabled
4340 in a specific instance by prepending the "no" keyword before it.
4341
4342
4343option clitcpka
4344no option clitcpka
4345 Enable or disable the sending of TCP keepalive packets on the client side
4346 May be used in sections : defaults | frontend | listen | backend
4347 yes | yes | yes | no
4348 Arguments : none
4349
4350 When there is a firewall or any session-aware component between a client and
4351 a server, and when the protocol involves very long sessions with long idle
4352 periods (eg: remote desktops), there is a risk that one of the intermediate
4353 components decides to expire a session which has remained idle for too long.
4354
4355 Enabling socket-level TCP keep-alives makes the system regularly send packets
4356 to the other end of the connection, leaving it active. The delay between
4357 keep-alive probes is controlled by the system only and depends both on the
4358 operating system and its tuning parameters.
4359
4360 It is important to understand that keep-alive packets are neither emitted nor
4361 received at the application level. It is only the network stacks which sees
4362 them. For this reason, even if one side of the proxy already uses keep-alives
4363 to maintain its connection alive, those keep-alive packets will not be
4364 forwarded to the other side of the proxy.
4365
4366 Please note that this has nothing to do with HTTP keep-alive.
4367
4368 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
4369 client side of a connection, which should help when session expirations are
4370 noticed between HAProxy and a client.
4371
4372 If this option has been enabled in a "defaults" section, it can be disabled
4373 in a specific instance by prepending the "no" keyword before it.
4374
4375 See also : "option srvtcpka", "option tcpka"
4376
4377
Willy Tarreau0ba27502007-12-24 16:55:16 +01004378option contstats
4379 Enable continuous traffic statistics updates
4380 May be used in sections : defaults | frontend | listen | backend
4381 yes | yes | yes | no
4382 Arguments : none
4383
4384 By default, counters used for statistics calculation are incremented
4385 only when a session finishes. It works quite well when serving small
4386 objects, but with big ones (for example large images or archives) or
4387 with A/V streaming, a graph generated from haproxy counters looks like
4388 a hedgehog. With this option enabled counters get incremented continuously,
4389 during a whole session. Recounting touches a hotpath directly so
4390 it is not enabled by default, as it has small performance impact (~0.5%).
4391
4392
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02004393option dontlog-normal
4394no option dontlog-normal
4395 Enable or disable logging of normal, successful connections
4396 May be used in sections : defaults | frontend | listen | backend
4397 yes | yes | yes | no
4398 Arguments : none
4399
4400 There are large sites dealing with several thousand connections per second
4401 and for which logging is a major pain. Some of them are even forced to turn
4402 logs off and cannot debug production issues. Setting this option ensures that
4403 normal connections, those which experience no error, no timeout, no retry nor
4404 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
4405 mode, the response status code is checked and return codes 5xx will still be
4406 logged.
4407
4408 It is strongly discouraged to use this option as most of the time, the key to
4409 complex issues is in the normal logs which will not be logged here. If you
4410 need to separate logs, see the "log-separate-errors" option instead.
4411
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004412 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02004413 logging.
4414
4415
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004416option dontlognull
4417no option dontlognull
4418 Enable or disable logging of null connections
4419 May be used in sections : defaults | frontend | listen | backend
4420 yes | yes | yes | no
4421 Arguments : none
4422
4423 In certain environments, there are components which will regularly connect to
4424 various systems to ensure that they are still alive. It can be the case from
4425 another load balancer as well as from monitoring systems. By default, even a
4426 simple port probe or scan will produce a log. If those connections pollute
4427 the logs too much, it is possible to enable option "dontlognull" to indicate
4428 that a connection on which no data has been transferred will not be logged,
Willy Tarreau0f228a02015-05-01 15:37:53 +02004429 which typically corresponds to those probes. Note that errors will still be
4430 returned to the client and accounted for in the stats. If this is not what is
4431 desired, option http-ignore-probes can be used instead.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004432
4433 It is generally recommended not to use this option in uncontrolled
4434 environments (eg: internet), otherwise scans and other malicious activities
4435 would not be logged.
4436
4437 If this option has been enabled in a "defaults" section, it can be disabled
4438 in a specific instance by prepending the "no" keyword before it.
4439
Willy Tarreau0f228a02015-05-01 15:37:53 +02004440 See also : "log", "http-ignore-probes", "monitor-net", "monitor-uri", and
4441 section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004442
4443
4444option forceclose
4445no option forceclose
4446 Enable or disable active connection closing after response is transferred.
4447 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaua31e5df2009-12-30 01:10:35 +01004448 yes | yes | yes | yes
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004449 Arguments : none
4450
4451 Some HTTP servers do not necessarily close the connections when they receive
4452 the "Connection: close" set by "option httpclose", and if the client does not
4453 close either, then the connection remains open till the timeout expires. This
4454 causes high number of simultaneous connections on the servers and shows high
4455 global session times in the logs.
4456
4457 When this happens, it is possible to use "option forceclose". It will
Willy Tarreau82eeaf22009-12-29 12:09:05 +01004458 actively close the outgoing server channel as soon as the server has finished
Cyril Bonté653dcd62014-02-20 00:13:15 +01004459 to respond and release some resources earlier than with "option httpclose".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004460
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004461 This option may also be combined with "option http-pretend-keepalive", which
4462 will disable sending of the "Connection: close" header, but will still cause
4463 the connection to be closed once the whole response is received.
4464
Cyril Bonté653dcd62014-02-20 00:13:15 +01004465 This option disables and replaces any previous "option httpclose", "option
4466 http-server-close", "option http-keep-alive", or "option http-tunnel".
Willy Tarreau02bce8b2014-01-30 00:15:28 +01004467
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004468 If this option has been enabled in a "defaults" section, it can be disabled
4469 in a specific instance by prepending the "no" keyword before it.
4470
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004471 See also : "option httpclose" and "option http-pretend-keepalive"
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004472
4473
Willy Tarreau87cf5142011-08-19 22:57:24 +02004474option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01004475 Enable insertion of the X-Forwarded-For header to requests sent to servers
4476 May be used in sections : defaults | frontend | listen | backend
4477 yes | yes | yes | yes
4478 Arguments :
4479 <network> is an optional argument used to disable this option for sources
4480 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02004481 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01004482 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004483
4484 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
4485 their client address. This is sometimes annoying when the client's IP address
4486 is expected in server logs. To solve this problem, the well-known HTTP header
4487 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
4488 This header contains a value representing the client's IP address. Since this
4489 header is always appended at the end of the existing header list, the server
4490 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02004491 the server's manual to find how to enable use of this standard header. Note
4492 that only the last occurrence of the header must be used, since it is really
4493 possible that the client has already brought one.
4494
Willy Tarreaud72758d2010-01-12 10:42:19 +01004495 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02004496 the default "X-Forwarded-For". This can be useful where you might already
Willy Tarreaud72758d2010-01-12 10:42:19 +01004497 have a "X-Forwarded-For" header from a different application (eg: stunnel),
4498 and you need preserve it. Also if your backend server doesn't use the
Ross Westaf72a1d2008-08-03 10:51:45 +02004499 "X-Forwarded-For" header and requires different one (eg: Zeus Web Servers
4500 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01004501
4502 Sometimes, a same HAProxy instance may be shared between a direct client
4503 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
4504 used to decrypt HTTPS traffic). It is possible to disable the addition of the
4505 header for a known source address or network by adding the "except" keyword
4506 followed by the network address. In this case, any source IP matching the
4507 network will not cause an addition of this header. Most common uses are with
4508 private networks or 127.0.0.1.
4509
Willy Tarreau87cf5142011-08-19 22:57:24 +02004510 Alternatively, the keyword "if-none" states that the header will only be
4511 added if it is not present. This should only be used in perfectly trusted
4512 environment, as this might cause a security issue if headers reaching haproxy
4513 are under the control of the end-user.
4514
Willy Tarreauc27debf2008-01-06 08:57:02 +01004515 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02004516 least one of them uses it, the header will be added. Note that the backend's
4517 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02004518 both are defined. In the case of the "if-none" argument, if at least one of
4519 the frontend or the backend does not specify it, it wants the addition to be
4520 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004521
Ross Westaf72a1d2008-08-03 10:51:45 +02004522 Examples :
Willy Tarreauc27debf2008-01-06 08:57:02 +01004523 # Public HTTP address also used by stunnel on the same machine
4524 frontend www
4525 mode http
4526 option forwardfor except 127.0.0.1 # stunnel already adds the header
4527
Ross Westaf72a1d2008-08-03 10:51:45 +02004528 # Those servers want the IP Address in X-Client
4529 backend www
4530 mode http
4531 option forwardfor header X-Client
4532
Willy Tarreau87cf5142011-08-19 22:57:24 +02004533 See also : "option httpclose", "option http-server-close",
Willy Tarreau70dffda2014-01-30 03:07:23 +01004534 "option forceclose", "option http-keep-alive"
Willy Tarreauc27debf2008-01-06 08:57:02 +01004535
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004536
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02004537option http-buffer-request
4538no option http-buffer-request
4539 Enable or disable waiting for whole HTTP request body before proceeding
4540 May be used in sections : defaults | frontend | listen | backend
4541 yes | yes | yes | yes
4542 Arguments : none
4543
4544 It is sometimes desirable to wait for the body of an HTTP request before
4545 taking a decision. This is what is being done by "balance url_param" for
4546 example. The first use case is to buffer requests from slow clients before
4547 connecting to the server. Another use case consists in taking the routing
4548 decision based on the request body's contents. This option placed in a
4549 frontend or backend forces the HTTP processing to wait until either the whole
4550 body is received, or the request buffer is full, or the first chunk is
4551 complete in case of chunked encoding. It can have undesired side effects with
4552 some applications abusing HTTP by expecting unbufferred transmissions between
4553 the frontend and the backend, so this should definitely not be used by
4554 default.
4555
4556 See also : "option http-no-delay"
4557
4558
Willy Tarreau0f228a02015-05-01 15:37:53 +02004559option http-ignore-probes
4560no option http-ignore-probes
4561 Enable or disable logging of null connections and request timeouts
4562 May be used in sections : defaults | frontend | listen | backend
4563 yes | yes | yes | no
4564 Arguments : none
4565
4566 Recently some browsers started to implement a "pre-connect" feature
4567 consisting in speculatively connecting to some recently visited web sites
4568 just in case the user would like to visit them. This results in many
4569 connections being established to web sites, which end up in 408 Request
4570 Timeout if the timeout strikes first, or 400 Bad Request when the browser
4571 decides to close them first. These ones pollute the log and feed the error
4572 counters. There was already "option dontlognull" but it's insufficient in
4573 this case. Instead, this option does the following things :
4574 - prevent any 400/408 message from being sent to the client if nothing
4575 was received over a connection before it was closed ;
4576 - prevent any log from being emitted in this situation ;
4577 - prevent any error counter from being incremented
4578
4579 That way the empty connection is silently ignored. Note that it is better
4580 not to use this unless it is clear that it is needed, because it will hide
4581 real problems. The most common reason for not receiving a request and seeing
4582 a 408 is due to an MTU inconsistency between the client and an intermediary
4583 element such as a VPN, which blocks too large packets. These issues are
4584 generally seen with POST requests as well as GET with large cookies. The logs
4585 are often the only way to detect them.
4586
4587 If this option has been enabled in a "defaults" section, it can be disabled
4588 in a specific instance by prepending the "no" keyword before it.
4589
4590 See also : "log", "dontlognull", "errorfile", and section 8 about logging.
4591
4592
Willy Tarreau16bfb022010-01-16 19:48:41 +01004593option http-keep-alive
4594no option http-keep-alive
4595 Enable or disable HTTP keep-alive from client to server
4596 May be used in sections : defaults | frontend | listen | backend
4597 yes | yes | yes | yes
4598 Arguments : none
4599
Willy Tarreau70dffda2014-01-30 03:07:23 +01004600 By default HAProxy operates in keep-alive mode with regards to persistent
4601 connections: for each connection it processes each request and response, and
4602 leaves the connection idle on both sides between the end of a response and the
4603 start of a new request. This mode may be changed by several options such as
4604 "option http-server-close", "option forceclose", "option httpclose" or
4605 "option http-tunnel". This option allows to set back the keep-alive mode,
4606 which can be useful when another mode was used in a defaults section.
4607
4608 Setting "option http-keep-alive" enables HTTP keep-alive mode on the client-
4609 and server- sides. This provides the lowest latency on the client side (slow
Willy Tarreau16bfb022010-01-16 19:48:41 +01004610 network) and the fastest session reuse on the server side at the expense
4611 of maintaining idle connections to the servers. In general, it is possible
4612 with this option to achieve approximately twice the request rate that the
4613 "http-server-close" option achieves on small objects. There are mainly two
4614 situations where this option may be useful :
4615
4616 - when the server is non-HTTP compliant and authenticates the connection
4617 instead of requests (eg: NTLM authentication)
4618
4619 - when the cost of establishing the connection to the server is significant
4620 compared to the cost of retrieving the associated object from the server.
4621
4622 This last case can happen when the server is a fast static server of cache.
4623 In this case, the server will need to be properly tuned to support high enough
4624 connection counts because connections will last until the client sends another
4625 request.
4626
4627 If the client request has to go to another backend or another server due to
4628 content switching or the load balancing algorithm, the idle connection will
Willy Tarreau9420b122013-12-15 18:58:25 +01004629 immediately be closed and a new one re-opened. Option "prefer-last-server" is
4630 available to try optimize server selection so that if the server currently
4631 attached to an idle connection is usable, it will be used.
Willy Tarreau16bfb022010-01-16 19:48:41 +01004632
4633 In general it is preferred to use "option http-server-close" with application
4634 servers, and some static servers might benefit from "option http-keep-alive".
4635
4636 At the moment, logs will not indicate whether requests came from the same
4637 session or not. The accept date reported in the logs corresponds to the end
4638 of the previous request, and the request time corresponds to the time spent
4639 waiting for a new request. The keep-alive request time is still bound to the
4640 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
4641 not set.
4642
Cyril Bonté653dcd62014-02-20 00:13:15 +01004643 This option disables and replaces any previous "option httpclose", "option
4644 http-server-close", "option forceclose" or "option http-tunnel". When backend
Willy Tarreau70dffda2014-01-30 03:07:23 +01004645 and frontend options differ, all of these 4 options have precedence over
Cyril Bonté653dcd62014-02-20 00:13:15 +01004646 "option http-keep-alive".
Willy Tarreau16bfb022010-01-16 19:48:41 +01004647
4648 See also : "option forceclose", "option http-server-close",
Willy Tarreau9420b122013-12-15 18:58:25 +01004649 "option prefer-last-server", "option http-pretend-keepalive",
4650 "option httpclose", and "1.1. The HTTP transaction model".
Willy Tarreau16bfb022010-01-16 19:48:41 +01004651
4652
Willy Tarreau96e31212011-05-30 18:10:30 +02004653option http-no-delay
4654no option http-no-delay
4655 Instruct the system to favor low interactive delays over performance in HTTP
4656 May be used in sections : defaults | frontend | listen | backend
4657 yes | yes | yes | yes
4658 Arguments : none
4659
4660 In HTTP, each payload is unidirectional and has no notion of interactivity.
4661 Any agent is expected to queue data somewhat for a reasonably low delay.
4662 There are some very rare server-to-server applications that abuse the HTTP
4663 protocol and expect the payload phase to be highly interactive, with many
4664 interleaved data chunks in both directions within a single request. This is
4665 absolutely not supported by the HTTP specification and will not work across
4666 most proxies or servers. When such applications attempt to do this through
4667 haproxy, it works but they will experience high delays due to the network
4668 optimizations which favor performance by instructing the system to wait for
4669 enough data to be available in order to only send full packets. Typical
4670 delays are around 200 ms per round trip. Note that this only happens with
4671 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
4672 affected.
4673
4674 When "option http-no-delay" is present in either the frontend or the backend
4675 used by a connection, all such optimizations will be disabled in order to
4676 make the exchanges as fast as possible. Of course this offers no guarantee on
4677 the functionality, as it may break at any other place. But if it works via
4678 HAProxy, it will work as fast as possible. This option should never be used
4679 by default, and should never be used at all unless such a buggy application
4680 is discovered. The impact of using this option is an increase of bandwidth
4681 usage and CPU usage, which may significantly lower performance in high
4682 latency environments.
4683
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02004684 See also : "option http-buffer-request"
4685
Willy Tarreau96e31212011-05-30 18:10:30 +02004686
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004687option http-pretend-keepalive
4688no option http-pretend-keepalive
4689 Define whether haproxy will announce keepalive to the server or not
4690 May be used in sections : defaults | frontend | listen | backend
4691 yes | yes | yes | yes
4692 Arguments : none
4693
4694 When running with "option http-server-close" or "option forceclose", haproxy
4695 adds a "Connection: close" header to the request forwarded to the server.
4696 Unfortunately, when some servers see this header, they automatically refrain
4697 from using the chunked encoding for responses of unknown length, while this
4698 is totally unrelated. The immediate effect is that this prevents haproxy from
4699 maintaining the client connection alive. A second effect is that a client or
4700 a cache could receive an incomplete response without being aware of it, and
4701 consider the response complete.
4702
4703 By setting "option http-pretend-keepalive", haproxy will make the server
4704 believe it will keep the connection alive. The server will then not fall back
4705 to the abnormal undesired above. When haproxy gets the whole response, it
4706 will close the connection with the server just as it would do with the
4707 "forceclose" option. That way the client gets a normal response and the
4708 connection is correctly closed on the server side.
4709
4710 It is recommended not to enable this option by default, because most servers
4711 will more efficiently close the connection themselves after the last packet,
4712 and release its buffers slightly earlier. Also, the added packet on the
4713 network could slightly reduce the overall peak performance. However it is
4714 worth noting that when this option is enabled, haproxy will have slightly
4715 less work to do. So if haproxy is the bottleneck on the whole architecture,
4716 enabling this option might save a few CPU cycles.
4717
4718 This option may be set both in a frontend and in a backend. It is enabled if
4719 at least one of the frontend or backend holding a connection has it enabled.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004720 This option may be combined with "option httpclose", which will cause
Willy Tarreau22a95342010-09-29 14:31:41 +02004721 keepalive to be announced to the server and close to be announced to the
4722 client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004723
4724 If this option has been enabled in a "defaults" section, it can be disabled
4725 in a specific instance by prepending the "no" keyword before it.
4726
Willy Tarreau16bfb022010-01-16 19:48:41 +01004727 See also : "option forceclose", "option http-server-close", and
4728 "option http-keep-alive"
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004729
Willy Tarreauc27debf2008-01-06 08:57:02 +01004730
Willy Tarreaub608feb2010-01-02 22:47:18 +01004731option http-server-close
4732no option http-server-close
4733 Enable or disable HTTP connection closing on the server side
4734 May be used in sections : defaults | frontend | listen | backend
4735 yes | yes | yes | yes
4736 Arguments : none
4737
Willy Tarreau70dffda2014-01-30 03:07:23 +01004738 By default HAProxy operates in keep-alive mode with regards to persistent
4739 connections: for each connection it processes each request and response, and
4740 leaves the connection idle on both sides between the end of a response and
4741 the start of a new request. This mode may be changed by several options such
4742 as "option http-server-close", "option forceclose", "option httpclose" or
4743 "option http-tunnel". Setting "option http-server-close" enables HTTP
4744 connection-close mode on the server side while keeping the ability to support
4745 HTTP keep-alive and pipelining on the client side. This provides the lowest
4746 latency on the client side (slow network) and the fastest session reuse on
4747 the server side to save server resources, similarly to "option forceclose".
4748 It also permits non-keepalive capable servers to be served in keep-alive mode
4749 to the clients if they conform to the requirements of RFC2616. Please note
4750 that some servers do not always conform to those requirements when they see
4751 "Connection: close" in the request. The effect will be that keep-alive will
4752 never be used. A workaround consists in enabling "option
4753 http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01004754
4755 At the moment, logs will not indicate whether requests came from the same
4756 session or not. The accept date reported in the logs corresponds to the end
4757 of the previous request, and the request time corresponds to the time spent
4758 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01004759 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
4760 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01004761
4762 This option may be set both in a frontend and in a backend. It is enabled if
4763 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01004764 It disables and replaces any previous "option httpclose", "option forceclose",
4765 "option http-tunnel" or "option http-keep-alive". Please check section 4
Willy Tarreau70dffda2014-01-30 03:07:23 +01004766 ("Proxies") to see how this option combines with others when frontend and
4767 backend options differ.
Willy Tarreaub608feb2010-01-02 22:47:18 +01004768
4769 If this option has been enabled in a "defaults" section, it can be disabled
4770 in a specific instance by prepending the "no" keyword before it.
4771
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02004772 See also : "option forceclose", "option http-pretend-keepalive",
Willy Tarreau16bfb022010-01-16 19:48:41 +01004773 "option httpclose", "option http-keep-alive", and
4774 "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01004775
4776
Willy Tarreau02bce8b2014-01-30 00:15:28 +01004777option http-tunnel
4778no option http-tunnel
4779 Disable or enable HTTP connection processing after first transaction
4780 May be used in sections : defaults | frontend | listen | backend
4781 yes | yes | yes | yes
4782 Arguments : none
4783
Willy Tarreau70dffda2014-01-30 03:07:23 +01004784 By default HAProxy operates in keep-alive mode with regards to persistent
4785 connections: for each connection it processes each request and response, and
4786 leaves the connection idle on both sides between the end of a response and
4787 the start of a new request. This mode may be changed by several options such
4788 as "option http-server-close", "option forceclose", "option httpclose" or
4789 "option http-tunnel".
4790
4791 Option "http-tunnel" disables any HTTP processing past the first request and
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03004792 the first response. This is the mode which was used by default in versions
Willy Tarreau70dffda2014-01-30 03:07:23 +01004793 1.0 to 1.5-dev21. It is the mode with the lowest processing overhead, which
4794 is normally not needed anymore unless in very specific cases such as when
4795 using an in-house protocol that looks like HTTP but is not compatible, or
4796 just to log one request per client in order to reduce log size. Note that
4797 everything which works at the HTTP level, including header parsing/addition,
4798 cookie processing or content switching will only work for the first request
4799 and will be ignored after the first response.
Willy Tarreau02bce8b2014-01-30 00:15:28 +01004800
4801 If this option has been enabled in a "defaults" section, it can be disabled
4802 in a specific instance by prepending the "no" keyword before it.
4803
4804 See also : "option forceclose", "option http-server-close",
4805 "option httpclose", "option http-keep-alive", and
4806 "1.1. The HTTP transaction model".
4807
4808
Willy Tarreau88d349d2010-01-25 12:15:43 +01004809option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01004810no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01004811 Make use of non-standard Proxy-Connection header instead of Connection
4812 May be used in sections : defaults | frontend | listen | backend
4813 yes | yes | yes | no
4814 Arguments : none
4815
4816 While RFC2616 explicitly states that HTTP/1.1 agents must use the
4817 Connection header to indicate their wish of persistent or non-persistent
4818 connections, both browsers and proxies ignore this header for proxied
4819 connections and make use of the undocumented, non-standard Proxy-Connection
4820 header instead. The issue begins when trying to put a load balancer between
4821 browsers and such proxies, because there will be a difference between what
4822 haproxy understands and what the client and the proxy agree on.
4823
4824 By setting this option in a frontend, haproxy can automatically switch to use
4825 that non-standard header if it sees proxied requests. A proxied request is
4826 defined here as one where the URI begins with neither a '/' nor a '*'. The
4827 choice of header only affects requests passing through proxies making use of
4828 one of the "httpclose", "forceclose" and "http-server-close" options. Note
4829 that this option can only be specified in a frontend and will affect the
4830 request along its whole life.
4831
Willy Tarreau844a7e72010-01-31 21:46:18 +01004832 Also, when this option is set, a request which requires authentication will
4833 automatically switch to use proxy authentication headers if it is itself a
4834 proxied request. That makes it possible to check or enforce authentication in
4835 front of an existing proxy.
4836
Willy Tarreau88d349d2010-01-25 12:15:43 +01004837 This option should normally never be used, except in front of a proxy.
4838
4839 See also : "option httpclose", "option forceclose" and "option
4840 http-server-close".
4841
4842
Willy Tarreaud63335a2010-02-26 12:56:52 +01004843option httpchk
4844option httpchk <uri>
4845option httpchk <method> <uri>
4846option httpchk <method> <uri> <version>
4847 Enable HTTP protocol to check on the servers health
4848 May be used in sections : defaults | frontend | listen | backend
4849 yes | no | yes | yes
4850 Arguments :
4851 <method> is the optional HTTP method used with the requests. When not set,
4852 the "OPTIONS" method is used, as it generally requires low server
4853 processing and is easy to filter out from the logs. Any method
4854 may be used, though it is not recommended to invent non-standard
4855 ones.
4856
4857 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
4858 which is accessible by default on almost any server, but may be
4859 changed to any other URI. Query strings are permitted.
4860
4861 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
4862 but some servers might behave incorrectly in HTTP 1.0, so turning
4863 it to HTTP/1.1 may sometimes help. Note that the Host field is
4864 mandatory in HTTP/1.1, and as a trick, it is possible to pass it
4865 after "\r\n" following the version string.
4866
4867 By default, server health checks only consist in trying to establish a TCP
4868 connection. When "option httpchk" is specified, a complete HTTP request is
4869 sent once the TCP connection is established, and responses 2xx and 3xx are
4870 considered valid, while all other ones indicate a server failure, including
4871 the lack of any response.
4872
4873 The port and interval are specified in the server configuration.
4874
4875 This option does not necessarily require an HTTP backend, it also works with
4876 plain TCP backends. This is particularly useful to check simple scripts bound
4877 to some dedicated ports using the inetd daemon.
4878
4879 Examples :
4880 # Relay HTTPS traffic to Apache instance and check service availability
4881 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
4882 backend https_relay
4883 mode tcp
4884 option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www
4885 server apache1 192.168.1.1:443 check port 80
4886
Simon Hormanafc47ee2013-11-25 10:46:35 +09004887 See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",
4888 "option pgsql-check", "http-check" and the "check", "port" and
4889 "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01004890
4891
Willy Tarreauc27debf2008-01-06 08:57:02 +01004892option httpclose
4893no option httpclose
4894 Enable or disable passive HTTP connection closing
4895 May be used in sections : defaults | frontend | listen | backend
4896 yes | yes | yes | yes
4897 Arguments : none
4898
Willy Tarreau70dffda2014-01-30 03:07:23 +01004899 By default HAProxy operates in keep-alive mode with regards to persistent
4900 connections: for each connection it processes each request and response, and
4901 leaves the connection idle on both sides between the end of a response and
4902 the start of a new request. This mode may be changed by several options such
Cyril Bonté653dcd62014-02-20 00:13:15 +01004903 as "option http-server-close", "option forceclose", "option httpclose" or
Willy Tarreau70dffda2014-01-30 03:07:23 +01004904 "option http-tunnel".
4905
4906 If "option httpclose" is set, HAProxy will work in HTTP tunnel mode and check
4907 if a "Connection: close" header is already set in each direction, and will
4908 add one if missing. Each end should react to this by actively closing the TCP
4909 connection after each transfer, thus resulting in a switch to the HTTP close
4910 mode. Any "Connection" header different from "close" will also be removed.
4911 Note that this option is deprecated since what it does is very cheap but not
4912 reliable. Using "option http-server-close" or "option forceclose" is strongly
4913 recommended instead.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004914
4915 It seldom happens that some servers incorrectly ignore this header and do not
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004916 close the connection even though they reply "Connection: close". For this
Willy Tarreau0dfdf192010-01-05 11:33:11 +01004917 reason, they are not compatible with older HTTP 1.0 browsers. If this happens
4918 it is possible to use the "option forceclose" which actively closes the
4919 request connection once the server responds. Option "forceclose" also
4920 releases the server connection earlier because it does not have to wait for
4921 the client to acknowledge it.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004922
4923 This option may be set both in a frontend and in a backend. It is enabled if
4924 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01004925 It disables and replaces any previous "option http-server-close",
4926 "option forceclose", "option http-keep-alive" or "option http-tunnel". Please
Willy Tarreau70dffda2014-01-30 03:07:23 +01004927 check section 4 ("Proxies") to see how this option combines with others when
4928 frontend and backend options differ.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004929
4930 If this option has been enabled in a "defaults" section, it can be disabled
4931 in a specific instance by prepending the "no" keyword before it.
4932
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02004933 See also : "option forceclose", "option http-server-close" and
4934 "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01004935
4936
Emeric Brun3a058f32009-06-30 18:26:00 +02004937option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01004938 Enable logging of HTTP request, session state and timers
4939 May be used in sections : defaults | frontend | listen | backend
4940 yes | yes | yes | yes
Emeric Brun3a058f32009-06-30 18:26:00 +02004941 Arguments :
4942 clf if the "clf" argument is added, then the output format will be
4943 the CLF format instead of HAProxy's default HTTP format. You can
4944 use this when you need to feed HAProxy's logs through a specific
4945 log analyser which only support the CLF format and which is not
4946 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004947
4948 By default, the log output format is very poor, as it only contains the
4949 source and destination addresses, and the instance name. By specifying
4950 "option httplog", each log line turns into a much richer format including,
4951 but not limited to, the HTTP request, the connection timers, the session
4952 status, the connections numbers, the captured headers and cookies, the
4953 frontend, backend and server name, and of course the source address and
4954 ports.
4955
4956 This option may be set either in the frontend or the backend.
4957
PiBa-NLbd556bf2014-12-11 21:31:54 +01004958 Specifying only "option httplog" will automatically clear the 'clf' mode
4959 if it was set by default.
Emeric Brun3a058f32009-06-30 18:26:00 +02004960
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004961 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004962
Willy Tarreau55165fe2009-05-10 12:02:55 +02004963
4964option http_proxy
4965no option http_proxy
4966 Enable or disable plain HTTP proxy mode
4967 May be used in sections : defaults | frontend | listen | backend
4968 yes | yes | yes | yes
4969 Arguments : none
4970
4971 It sometimes happens that people need a pure HTTP proxy which understands
4972 basic proxy requests without caching nor any fancy feature. In this case,
4973 it may be worth setting up an HAProxy instance with the "option http_proxy"
4974 set. In this mode, no server is declared, and the connection is forwarded to
4975 the IP address and port found in the URL after the "http://" scheme.
4976
4977 No host address resolution is performed, so this only works when pure IP
4978 addresses are passed. Since this option's usage perimeter is rather limited,
4979 it will probably be used only by experts who know they need exactly it. Last,
4980 if the clients are susceptible of sending keep-alive requests, it will be
Cyril Bonté2409e682010-12-14 22:47:51 +01004981 needed to add "option httpclose" to ensure that all requests will correctly
Willy Tarreau55165fe2009-05-10 12:02:55 +02004982 be analyzed.
4983
4984 If this option has been enabled in a "defaults" section, it can be disabled
4985 in a specific instance by prepending the "no" keyword before it.
4986
4987 Example :
4988 # this backend understands HTTP proxy requests and forwards them directly.
4989 backend direct_forward
4990 option httpclose
4991 option http_proxy
4992
4993 See also : "option httpclose"
4994
Willy Tarreau211ad242009-10-03 21:45:07 +02004995
Jamie Gloudon801a0a32012-08-25 00:18:33 -04004996option independent-streams
4997no option independent-streams
4998 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02004999 May be used in sections : defaults | frontend | listen | backend
5000 yes | yes | yes | yes
5001 Arguments : none
5002
5003 By default, when data is sent over a socket, both the write timeout and the
5004 read timeout for that socket are refreshed, because we consider that there is
5005 activity on that socket, and we have no other means of guessing if we should
5006 receive data or not.
5007
5008 While this default behaviour is desirable for almost all applications, there
5009 exists a situation where it is desirable to disable it, and only refresh the
5010 read timeout if there are incoming data. This happens on sessions with large
5011 timeouts and low amounts of exchanged data such as telnet session. If the
5012 server suddenly disappears, the output data accumulates in the system's
5013 socket buffers, both timeouts are correctly refreshed, and there is no way
5014 to know the server does not receive them, so we don't timeout. However, when
5015 the underlying protocol always echoes sent data, it would be enough by itself
5016 to detect the issue using the read timeout. Note that this problem does not
5017 happen with more verbose protocols because data won't accumulate long in the
5018 socket buffers.
5019
5020 When this option is set on the frontend, it will disable read timeout updates
5021 on data sent to the client. There probably is little use of this case. When
5022 the option is set on the backend, it will disable read timeout updates on
5023 data sent to the server. Doing so will typically break large HTTP posts from
5024 slow lines, so use it with caution.
5025
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005026 Note: older versions used to call this setting "option independent-streams"
Jamie Gloudon801a0a32012-08-25 00:18:33 -04005027 with a spelling mistake. This spelling is still supported but
5028 deprecated.
5029
Willy Tarreauce887fd2012-05-12 12:50:00 +02005030 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02005031
5032
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02005033option ldap-check
5034 Use LDAPv3 health checks for server testing
5035 May be used in sections : defaults | frontend | listen | backend
5036 yes | no | yes | yes
5037 Arguments : none
5038
5039 It is possible to test that the server correctly talks LDAPv3 instead of just
5040 testing that it accepts the TCP connection. When this option is set, an
5041 LDAPv3 anonymous simple bind message is sent to the server, and the response
5042 is analyzed to find an LDAPv3 bind response message.
5043
5044 The server is considered valid only when the LDAP response contains success
5045 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
5046
5047 Logging of bind requests is server dependent see your documentation how to
5048 configure it.
5049
5050 Example :
5051 option ldap-check
5052
5053 See also : "option httpchk"
5054
5055
Simon Horman98637e52014-06-20 12:30:16 +09005056option external-check
5057 Use external processes for server health checks
5058 May be used in sections : defaults | frontend | listen | backend
5059 yes | no | yes | yes
5060
5061 It is possible to test the health of a server using an external command.
5062 This is achieved by running the executable set using "external-check
5063 command".
5064
5065 Requires the "external-check" global to be set.
5066
5067 See also : "external-check", "external-check command", "external-check path"
5068
5069
Willy Tarreau211ad242009-10-03 21:45:07 +02005070option log-health-checks
5071no option log-health-checks
Willy Tarreaubef1b322014-05-13 21:01:39 +02005072 Enable or disable logging of health checks status updates
Willy Tarreau211ad242009-10-03 21:45:07 +02005073 May be used in sections : defaults | frontend | listen | backend
5074 yes | no | yes | yes
5075 Arguments : none
5076
Willy Tarreaubef1b322014-05-13 21:01:39 +02005077 By default, failed health check are logged if server is UP and successful
5078 health checks are logged if server is DOWN, so the amount of additional
5079 information is limited.
Willy Tarreau211ad242009-10-03 21:45:07 +02005080
Willy Tarreaubef1b322014-05-13 21:01:39 +02005081 When this option is enabled, any change of the health check status or to
5082 the server's health will be logged, so that it becomes possible to know
5083 that a server was failing occasional checks before crashing, or exactly when
5084 it failed to respond a valid HTTP status, then when the port started to
5085 reject connections, then when the server stopped responding at all.
5086
5087 Note that status changes not caused by health checks (eg: enable/disable on
5088 the CLI) are intentionally not logged by this option.
Willy Tarreau211ad242009-10-03 21:45:07 +02005089
Willy Tarreaubef1b322014-05-13 21:01:39 +02005090 See also: "option httpchk", "option ldap-check", "option mysql-check",
5091 "option pgsql-check", "option redis-check", "option smtpchk",
5092 "option tcp-check", "log" and section 8 about logging.
Willy Tarreau211ad242009-10-03 21:45:07 +02005093
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005094
5095option log-separate-errors
5096no option log-separate-errors
5097 Change log level for non-completely successful connections
5098 May be used in sections : defaults | frontend | listen | backend
5099 yes | yes | yes | no
5100 Arguments : none
5101
5102 Sometimes looking for errors in logs is not easy. This option makes haproxy
5103 raise the level of logs containing potentially interesting information such
5104 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
5105 level changes from "info" to "err". This makes it possible to log them
5106 separately to a different file with most syslog daemons. Be careful not to
5107 remove them from the original file, otherwise you would lose ordering which
5108 provides very important information.
5109
5110 Using this option, large sites dealing with several thousand connections per
5111 second may log normal traffic to a rotating buffer and only archive smaller
5112 error logs.
5113
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005114 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005115 logging.
5116
Willy Tarreauc27debf2008-01-06 08:57:02 +01005117
5118option logasap
5119no option logasap
5120 Enable or disable early logging of HTTP requests
5121 May be used in sections : defaults | frontend | listen | backend
5122 yes | yes | yes | no
5123 Arguments : none
5124
5125 By default, HTTP requests are logged upon termination so that the total
5126 transfer time and the number of bytes appear in the logs. When large objects
5127 are being transferred, it may take a while before the request appears in the
5128 logs. Using "option logasap", the request gets logged as soon as the server
5129 sends the complete headers. The only missing information in the logs will be
5130 the total number of bytes which will indicate everything except the amount
5131 of data transferred, and the total time which will not take the transfer
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01005132 time into account. In such a situation, it's a good practice to capture the
Willy Tarreauc27debf2008-01-06 08:57:02 +01005133 "Content-Length" response header so that the logs at least indicate how many
5134 bytes are expected to be transferred.
5135
Willy Tarreaucc6c8912009-02-22 10:53:55 +01005136 Examples :
5137 listen http_proxy 0.0.0.0:80
5138 mode http
5139 option httplog
5140 option logasap
5141 log 192.168.2.200 local3
5142
5143 >>> Feb 6 12:14:14 localhost \
5144 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
5145 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
5146 "GET /image.iso HTTP/1.0"
5147
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005148 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01005149 logging.
5150
5151
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02005152option mysql-check [ user <username> [ post-41 ] ]
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005153 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005154 May be used in sections : defaults | frontend | listen | backend
5155 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005156 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005157 <username> This is the username which will be used when connecting to MySQL
5158 server.
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02005159 post-41 Send post v4.1 client compatible checks
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005160
5161 If you specify a username, the check consists of sending two MySQL packet,
5162 one Client Authentication packet, and one QUIT packet, to correctly close
5163 MySQL session. We then parse the MySQL Handshake Initialisation packet and/or
5164 Error packet. It is a basic but useful test which does not produce error nor
5165 aborted connect on the server. However, it requires adding an authorization
5166 in the MySQL table, like this :
5167
5168 USE mysql;
5169 INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');
5170 FLUSH PRIVILEGES;
5171
5172 If you don't specify a username (it is deprecated and not recommended), the
5173 check only consists in parsing the Mysql Handshake Initialisation packet or
5174 Error packet, we don't send anything in this mode. It was reported that it
5175 can generate lockout if check is too frequent and/or if there is not enough
5176 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
5177 value as if a connection is established successfully within fewer than MySQL
5178 "max_connect_errors" attempts after a previous connection was interrupted,
5179 the error count for the host is cleared to zero. If HAProxy's server get
5180 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
5181
5182 Remember that this does not check database presence nor database consistency.
5183 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005184
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02005185 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005186
5187 Most often, an incoming MySQL server needs to see the client's IP address for
5188 various purposes, including IP privilege matching and connection logging.
5189 When possible, it is often wise to masquerade the client's IP address when
5190 connecting to the server using the "usesrc" argument of the "source" keyword,
5191 which requires the cttproxy feature to be compiled in, and the MySQL server
5192 to route the client via the machine hosting haproxy.
5193
5194 See also: "option httpchk"
5195
5196
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005197option nolinger
5198no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005199 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005200 May be used in sections: defaults | frontend | listen | backend
5201 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005202 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005203
5204 When clients or servers abort connections in a dirty way (eg: they are
5205 physically disconnected), the session timeouts triggers and the session is
5206 closed. But it will remain in FIN_WAIT1 state for some time in the system,
5207 using some resources and possibly limiting the ability to establish newer
5208 connections.
5209
5210 When this happens, it is possible to activate "option nolinger" which forces
5211 the system to immediately remove any socket's pending data on close. Thus,
5212 the session is instantly purged from the system's tables. This usually has
5213 side effects such as increased number of TCP resets due to old retransmits
5214 getting immediately rejected. Some firewalls may sometimes complain about
5215 this too.
5216
5217 For this reason, it is not recommended to use this option when not absolutely
5218 needed. You know that you need it when you have thousands of FIN_WAIT1
5219 sessions on your system (TIME_WAIT ones do not count).
5220
5221 This option may be used both on frontends and backends, depending on the side
5222 where it is required. Use it on the frontend for clients, and on the backend
5223 for servers.
5224
5225 If this option has been enabled in a "defaults" section, it can be disabled
5226 in a specific instance by prepending the "no" keyword before it.
5227
5228
Willy Tarreau55165fe2009-05-10 12:02:55 +02005229option originalto [ except <network> ] [ header <name> ]
5230 Enable insertion of the X-Original-To header to requests sent to servers
5231 May be used in sections : defaults | frontend | listen | backend
5232 yes | yes | yes | yes
5233 Arguments :
5234 <network> is an optional argument used to disable this option for sources
5235 matching <network>
5236 <name> an optional argument to specify a different "X-Original-To"
5237 header name.
5238
5239 Since HAProxy can work in transparent mode, every request from a client can
5240 be redirected to the proxy and HAProxy itself can proxy every request to a
5241 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
5242 be lost. This is annoying when you want access rules based on destination ip
5243 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
5244 added by HAProxy to all requests sent to the server. This header contains a
5245 value representing the original destination IP address. Since this must be
5246 configured to always use the last occurrence of this header only. Note that
5247 only the last occurrence of the header must be used, since it is really
5248 possible that the client has already brought one.
5249
5250 The keyword "header" may be used to supply a different header name to replace
5251 the default "X-Original-To". This can be useful where you might already
5252 have a "X-Original-To" header from a different application, and you need
5253 preserve it. Also if your backend server doesn't use the "X-Original-To"
5254 header and requires different one.
5255
5256 Sometimes, a same HAProxy instance may be shared between a direct client
5257 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
5258 used to decrypt HTTPS traffic). It is possible to disable the addition of the
5259 header for a known source address or network by adding the "except" keyword
5260 followed by the network address. In this case, any source IP matching the
5261 network will not cause an addition of this header. Most common uses are with
5262 private networks or 127.0.0.1.
5263
5264 This option may be specified either in the frontend or in the backend. If at
5265 least one of them uses it, the header will be added. Note that the backend's
5266 setting of the header subargument takes precedence over the frontend's if
5267 both are defined.
5268
Willy Tarreau55165fe2009-05-10 12:02:55 +02005269 Examples :
5270 # Original Destination address
5271 frontend www
5272 mode http
5273 option originalto except 127.0.0.1
5274
5275 # Those servers want the IP Address in X-Client-Dst
5276 backend www
5277 mode http
5278 option originalto header X-Client-Dst
5279
Willy Tarreau87cf5142011-08-19 22:57:24 +02005280 See also : "option httpclose", "option http-server-close",
5281 "option forceclose"
Willy Tarreau55165fe2009-05-10 12:02:55 +02005282
5283
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005284option persist
5285no option persist
5286 Enable or disable forced persistence on down servers
5287 May be used in sections: defaults | frontend | listen | backend
5288 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005289 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005290
5291 When an HTTP request reaches a backend with a cookie which references a dead
5292 server, by default it is redispatched to another server. It is possible to
5293 force the request to be sent to the dead server first using "option persist"
5294 if absolutely needed. A common use case is when servers are under extreme
5295 load and spend their time flapping. In this case, the users would still be
5296 directed to the server they opened the session on, in the hope they would be
5297 correctly served. It is recommended to use "option redispatch" in conjunction
5298 with this option so that in the event it would not be possible to connect to
5299 the server at all (server definitely dead), the client would finally be
5300 redirected to another valid server.
5301
5302 If this option has been enabled in a "defaults" section, it can be disabled
5303 in a specific instance by prepending the "no" keyword before it.
5304
Willy Tarreau4de91492010-01-22 19:10:05 +01005305 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005306
5307
Willy Tarreau0c122822013-12-15 18:49:01 +01005308option pgsql-check [ user <username> ]
5309 Use PostgreSQL health checks for server testing
5310 May be used in sections : defaults | frontend | listen | backend
5311 yes | no | yes | yes
5312 Arguments :
5313 <username> This is the username which will be used when connecting to
5314 PostgreSQL server.
5315
5316 The check sends a PostgreSQL StartupMessage and waits for either
5317 Authentication request or ErrorResponse message. It is a basic but useful
5318 test which does not produce error nor aborted connect on the server.
5319 This check is identical with the "mysql-check".
5320
5321 See also: "option httpchk"
5322
5323
Willy Tarreau9420b122013-12-15 18:58:25 +01005324option prefer-last-server
5325no option prefer-last-server
5326 Allow multiple load balanced requests to remain on the same server
5327 May be used in sections: defaults | frontend | listen | backend
5328 yes | no | yes | yes
5329 Arguments : none
5330
5331 When the load balancing algorithm in use is not deterministic, and a previous
5332 request was sent to a server to which haproxy still holds a connection, it is
5333 sometimes desirable that subsequent requests on a same session go to the same
5334 server as much as possible. Note that this is different from persistence, as
5335 we only indicate a preference which haproxy tries to apply without any form
5336 of warranty. The real use is for keep-alive connections sent to servers. When
5337 this option is used, haproxy will try to reuse the same connection that is
5338 attached to the server instead of rebalancing to another server, causing a
5339 close of the connection. This can make sense for static file servers. It does
Willy Tarreau068621e2013-12-23 15:11:25 +01005340 not make much sense to use this in combination with hashing algorithms. Note,
5341 haproxy already automatically tries to stick to a server which sends a 401 or
5342 to a proxy which sends a 407 (authentication required). This is mandatory for
5343 use with the broken NTLM authentication challenge, and significantly helps in
5344 troubleshooting some faulty applications. Option prefer-last-server might be
5345 desirable in these environments as well, to avoid redistributing the traffic
5346 after every other response.
Willy Tarreau9420b122013-12-15 18:58:25 +01005347
5348 If this option has been enabled in a "defaults" section, it can be disabled
5349 in a specific instance by prepending the "no" keyword before it.
5350
5351 See also: "option http-keep-alive"
5352
5353
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005354option redispatch
Joseph Lynch726ab712015-05-11 23:25:34 -07005355option redispatch <interval>
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005356no option redispatch
5357 Enable or disable session redistribution in case of connection failure
5358 May be used in sections: defaults | frontend | listen | backend
5359 yes | no | yes | yes
Joseph Lynch726ab712015-05-11 23:25:34 -07005360 Arguments :
5361 <interval> The optional integer value that controls how often redispatches
5362 occur when retrying connections. Positive value P indicates a
5363 redispatch is desired on every Pth retry, and negative value
5364 N indicate a redispath is desired on the Nth retry prior to the
5365 last retry. For example, the default of -1 preserves the
5366 historical behaviour of redispatching on the last retry, a
5367 positive value of 1 would indicate a redispatch on every retry,
5368 and a positive value of 3 would indicate a redispatch on every
5369 third retry. You can disable redispatches with a value of 0.
5370
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005371
5372 In HTTP mode, if a server designated by a cookie is down, clients may
5373 definitely stick to it because they cannot flush the cookie, so they will not
5374 be able to access the service anymore.
5375
5376 Specifying "option redispatch" will allow the proxy to break their
5377 persistence and redistribute them to a working server.
5378
Joseph Lynch726ab712015-05-11 23:25:34 -07005379 It also allows to retry connections to another server in case of multiple
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005380 connection failures. Of course, it requires having "retries" set to a nonzero
5381 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01005382
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005383 This form is the preferred form, which replaces both the "redispatch" and
5384 "redisp" keywords.
5385
5386 If this option has been enabled in a "defaults" section, it can be disabled
5387 in a specific instance by prepending the "no" keyword before it.
5388
Willy Tarreau4de91492010-01-22 19:10:05 +01005389 See also : "redispatch", "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005390
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005391
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02005392option redis-check
5393 Use redis health checks for server testing
5394 May be used in sections : defaults | frontend | listen | backend
5395 yes | no | yes | yes
5396 Arguments : none
5397
5398 It is possible to test that the server correctly talks REDIS protocol instead
5399 of just testing that it accepts the TCP connection. When this option is set,
5400 a PING redis command is sent to the server, and the response is analyzed to
5401 find the "+PONG" response message.
5402
5403 Example :
5404 option redis-check
5405
5406 See also : "option httpchk"
5407
5408
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005409option smtpchk
5410option smtpchk <hello> <domain>
5411 Use SMTP health checks for server testing
5412 May be used in sections : defaults | frontend | listen | backend
5413 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01005414 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005415 <hello> is an optional argument. It is the "hello" command to use. It can
5416 be either "HELO" (for SMTP) or "EHLO" (for ESTMP). All other
5417 values will be turned into the default command ("HELO").
5418
5419 <domain> is the domain name to present to the server. It may only be
5420 specified (and is mandatory) if the hello command has been
5421 specified. By default, "localhost" is used.
5422
5423 When "option smtpchk" is set, the health checks will consist in TCP
5424 connections followed by an SMTP command. By default, this command is
5425 "HELO localhost". The server's return code is analyzed and only return codes
5426 starting with a "2" will be considered as valid. All other responses,
5427 including a lack of response will constitute an error and will indicate a
5428 dead server.
5429
5430 This test is meant to be used with SMTP servers or relays. Depending on the
5431 request, it is possible that some servers do not log each connection attempt,
5432 so you may want to experiment to improve the behaviour. Using telnet on port
5433 25 is often easier than adjusting the configuration.
5434
5435 Most often, an incoming SMTP server needs to see the client's IP address for
5436 various purposes, including spam filtering, anti-spoofing and logging. When
5437 possible, it is often wise to masquerade the client's IP address when
5438 connecting to the server using the "usesrc" argument of the "source" keyword,
5439 which requires the cttproxy feature to be compiled in.
5440
5441 Example :
5442 option smtpchk HELO mydomain.org
5443
5444 See also : "option httpchk", "source"
5445
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005446
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02005447option socket-stats
5448no option socket-stats
5449
5450 Enable or disable collecting & providing separate statistics for each socket.
5451 May be used in sections : defaults | frontend | listen | backend
5452 yes | yes | yes | no
5453
5454 Arguments : none
5455
5456
Willy Tarreauff4f82d2009-02-06 11:28:13 +01005457option splice-auto
5458no option splice-auto
5459 Enable or disable automatic kernel acceleration on sockets in both directions
5460 May be used in sections : defaults | frontend | listen | backend
5461 yes | yes | yes | yes
5462 Arguments : none
5463
5464 When this option is enabled either on a frontend or on a backend, haproxy
5465 will automatically evaluate the opportunity to use kernel tcp splicing to
5466 forward data between the client and the server, in either direction. Haproxy
5467 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005468 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01005469 are not much aggressive in order to limit excessive use of splicing. This
5470 option requires splicing to be enabled at compile time, and may be globally
5471 disabled with the global option "nosplice". Since splice uses pipes, using it
5472 requires that there are enough spare pipes.
5473
5474 Important note: kernel-based TCP splicing is a Linux-specific feature which
5475 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
5476 transfer data between sockets without copying these data to user-space, thus
5477 providing noticeable performance gains and CPU cycles savings. Since many
5478 early implementations are buggy, corrupt data and/or are inefficient, this
5479 feature is not enabled by default, and it should be used with extreme care.
5480 While it is not possible to detect the correctness of an implementation,
5481 2.6.29 is the first version offering a properly working implementation. In
5482 case of doubt, splicing may be globally disabled using the global "nosplice"
5483 keyword.
5484
5485 Example :
5486 option splice-auto
5487
5488 If this option has been enabled in a "defaults" section, it can be disabled
5489 in a specific instance by prepending the "no" keyword before it.
5490
5491 See also : "option splice-request", "option splice-response", and global
5492 options "nosplice" and "maxpipes"
5493
5494
5495option splice-request
5496no option splice-request
5497 Enable or disable automatic kernel acceleration on sockets for requests
5498 May be used in sections : defaults | frontend | listen | backend
5499 yes | yes | yes | yes
5500 Arguments : none
5501
5502 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005503 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01005504 the client to the server. It might still use the recv/send scheme if there
5505 are no spare pipes left. This option requires splicing to be enabled at
5506 compile time, and may be globally disabled with the global option "nosplice".
5507 Since splice uses pipes, using it requires that there are enough spare pipes.
5508
5509 Important note: see "option splice-auto" for usage limitations.
5510
5511 Example :
5512 option splice-request
5513
5514 If this option has been enabled in a "defaults" section, it can be disabled
5515 in a specific instance by prepending the "no" keyword before it.
5516
5517 See also : "option splice-auto", "option splice-response", and global options
5518 "nosplice" and "maxpipes"
5519
5520
5521option splice-response
5522no option splice-response
5523 Enable or disable automatic kernel acceleration on sockets for responses
5524 May be used in sections : defaults | frontend | listen | backend
5525 yes | yes | yes | yes
5526 Arguments : none
5527
5528 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005529 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01005530 the server to the client. It might still use the recv/send scheme if there
5531 are no spare pipes left. This option requires splicing to be enabled at
5532 compile time, and may be globally disabled with the global option "nosplice".
5533 Since splice uses pipes, using it requires that there are enough spare pipes.
5534
5535 Important note: see "option splice-auto" for usage limitations.
5536
5537 Example :
5538 option splice-response
5539
5540 If this option has been enabled in a "defaults" section, it can be disabled
5541 in a specific instance by prepending the "no" keyword before it.
5542
5543 See also : "option splice-auto", "option splice-request", and global options
5544 "nosplice" and "maxpipes"
5545
5546
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005547option srvtcpka
5548no option srvtcpka
5549 Enable or disable the sending of TCP keepalive packets on the server side
5550 May be used in sections : defaults | frontend | listen | backend
5551 yes | no | yes | yes
5552 Arguments : none
5553
5554 When there is a firewall or any session-aware component between a client and
5555 a server, and when the protocol involves very long sessions with long idle
5556 periods (eg: remote desktops), there is a risk that one of the intermediate
5557 components decides to expire a session which has remained idle for too long.
5558
5559 Enabling socket-level TCP keep-alives makes the system regularly send packets
5560 to the other end of the connection, leaving it active. The delay between
5561 keep-alive probes is controlled by the system only and depends both on the
5562 operating system and its tuning parameters.
5563
5564 It is important to understand that keep-alive packets are neither emitted nor
5565 received at the application level. It is only the network stacks which sees
5566 them. For this reason, even if one side of the proxy already uses keep-alives
5567 to maintain its connection alive, those keep-alive packets will not be
5568 forwarded to the other side of the proxy.
5569
5570 Please note that this has nothing to do with HTTP keep-alive.
5571
5572 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
5573 server side of a connection, which should help when session expirations are
5574 noticed between HAProxy and a server.
5575
5576 If this option has been enabled in a "defaults" section, it can be disabled
5577 in a specific instance by prepending the "no" keyword before it.
5578
5579 See also : "option clitcpka", "option tcpka"
5580
5581
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005582option ssl-hello-chk
5583 Use SSLv3 client hello health checks for server testing
5584 May be used in sections : defaults | frontend | listen | backend
5585 yes | no | yes | yes
5586 Arguments : none
5587
5588 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
5589 possible to test that the server correctly talks SSL instead of just testing
5590 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
5591 SSLv3 client hello messages are sent once the connection is established to
5592 the server, and the response is analyzed to find an SSL server hello message.
5593 The server is considered valid only when the response contains this server
5594 hello message.
5595
5596 All servers tested till there correctly reply to SSLv3 client hello messages,
5597 and most servers tested do not even log the requests containing only hello
5598 messages, which is appreciable.
5599
Willy Tarreau763a95b2012-10-04 23:15:39 +02005600 Note that this check works even when SSL support was not built into haproxy
5601 because it forges the SSL message. When SSL support is available, it is best
5602 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005603
Willy Tarreau763a95b2012-10-04 23:15:39 +02005604 See also: "option httpchk", "check-ssl"
5605
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005606
Willy Tarreaued179852013-12-16 01:07:00 +01005607option tcp-check
5608 Perform health checks using tcp-check send/expect sequences
5609 May be used in sections: defaults | frontend | listen | backend
5610 yes | no | yes | yes
5611
5612 This health check method is intended to be combined with "tcp-check" command
5613 lists in order to support send/expect types of health check sequences.
5614
5615 TCP checks currently support 4 modes of operations :
5616 - no "tcp-check" directive : the health check only consists in a connection
5617 attempt, which remains the default mode.
5618
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005619 - "tcp-check send" or "tcp-check send-binary" only is mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01005620 used to send a string along with a connection opening. With some
5621 protocols, it helps sending a "QUIT" message for example that prevents
5622 the server from logging a connection error for each health check. The
5623 check result will still be based on the ability to open the connection
5624 only.
5625
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005626 - "tcp-check expect" only is mentioned : this is used to test a banner.
Willy Tarreaued179852013-12-16 01:07:00 +01005627 The connection is opened and haproxy waits for the server to present some
5628 contents which must validate some rules. The check result will be based
5629 on the matching between the contents and the rules. This is suited for
5630 POP, IMAP, SMTP, FTP, SSH, TELNET.
5631
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005632 - both "tcp-check send" and "tcp-check expect" are mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01005633 used to test a hello-type protocol. Haproxy sends a message, the server
5634 responds and its response is analysed. the check result will be based on
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005635 the matching between the response contents and the rules. This is often
Willy Tarreaued179852013-12-16 01:07:00 +01005636 suited for protocols which require a binding or a request/response model.
5637 LDAP, MySQL, Redis and SSL are example of such protocols, though they
5638 already all have their dedicated checks with a deeper understanding of
5639 the respective protocols.
5640 In this mode, many questions may be sent and many answers may be
5641 analysed.
5642
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02005643 A fifth mode can be used to insert comments in different steps of the
5644 script.
5645
5646 For each tcp-check rule you create, you can add a "comment" directive,
5647 followed by a string. This string will be reported in the log and stderr
5648 in debug mode. It is useful to make user-friendly error reporting.
5649 The "comment" is of course optional.
5650
5651
Willy Tarreaued179852013-12-16 01:07:00 +01005652 Examples :
5653 # perform a POP check (analyse only server's banner)
5654 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02005655 tcp-check expect string +OK\ POP3\ ready comment POP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01005656
5657 # perform an IMAP check (analyse only server's banner)
5658 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02005659 tcp-check expect string *\ OK\ IMAP4\ ready comment IMAP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01005660
5661 # look for the redis master server after ensuring it speaks well
5662 # redis protocol, then it exits properly.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005663 # (send a command then analyse the response 3 times)
Willy Tarreaued179852013-12-16 01:07:00 +01005664 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02005665 tcp-check comment PING\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01005666 tcp-check send PING\r\n
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02005667 tcp-check expect +PONGe
5668 tcp-check comment role\ check
Willy Tarreaued179852013-12-16 01:07:00 +01005669 tcp-check send info\ replication\r\n
5670 tcp-check expect string role:master
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02005671 tcp-check comment QUIT\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01005672 tcp-check send QUIT\r\n
5673 tcp-check expect string +OK
5674
5675 forge a HTTP request, then analyse the response
5676 (send many headers before analyzing)
5677 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02005678 tcp-check comment forge\ and\ send\ HTTP\ request
Willy Tarreaued179852013-12-16 01:07:00 +01005679 tcp-check send HEAD\ /\ HTTP/1.1\r\n
5680 tcp-check send Host:\ www.mydomain.com\r\n
5681 tcp-check send User-Agent:\ HAProxy\ tcpcheck\r\n
5682 tcp-check send \r\n
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02005683 tcp-check expect rstring HTTP/1\..\ (2..|3..) comment check\ HTTP\ response
Willy Tarreaued179852013-12-16 01:07:00 +01005684
5685
5686 See also : "tcp-check expect", "tcp-check send"
5687
5688
Willy Tarreau9ea05a72009-06-14 12:07:01 +02005689option tcp-smart-accept
5690no option tcp-smart-accept
5691 Enable or disable the saving of one ACK packet during the accept sequence
5692 May be used in sections : defaults | frontend | listen | backend
5693 yes | yes | yes | no
5694 Arguments : none
5695
5696 When an HTTP connection request comes in, the system acknowledges it on
5697 behalf of HAProxy, then the client immediately sends its request, and the
5698 system acknowledges it too while it is notifying HAProxy about the new
5699 connection. HAProxy then reads the request and responds. This means that we
5700 have one TCP ACK sent by the system for nothing, because the request could
5701 very well be acknowledged by HAProxy when it sends its response.
5702
5703 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
5704 sending this useless ACK on platforms which support it (currently at least
5705 Linux). It must not cause any problem, because the system will send it anyway
5706 after 40 ms if the response takes more time than expected to come.
5707
5708 During complex network debugging sessions, it may be desirable to disable
5709 this optimization because delayed ACKs can make troubleshooting more complex
5710 when trying to identify where packets are delayed. It is then possible to
5711 fall back to normal behaviour by specifying "no option tcp-smart-accept".
5712
5713 It is also possible to force it for non-HTTP proxies by simply specifying
5714 "option tcp-smart-accept". For instance, it can make sense with some services
5715 such as SMTP where the server speaks first.
5716
5717 It is recommended to avoid forcing this option in a defaults section. In case
5718 of doubt, consider setting it back to automatic values by prepending the
5719 "default" keyword before it, or disabling it using the "no" keyword.
5720
Willy Tarreaud88edf22009-06-14 15:48:17 +02005721 See also : "option tcp-smart-connect"
5722
5723
5724option tcp-smart-connect
5725no option tcp-smart-connect
5726 Enable or disable the saving of one ACK packet during the connect sequence
5727 May be used in sections : defaults | frontend | listen | backend
5728 yes | no | yes | yes
5729 Arguments : none
5730
5731 On certain systems (at least Linux), HAProxy can ask the kernel not to
5732 immediately send an empty ACK upon a connection request, but to directly
5733 send the buffer request instead. This saves one packet on the network and
5734 thus boosts performance. It can also be useful for some servers, because they
5735 immediately get the request along with the incoming connection.
5736
5737 This feature is enabled when "option tcp-smart-connect" is set in a backend.
5738 It is not enabled by default because it makes network troubleshooting more
5739 complex.
5740
5741 It only makes sense to enable it with protocols where the client speaks first
5742 such as HTTP. In other situations, if there is no data to send in place of
5743 the ACK, a normal ACK is sent.
5744
5745 If this option has been enabled in a "defaults" section, it can be disabled
5746 in a specific instance by prepending the "no" keyword before it.
5747
5748 See also : "option tcp-smart-accept"
5749
Willy Tarreau9ea05a72009-06-14 12:07:01 +02005750
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005751option tcpka
5752 Enable or disable the sending of TCP keepalive packets on both sides
5753 May be used in sections : defaults | frontend | listen | backend
5754 yes | yes | yes | yes
5755 Arguments : none
5756
5757 When there is a firewall or any session-aware component between a client and
5758 a server, and when the protocol involves very long sessions with long idle
5759 periods (eg: remote desktops), there is a risk that one of the intermediate
5760 components decides to expire a session which has remained idle for too long.
5761
5762 Enabling socket-level TCP keep-alives makes the system regularly send packets
5763 to the other end of the connection, leaving it active. The delay between
5764 keep-alive probes is controlled by the system only and depends both on the
5765 operating system and its tuning parameters.
5766
5767 It is important to understand that keep-alive packets are neither emitted nor
5768 received at the application level. It is only the network stacks which sees
5769 them. For this reason, even if one side of the proxy already uses keep-alives
5770 to maintain its connection alive, those keep-alive packets will not be
5771 forwarded to the other side of the proxy.
5772
5773 Please note that this has nothing to do with HTTP keep-alive.
5774
5775 Using option "tcpka" enables the emission of TCP keep-alive probes on both
5776 the client and server sides of a connection. Note that this is meaningful
5777 only in "defaults" or "listen" sections. If this option is used in a
5778 frontend, only the client side will get keep-alives, and if this option is
5779 used in a backend, only the server side will get keep-alives. For this
5780 reason, it is strongly recommended to explicitly use "option clitcpka" and
5781 "option srvtcpka" when the configuration is split between frontends and
5782 backends.
5783
5784 See also : "option clitcpka", "option srvtcpka"
5785
Willy Tarreau844e3c52008-01-11 16:28:18 +01005786
5787option tcplog
5788 Enable advanced logging of TCP connections with session state and timers
5789 May be used in sections : defaults | frontend | listen | backend
5790 yes | yes | yes | yes
5791 Arguments : none
5792
5793 By default, the log output format is very poor, as it only contains the
5794 source and destination addresses, and the instance name. By specifying
5795 "option tcplog", each log line turns into a much richer format including, but
5796 not limited to, the connection timers, the session status, the connections
5797 numbers, the frontend, backend and server name, and of course the source
5798 address and ports. This option is useful for pure TCP proxies in order to
5799 find which of the client or server disconnects or times out. For normal HTTP
5800 proxies, it's better to use "option httplog" which is even more complete.
5801
5802 This option may be set either in the frontend or the backend.
5803
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005804 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01005805
5806
Willy Tarreau844e3c52008-01-11 16:28:18 +01005807option transparent
5808no option transparent
5809 Enable client-side transparent proxying
5810 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01005811 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01005812 Arguments : none
5813
5814 This option was introduced in order to provide layer 7 persistence to layer 3
5815 load balancers. The idea is to use the OS's ability to redirect an incoming
5816 connection for a remote address to a local process (here HAProxy), and let
5817 this process know what address was initially requested. When this option is
5818 used, sessions without cookies will be forwarded to the original destination
5819 IP address of the incoming request (which should match that of another
5820 equipment), while requests with cookies will still be forwarded to the
5821 appropriate server.
5822
5823 Note that contrary to a common belief, this option does NOT make HAProxy
5824 present the client's IP to the server when establishing the connection.
5825
Willy Tarreaua1146052011-03-01 09:51:54 +01005826 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005827 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01005828
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005829
Simon Horman98637e52014-06-20 12:30:16 +09005830external-check command <command>
5831 Executable to run when performing an external-check
5832 May be used in sections : defaults | frontend | listen | backend
5833 yes | no | yes | yes
5834
5835 Arguments :
5836 <command> is the external command to run
5837
Simon Horman98637e52014-06-20 12:30:16 +09005838 The arguments passed to the to the command are:
5839
Cyril Bonté777be862014-12-02 21:21:35 +01005840 <proxy_address> <proxy_port> <server_address> <server_port>
Simon Horman98637e52014-06-20 12:30:16 +09005841
Cyril Bonté777be862014-12-02 21:21:35 +01005842 The <proxy_address> and <proxy_port> are derived from the first listener
5843 that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
5844 listener the proxy_address will be the path of the socket and the
5845 <proxy_port> will be the string "NOT_USED". In a backend section, it's not
5846 possible to determine a listener, and both <proxy_address> and <proxy_port>
5847 will have the string value "NOT_USED".
Simon Horman98637e52014-06-20 12:30:16 +09005848
Cyril Bonté72cda2a2014-12-27 22:28:39 +01005849 Some values are also provided through environment variables.
5850
5851 Environment variables :
5852 HAPROXY_PROXY_ADDR The first bind address if available (or empty if not
5853 applicable, for example in a "backend" section).
5854
5855 HAPROXY_PROXY_ID The backend id.
5856
5857 HAPROXY_PROXY_NAME The backend name.
5858
5859 HAPROXY_PROXY_PORT The first bind port if available (or empty if not
5860 applicable, for example in a "backend" section or
5861 for a UNIX socket).
5862
5863 HAPROXY_SERVER_ADDR The server address.
5864
5865 HAPROXY_SERVER_CURCONN The current number of connections on the server.
5866
5867 HAPROXY_SERVER_ID The server id.
5868
5869 HAPROXY_SERVER_MAXCONN The server max connections.
5870
5871 HAPROXY_SERVER_NAME The server name.
5872
5873 HAPROXY_SERVER_PORT The server port if available (or empty for a UNIX
5874 socket).
5875
5876 PATH The PATH environment variable used when executing
5877 the command may be set using "external-check path".
5878
Simon Horman98637e52014-06-20 12:30:16 +09005879 If the command executed and exits with a zero status then the check is
5880 considered to have passed, otherwise the check is considered to have
5881 failed.
5882
5883 Example :
5884 external-check command /bin/true
5885
5886 See also : "external-check", "option external-check", "external-check path"
5887
5888
5889external-check path <path>
5890 The value of the PATH environment variable used when running an external-check
5891 May be used in sections : defaults | frontend | listen | backend
5892 yes | no | yes | yes
5893
5894 Arguments :
5895 <path> is the path used when executing external command to run
5896
5897 The default path is "".
5898
5899 Example :
5900 external-check path "/usr/bin:/bin"
5901
5902 See also : "external-check", "option external-check",
5903 "external-check command"
5904
5905
Emeric Brun647caf12009-06-30 17:57:00 +02005906persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02005907persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02005908 Enable RDP cookie-based persistence
5909 May be used in sections : defaults | frontend | listen | backend
5910 yes | no | yes | yes
5911 Arguments :
5912 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02005913 default cookie name "msts" will be used. There currently is no
5914 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02005915
5916 This statement enables persistence based on an RDP cookie. The RDP cookie
5917 contains all information required to find the server in the list of known
5918 servers. So when this option is set in the backend, the request is analysed
5919 and if an RDP cookie is found, it is decoded. If it matches a known server
5920 which is still UP (or if "option persist" is set), then the connection is
5921 forwarded to this server.
5922
5923 Note that this only makes sense in a TCP backend, but for this to work, the
5924 frontend must have waited long enough to ensure that an RDP cookie is present
5925 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005926 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02005927 a single "listen" section.
5928
Willy Tarreau61e28f22010-05-16 22:31:05 +02005929 Also, it is important to understand that the terminal server will emit this
5930 RDP cookie only if it is configured for "token redirection mode", which means
5931 that the "IP address redirection" option is disabled.
5932
Emeric Brun647caf12009-06-30 17:57:00 +02005933 Example :
5934 listen tse-farm
5935 bind :3389
5936 # wait up to 5s for an RDP cookie in the request
5937 tcp-request inspect-delay 5s
5938 tcp-request content accept if RDP_COOKIE
5939 # apply RDP cookie persistence
5940 persist rdp-cookie
5941 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02005942 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02005943 balance rdp-cookie
5944 server srv1 1.1.1.1:3389
5945 server srv2 1.1.1.2:3389
5946
Simon Hormanab814e02011-06-24 14:50:20 +09005947 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
5948 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02005949
5950
Willy Tarreau3a7d2072009-03-05 23:48:25 +01005951rate-limit sessions <rate>
5952 Set a limit on the number of new sessions accepted per second on a frontend
5953 May be used in sections : defaults | frontend | listen | backend
5954 yes | yes | yes | no
5955 Arguments :
5956 <rate> The <rate> parameter is an integer designating the maximum number
5957 of new sessions per second to accept on the frontend.
5958
5959 When the frontend reaches the specified number of new sessions per second, it
5960 stops accepting new connections until the rate drops below the limit again.
5961 During this time, the pending sessions will be kept in the socket's backlog
5962 (in system buffers) and haproxy will not even be aware that sessions are
5963 pending. When applying very low limit on a highly loaded service, it may make
5964 sense to increase the socket's backlog using the "backlog" keyword.
5965
5966 This feature is particularly efficient at blocking connection-based attacks
5967 or service abuse on fragile servers. Since the session rate is measured every
5968 millisecond, it is extremely accurate. Also, the limit applies immediately,
5969 no delay is needed at all to detect the threshold.
5970
5971 Example : limit the connection rate on SMTP to 10 per second max
5972 listen smtp
5973 mode tcp
5974 bind :25
5975 rate-limit sessions 10
5976 server 127.0.0.1:1025
5977
Willy Tarreaua17c2d92011-07-25 08:16:20 +02005978 Note : when the maximum rate is reached, the frontend's status is not changed
5979 but its sockets appear as "WAITING" in the statistics if the
5980 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +01005981
5982 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
5983
5984
Willy Tarreau2e1dca82012-09-12 08:43:15 +02005985redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
5986redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
5987redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +02005988 Return an HTTP redirection if/unless a condition is matched
5989 May be used in sections : defaults | frontend | listen | backend
5990 no | yes | yes | yes
5991
5992 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +01005993 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02005994
Willy Tarreau0140f252008-11-19 21:07:09 +01005995 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +02005996 <loc> With "redirect location", the exact value in <loc> is placed into
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01005997 the HTTP "Location" header. When used in an "http-request" rule,
5998 <loc> value follows the log-format rules and can include some
5999 dynamic values (see Custom Log Format in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006000
6001 <pfx> With "redirect prefix", the "Location" header is built from the
6002 concatenation of <pfx> and the complete URI path, including the
6003 query string, unless the "drop-query" option is specified (see
6004 below). As a special case, if <pfx> equals exactly "/", then
6005 nothing is inserted before the original URI. It allows one to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006006 redirect to the same URL (for instance, to insert a cookie). When
6007 used in an "http-request" rule, <pfx> value follows the log-format
6008 rules and can include some dynamic values (see Custom Log Format
6009 in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006010
6011 <sch> With "redirect scheme", then the "Location" header is built by
6012 concatenating <sch> with "://" then the first occurrence of the
6013 "Host" header, and then the URI path, including the query string
6014 unless the "drop-query" option is specified (see below). If no
6015 path is found or if the path is "*", then "/" is used instead. If
6016 no "Host" header is found, then an empty host component will be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006017 returned, which most recent browsers interpret as redirecting to
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006018 the same host. This directive is mostly used to redirect HTTP to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006019 HTTPS. When used in an "http-request" rule, <sch> value follows
6020 the log-format rules and can include some dynamic values (see
6021 Custom Log Format in section 8.2.4).
Willy Tarreau0140f252008-11-19 21:07:09 +01006022
6023 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +01006024 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
6025 with 302 used by default if no code is specified. 301 means
6026 "Moved permanently", and a browser may cache the Location. 302
6027 means "Moved permanently" and means that the browser should not
6028 cache the redirection. 303 is equivalent to 302 except that the
6029 browser will fetch the location with a GET method. 307 is just
6030 like 302 but makes it clear that the same method must be reused.
6031 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +01006032
6033 <option> There are several options which can be specified to adjust the
6034 expected behaviour of a redirection :
6035
6036 - "drop-query"
6037 When this keyword is used in a prefix-based redirection, then the
6038 location will be set without any possible query-string, which is useful
6039 for directing users to a non-secure page for instance. It has no effect
6040 with a location-type redirect.
6041
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01006042 - "append-slash"
6043 This keyword may be used in conjunction with "drop-query" to redirect
6044 users who use a URL not ending with a '/' to the same one with the '/'.
6045 It can be useful to ensure that search engines will only see one URL.
6046 For this, a return code 301 is preferred.
6047
Willy Tarreau0140f252008-11-19 21:07:09 +01006048 - "set-cookie NAME[=value]"
6049 A "Set-Cookie" header will be added with NAME (and optionally "=value")
6050 to the response. This is sometimes used to indicate that a user has
6051 been seen, for instance to protect against some types of DoS. No other
6052 cookie option is added, so the cookie will be a session cookie. Note
6053 that for a browser, a sole cookie name without an equal sign is
6054 different from a cookie with an equal sign.
6055
6056 - "clear-cookie NAME[=]"
6057 A "Set-Cookie" header will be added with NAME (and optionally "="), but
6058 with the "Max-Age" attribute set to zero. This will tell the browser to
6059 delete this cookie. It is useful for instance on logout pages. It is
6060 important to note that clearing the cookie "NAME" will not remove a
6061 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
6062 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006063
6064 Example: move the login URL only to HTTPS.
6065 acl clear dst_port 80
6066 acl secure dst_port 8080
6067 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +01006068 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +01006069 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +01006070 acl cookie_set hdr_sub(cookie) SEEN=1
6071
6072 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +01006073 redirect prefix https://mysite.com if login_page !secure
6074 redirect prefix http://mysite.com drop-query if login_page !uid_given
6075 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +01006076 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006077
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01006078 Example: send redirects for request for articles without a '/'.
6079 acl missing_slash path_reg ^/article/[^/]*$
6080 redirect code 301 prefix / drop-query append-slash if missing_slash
6081
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006082 Example: redirect all HTTP traffic to HTTPS when SSL is handled by haproxy.
David BERARDe7153042012-11-03 00:11:31 +01006083 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006084
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006085 Example: append 'www.' prefix in front of all hosts not having it
6086 http-request redirect code 301 location www.%[hdr(host)]%[req.uri] \
6087 unless { hdr_beg(host) -i www }
6088
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006089 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006090
6091
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006092redisp (deprecated)
6093redispatch (deprecated)
6094 Enable or disable session redistribution in case of connection failure
6095 May be used in sections: defaults | frontend | listen | backend
6096 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006097 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006098
6099 In HTTP mode, if a server designated by a cookie is down, clients may
6100 definitely stick to it because they cannot flush the cookie, so they will not
6101 be able to access the service anymore.
6102
6103 Specifying "redispatch" will allow the proxy to break their persistence and
6104 redistribute them to a working server.
6105
6106 It also allows to retry last connection to another server in case of multiple
6107 connection failures. Of course, it requires having "retries" set to a nonzero
6108 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01006109
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006110 This form is deprecated, do not use it in any new configuration, use the new
6111 "option redispatch" instead.
6112
6113 See also : "option redispatch"
6114
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006115
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006116reqadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01006117 Add a header at the end of the HTTP request
6118 May be used in sections : defaults | frontend | listen | backend
6119 no | yes | yes | yes
6120 Arguments :
6121 <string> is the complete line to be added. Any space or known delimiter
6122 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006123 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006124
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006125 <cond> is an optional matching condition built from ACLs. It makes it
6126 possible to ignore this rule when other conditions are not met.
6127
Willy Tarreau303c0352008-01-17 19:01:39 +01006128 A new line consisting in <string> followed by a line feed will be added after
6129 the last header of an HTTP request.
6130
6131 Header transformations only apply to traffic which passes through HAProxy,
6132 and not to traffic generated by HAProxy, such as health-checks or error
6133 responses.
6134
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006135 Example : add "X-Proto: SSL" to requests coming via port 81
6136 acl is-ssl dst_port 81
6137 reqadd X-Proto:\ SSL if is-ssl
6138
6139 See also: "rspadd", section 6 about HTTP header manipulation, and section 7
6140 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006141
6142
Willy Tarreau5321c422010-01-28 20:35:13 +01006143reqallow <search> [{if | unless} <cond>]
6144reqiallow <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006145 Definitely allow an HTTP request if a line matches a regular expression
6146 May be used in sections : defaults | frontend | listen | backend
6147 no | yes | yes | yes
6148 Arguments :
6149 <search> is the regular expression applied to HTTP headers and to the
6150 request line. This is an extended regular expression. Parenthesis
6151 grouping is supported and no preliminary backslash is required.
6152 Any space or known delimiter must be escaped using a backslash
6153 ('\'). The pattern applies to a full line at a time. The
6154 "reqallow" keyword strictly matches case while "reqiallow"
6155 ignores case.
6156
Willy Tarreau5321c422010-01-28 20:35:13 +01006157 <cond> is an optional matching condition built from ACLs. It makes it
6158 possible to ignore this rule when other conditions are not met.
6159
Willy Tarreau303c0352008-01-17 19:01:39 +01006160 A request containing any line which matches extended regular expression
6161 <search> will mark the request as allowed, even if any later test would
6162 result in a deny. The test applies both to the request line and to request
6163 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01006164 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01006165
6166 It is easier, faster and more powerful to use ACLs to write access policies.
6167 Reqdeny, reqallow and reqpass should be avoided in new designs.
6168
6169 Example :
6170 # allow www.* but refuse *.local
6171 reqiallow ^Host:\ www\.
6172 reqideny ^Host:\ .*\.local
6173
Willy Tarreau5321c422010-01-28 20:35:13 +01006174 See also: "reqdeny", "block", section 6 about HTTP header manipulation, and
6175 section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006176
6177
Willy Tarreau5321c422010-01-28 20:35:13 +01006178reqdel <search> [{if | unless} <cond>]
6179reqidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006180 Delete all headers matching a regular expression in an HTTP request
6181 May be used in sections : defaults | frontend | listen | backend
6182 no | yes | yes | yes
6183 Arguments :
6184 <search> is the regular expression applied to HTTP headers and to the
6185 request line. This is an extended regular expression. Parenthesis
6186 grouping is supported and no preliminary backslash is required.
6187 Any space or known delimiter must be escaped using a backslash
6188 ('\'). The pattern applies to a full line at a time. The "reqdel"
6189 keyword strictly matches case while "reqidel" ignores case.
6190
Willy Tarreau5321c422010-01-28 20:35:13 +01006191 <cond> is an optional matching condition built from ACLs. It makes it
6192 possible to ignore this rule when other conditions are not met.
6193
Willy Tarreau303c0352008-01-17 19:01:39 +01006194 Any header line matching extended regular expression <search> in the request
6195 will be completely deleted. Most common use of this is to remove unwanted
6196 and/or dangerous headers or cookies from a request before passing it to the
6197 next servers.
6198
6199 Header transformations only apply to traffic which passes through HAProxy,
6200 and not to traffic generated by HAProxy, such as health-checks or error
6201 responses. Keep in mind that header names are not case-sensitive.
6202
6203 Example :
6204 # remove X-Forwarded-For header and SERVER cookie
6205 reqidel ^X-Forwarded-For:.*
6206 reqidel ^Cookie:.*SERVER=
6207
Willy Tarreau5321c422010-01-28 20:35:13 +01006208 See also: "reqadd", "reqrep", "rspdel", section 6 about HTTP header
6209 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006210
6211
Willy Tarreau5321c422010-01-28 20:35:13 +01006212reqdeny <search> [{if | unless} <cond>]
6213reqideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006214 Deny an HTTP request if a line matches a regular expression
6215 May be used in sections : defaults | frontend | listen | backend
6216 no | yes | yes | yes
6217 Arguments :
6218 <search> is the regular expression applied to HTTP headers and to the
6219 request line. This is an extended regular expression. Parenthesis
6220 grouping is supported and no preliminary backslash is required.
6221 Any space or known delimiter must be escaped using a backslash
6222 ('\'). The pattern applies to a full line at a time. The
6223 "reqdeny" keyword strictly matches case while "reqideny" ignores
6224 case.
6225
Willy Tarreau5321c422010-01-28 20:35:13 +01006226 <cond> is an optional matching condition built from ACLs. It makes it
6227 possible to ignore this rule when other conditions are not met.
6228
Willy Tarreau303c0352008-01-17 19:01:39 +01006229 A request containing any line which matches extended regular expression
6230 <search> will mark the request as denied, even if any later test would
6231 result in an allow. The test applies both to the request line and to request
6232 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01006233 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01006234
Willy Tarreauced27012008-01-17 20:35:34 +01006235 A denied request will generate an "HTTP 403 forbidden" response once the
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006236 complete request has been parsed. This is consistent with what is practiced
Willy Tarreaud72758d2010-01-12 10:42:19 +01006237 using ACLs.
Willy Tarreauced27012008-01-17 20:35:34 +01006238
Willy Tarreau303c0352008-01-17 19:01:39 +01006239 It is easier, faster and more powerful to use ACLs to write access policies.
6240 Reqdeny, reqallow and reqpass should be avoided in new designs.
6241
6242 Example :
6243 # refuse *.local, then allow www.*
6244 reqideny ^Host:\ .*\.local
6245 reqiallow ^Host:\ www\.
6246
Willy Tarreau5321c422010-01-28 20:35:13 +01006247 See also: "reqallow", "rspdeny", "block", section 6 about HTTP header
6248 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006249
6250
Willy Tarreau5321c422010-01-28 20:35:13 +01006251reqpass <search> [{if | unless} <cond>]
6252reqipass <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006253 Ignore any HTTP request line matching a regular expression in next rules
6254 May be used in sections : defaults | frontend | listen | backend
6255 no | yes | yes | yes
6256 Arguments :
6257 <search> is the regular expression applied to HTTP headers and to the
6258 request line. This is an extended regular expression. Parenthesis
6259 grouping is supported and no preliminary backslash is required.
6260 Any space or known delimiter must be escaped using a backslash
6261 ('\'). The pattern applies to a full line at a time. The
6262 "reqpass" keyword strictly matches case while "reqipass" ignores
6263 case.
6264
Willy Tarreau5321c422010-01-28 20:35:13 +01006265 <cond> is an optional matching condition built from ACLs. It makes it
6266 possible to ignore this rule when other conditions are not met.
6267
Willy Tarreau303c0352008-01-17 19:01:39 +01006268 A request containing any line which matches extended regular expression
6269 <search> will skip next rules, without assigning any deny or allow verdict.
6270 The test applies both to the request line and to request headers. Keep in
6271 mind that URLs in request line are case-sensitive while header names are not.
6272
6273 It is easier, faster and more powerful to use ACLs to write access policies.
6274 Reqdeny, reqallow and reqpass should be avoided in new designs.
6275
6276 Example :
6277 # refuse *.local, then allow www.*, but ignore "www.private.local"
6278 reqipass ^Host:\ www.private\.local
6279 reqideny ^Host:\ .*\.local
6280 reqiallow ^Host:\ www\.
6281
Willy Tarreau5321c422010-01-28 20:35:13 +01006282 See also: "reqallow", "reqdeny", "block", section 6 about HTTP header
6283 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006284
6285
Willy Tarreau5321c422010-01-28 20:35:13 +01006286reqrep <search> <string> [{if | unless} <cond>]
6287reqirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006288 Replace a regular expression with a string in an HTTP request line
6289 May be used in sections : defaults | frontend | listen | backend
6290 no | yes | yes | yes
6291 Arguments :
6292 <search> is the regular expression applied to HTTP headers and to the
6293 request line. This is an extended regular expression. Parenthesis
6294 grouping is supported and no preliminary backslash is required.
6295 Any space or known delimiter must be escaped using a backslash
6296 ('\'). The pattern applies to a full line at a time. The "reqrep"
6297 keyword strictly matches case while "reqirep" ignores case.
6298
6299 <string> is the complete line to be added. Any space or known delimiter
6300 must be escaped using a backslash ('\'). References to matched
6301 pattern groups are possible using the common \N form, with N
6302 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006303 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006304
Willy Tarreau5321c422010-01-28 20:35:13 +01006305 <cond> is an optional matching condition built from ACLs. It makes it
6306 possible to ignore this rule when other conditions are not met.
6307
Willy Tarreau303c0352008-01-17 19:01:39 +01006308 Any line matching extended regular expression <search> in the request (both
6309 the request line and header lines) will be completely replaced with <string>.
6310 Most common use of this is to rewrite URLs or domain names in "Host" headers.
6311
6312 Header transformations only apply to traffic which passes through HAProxy,
6313 and not to traffic generated by HAProxy, such as health-checks or error
6314 responses. Note that for increased readability, it is suggested to add enough
6315 spaces between the request and the response. Keep in mind that URLs in
6316 request line are case-sensitive while header names are not.
6317
6318 Example :
6319 # replace "/static/" with "/" at the beginning of any request path.
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04006320 reqrep ^([^\ :]*)\ /static/(.*) \1\ /\2
Willy Tarreau303c0352008-01-17 19:01:39 +01006321 # replace "www.mydomain.com" with "www" in the host name.
6322 reqirep ^Host:\ www.mydomain.com Host:\ www
6323
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04006324 See also: "reqadd", "reqdel", "rsprep", "tune.bufsize", section 6 about
6325 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006326
6327
Willy Tarreau5321c422010-01-28 20:35:13 +01006328reqtarpit <search> [{if | unless} <cond>]
6329reqitarpit <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006330 Tarpit an HTTP request containing a line matching a regular expression
6331 May be used in sections : defaults | frontend | listen | backend
6332 no | yes | yes | yes
6333 Arguments :
6334 <search> is the regular expression applied to HTTP headers and to the
6335 request line. This is an extended regular expression. Parenthesis
6336 grouping is supported and no preliminary backslash is required.
6337 Any space or known delimiter must be escaped using a backslash
6338 ('\'). The pattern applies to a full line at a time. The
6339 "reqtarpit" keyword strictly matches case while "reqitarpit"
6340 ignores case.
6341
Willy Tarreau5321c422010-01-28 20:35:13 +01006342 <cond> is an optional matching condition built from ACLs. It makes it
6343 possible to ignore this rule when other conditions are not met.
6344
Willy Tarreau303c0352008-01-17 19:01:39 +01006345 A request containing any line which matches extended regular expression
6346 <search> will be tarpitted, which means that it will connect to nowhere, will
Willy Tarreauced27012008-01-17 20:35:34 +01006347 be kept open for a pre-defined time, then will return an HTTP error 500 so
6348 that the attacker does not suspect it has been tarpitted. The status 500 will
6349 be reported in the logs, but the completion flags will indicate "PT". The
Willy Tarreau303c0352008-01-17 19:01:39 +01006350 delay is defined by "timeout tarpit", or "timeout connect" if the former is
6351 not set.
6352
6353 The goal of the tarpit is to slow down robots attacking servers with
6354 identifiable requests. Many robots limit their outgoing number of connections
6355 and stay connected waiting for a reply which can take several minutes to
6356 come. Depending on the environment and attack, it may be particularly
6357 efficient at reducing the load on the network and firewalls.
6358
Willy Tarreau5321c422010-01-28 20:35:13 +01006359 Examples :
Willy Tarreau303c0352008-01-17 19:01:39 +01006360 # ignore user-agents reporting any flavour of "Mozilla" or "MSIE", but
6361 # block all others.
6362 reqipass ^User-Agent:\.*(Mozilla|MSIE)
6363 reqitarpit ^User-Agent:
6364
Willy Tarreau5321c422010-01-28 20:35:13 +01006365 # block bad guys
6366 acl badguys src 10.1.0.3 172.16.13.20/28
6367 reqitarpit . if badguys
6368
6369 See also: "reqallow", "reqdeny", "reqpass", section 6 about HTTP header
6370 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006371
6372
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02006373retries <value>
6374 Set the number of retries to perform on a server after a connection failure
6375 May be used in sections: defaults | frontend | listen | backend
6376 yes | no | yes | yes
6377 Arguments :
6378 <value> is the number of times a connection attempt should be retried on
6379 a server when a connection either is refused or times out. The
6380 default value is 3.
6381
6382 It is important to understand that this value applies to the number of
6383 connection attempts, not full requests. When a connection has effectively
6384 been established to a server, there will be no more retry.
6385
6386 In order to avoid immediate reconnections to a server which is restarting,
Joseph Lynch726ab712015-05-11 23:25:34 -07006387 a turn-around timer of min("timeout connect", one second) is applied before
6388 a retry occurs.
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02006389
6390 When "option redispatch" is set, the last retry may be performed on another
6391 server even if a cookie references a different server.
6392
6393 See also : "option redispatch"
6394
6395
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006396rspadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01006397 Add a header at the end of the HTTP response
6398 May be used in sections : defaults | frontend | listen | backend
6399 no | yes | yes | yes
6400 Arguments :
6401 <string> is the complete line to be added. Any space or known delimiter
6402 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006403 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006404
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006405 <cond> is an optional matching condition built from ACLs. It makes it
6406 possible to ignore this rule when other conditions are not met.
6407
Willy Tarreau303c0352008-01-17 19:01:39 +01006408 A new line consisting in <string> followed by a line feed will be added after
6409 the last header of an HTTP response.
6410
6411 Header transformations only apply to traffic which passes through HAProxy,
6412 and not to traffic generated by HAProxy, such as health-checks or error
6413 responses.
6414
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006415 See also: "reqadd", section 6 about HTTP header manipulation, and section 7
6416 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006417
6418
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006419rspdel <search> [{if | unless} <cond>]
6420rspidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006421 Delete all headers matching a regular expression in an HTTP response
6422 May be used in sections : defaults | frontend | listen | backend
6423 no | yes | yes | yes
6424 Arguments :
6425 <search> is the regular expression applied to HTTP headers and to the
6426 response line. This is an extended regular expression, so
6427 parenthesis grouping is supported and no preliminary backslash
6428 is required. Any space or known delimiter must be escaped using
6429 a backslash ('\'). The pattern applies to a full line at a time.
6430 The "rspdel" keyword strictly matches case while "rspidel"
6431 ignores case.
6432
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006433 <cond> is an optional matching condition built from ACLs. It makes it
6434 possible to ignore this rule when other conditions are not met.
6435
Willy Tarreau303c0352008-01-17 19:01:39 +01006436 Any header line matching extended regular expression <search> in the response
6437 will be completely deleted. Most common use of this is to remove unwanted
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02006438 and/or sensitive headers or cookies from a response before passing it to the
Willy Tarreau303c0352008-01-17 19:01:39 +01006439 client.
6440
6441 Header transformations only apply to traffic which passes through HAProxy,
6442 and not to traffic generated by HAProxy, such as health-checks or error
6443 responses. Keep in mind that header names are not case-sensitive.
6444
6445 Example :
6446 # remove the Server header from responses
Willy Tarreau5e80e022013-05-25 08:31:25 +02006447 rspidel ^Server:.*
Willy Tarreau303c0352008-01-17 19:01:39 +01006448
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006449 See also: "rspadd", "rsprep", "reqdel", section 6 about HTTP header
6450 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006451
6452
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006453rspdeny <search> [{if | unless} <cond>]
6454rspideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006455 Block an HTTP response if a line matches a regular expression
6456 May be used in sections : defaults | frontend | listen | backend
6457 no | yes | yes | yes
6458 Arguments :
6459 <search> is the regular expression applied to HTTP headers and to the
6460 response line. This is an extended regular expression, so
6461 parenthesis grouping is supported and no preliminary backslash
6462 is required. Any space or known delimiter must be escaped using
6463 a backslash ('\'). The pattern applies to a full line at a time.
6464 The "rspdeny" keyword strictly matches case while "rspideny"
6465 ignores case.
6466
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006467 <cond> is an optional matching condition built from ACLs. It makes it
6468 possible to ignore this rule when other conditions are not met.
6469
Willy Tarreau303c0352008-01-17 19:01:39 +01006470 A response containing any line which matches extended regular expression
6471 <search> will mark the request as denied. The test applies both to the
6472 response line and to response headers. Keep in mind that header names are not
6473 case-sensitive.
6474
6475 Main use of this keyword is to prevent sensitive information leak and to
Willy Tarreauced27012008-01-17 20:35:34 +01006476 block the response before it reaches the client. If a response is denied, it
6477 will be replaced with an HTTP 502 error so that the client never retrieves
6478 any sensitive data.
Willy Tarreau303c0352008-01-17 19:01:39 +01006479
6480 It is easier, faster and more powerful to use ACLs to write access policies.
6481 Rspdeny should be avoided in new designs.
6482
6483 Example :
6484 # Ensure that no content type matching ms-word will leak
6485 rspideny ^Content-type:\.*/ms-word
6486
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006487 See also: "reqdeny", "acl", "block", section 6 about HTTP header manipulation
6488 and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006489
6490
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006491rsprep <search> <string> [{if | unless} <cond>]
6492rspirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006493 Replace a regular expression with a string in an HTTP response line
6494 May be used in sections : defaults | frontend | listen | backend
6495 no | yes | yes | yes
6496 Arguments :
6497 <search> is the regular expression applied to HTTP headers and to the
6498 response line. This is an extended regular expression, so
6499 parenthesis grouping is supported and no preliminary backslash
6500 is required. Any space or known delimiter must be escaped using
6501 a backslash ('\'). The pattern applies to a full line at a time.
6502 The "rsprep" keyword strictly matches case while "rspirep"
6503 ignores case.
6504
6505 <string> is the complete line to be added. Any space or known delimiter
6506 must be escaped using a backslash ('\'). References to matched
6507 pattern groups are possible using the common \N form, with N
6508 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006509 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006510
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006511 <cond> is an optional matching condition built from ACLs. It makes it
6512 possible to ignore this rule when other conditions are not met.
6513
Willy Tarreau303c0352008-01-17 19:01:39 +01006514 Any line matching extended regular expression <search> in the response (both
6515 the response line and header lines) will be completely replaced with
6516 <string>. Most common use of this is to rewrite Location headers.
6517
6518 Header transformations only apply to traffic which passes through HAProxy,
6519 and not to traffic generated by HAProxy, such as health-checks or error
6520 responses. Note that for increased readability, it is suggested to add enough
6521 spaces between the request and the response. Keep in mind that header names
6522 are not case-sensitive.
6523
6524 Example :
6525 # replace "Location: 127.0.0.1:8080" with "Location: www.mydomain.com"
6526 rspirep ^Location:\ 127.0.0.1:8080 Location:\ www.mydomain.com
6527
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006528 See also: "rspadd", "rspdel", "reqrep", section 6 about HTTP header
6529 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006530
6531
David du Colombier486df472011-03-17 10:40:26 +01006532server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006533 Declare a server in a backend
6534 May be used in sections : defaults | frontend | listen | backend
6535 no | no | yes | yes
6536 Arguments :
6537 <name> is the internal name assigned to this server. This name will
Cyril Bonté941a0c62012-10-15 19:44:24 +02006538 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -05006539 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006540
David du Colombier486df472011-03-17 10:40:26 +01006541 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
6542 resolvable hostname is supported, but this name will be resolved
6543 during start-up. Address "0.0.0.0" or "*" has a special meaning.
6544 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +02006545 address as the one from the client connection. This is useful in
6546 transparent proxy architectures where the client's connection is
6547 intercepted and haproxy must forward to the original destination
6548 address. This is more or less what the "transparent" keyword does
6549 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +01006550 to report statistics. Optionally, an address family prefix may be
6551 used before the address to force the family regardless of the
6552 address format, which can be useful to specify a path to a unix
6553 socket with no slash ('/'). Currently supported prefixes are :
6554 - 'ipv4@' -> address is always IPv4
6555 - 'ipv6@' -> address is always IPv6
6556 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02006557 - 'abns@' -> address is in abstract namespace (Linux only)
William Lallemandb2f07452015-05-12 14:27:13 +02006558 You may want to reference some environment variables in the
6559 address parameter, see section 2.3 about environment
6560 variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006561
Willy Tarreaub6205fd2012-09-24 12:27:33 +02006562 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006563 be sent to this port. If unset, the same port the client
6564 connected to will be used. The port may also be prefixed by a "+"
6565 or a "-". In this case, the server's port will be determined by
6566 adding this value to the client's port.
6567
6568 <param*> is a list of parameters for this server. The "server" keywords
6569 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006570 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006571
6572 Examples :
6573 server first 10.1.1.1:1080 cookie first check inter 1000
6574 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +01006575 server transp ipv4@
William Lallemandb2f07452015-05-12 14:27:13 +02006576 server backup "${SRV_BACKUP}:1080" backup
6577 server www1_dc1 "${LAN_DC1}.101:80"
6578 server www1_dc2 "${LAN_DC2}.101:80"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006579
Mark Lamourinec2247f02012-01-04 13:02:01 -05006580 See also: "default-server", "http-send-name-header" and section 5 about
6581 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006582
6583
6584source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02006585source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +01006586source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006587 Set the source address for outgoing connections
6588 May be used in sections : defaults | frontend | listen | backend
6589 yes | no | yes | yes
6590 Arguments :
6591 <addr> is the IPv4 address HAProxy will bind to before connecting to a
6592 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +01006593
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006594 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +01006595 the most appropriate address to reach its destination. Optionally
6596 an address family prefix may be used before the address to force
6597 the family regardless of the address format, which can be useful
6598 to specify a path to a unix socket with no slash ('/'). Currently
6599 supported prefixes are :
6600 - 'ipv4@' -> address is always IPv4
6601 - 'ipv6@' -> address is always IPv6
6602 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02006603 - 'abns@' -> address is in abstract namespace (Linux only)
William Lallemandb2f07452015-05-12 14:27:13 +02006604 You may want to reference some environment variables in the address
6605 parameter, see section 2.3 about environment variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006606
6607 <port> is an optional port. It is normally not needed but may be useful
6608 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02006609 the system will select a free port. Note that port ranges are not
6610 supported in the backend. If you want to force port ranges, you
6611 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006612
6613 <addr2> is the IP address to present to the server when connections are
6614 forwarded in full transparent proxy mode. This is currently only
6615 supported on some patched Linux kernels. When this address is
6616 specified, clients connecting to the server will be presented
6617 with this address, while health checks will still use the address
6618 <addr>.
6619
6620 <port2> is the optional port to present to the server when connections
6621 are forwarded in full transparent proxy mode (see <addr2> above).
6622 The default value of zero means the system will select a free
6623 port.
6624
Willy Tarreaubce70882009-09-07 11:51:47 +02006625 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
6626 This is the name of a comma-separated header list which can
6627 contain multiple IP addresses. By default, the last occurrence is
6628 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +01006629 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +02006630 by previous proxy, typically Stunnel. In order to use another
6631 occurrence from the last one, please see the <occ> parameter
6632 below. When the header (or occurrence) is not found, no binding
6633 is performed so that the proxy's default IP address is used. Also
6634 keep in mind that the header name is case insensitive, as for any
6635 HTTP header.
6636
6637 <occ> is the occurrence number of a value to be used in a multi-value
6638 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04006639 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +02006640 address. Positive values indicate a position from the first
6641 occurrence, 1 being the first one. Negative values indicate
6642 positions relative to the last one, -1 being the last one. This
6643 is helpful for situations where an X-Forwarded-For header is set
6644 at the entry point of an infrastructure and must be used several
6645 proxy layers away. When this value is not specified, -1 is
6646 assumed. Passing a zero here disables the feature.
6647
Willy Tarreaud53f96b2009-02-04 18:46:54 +01006648 <name> is an optional interface name to which to bind to for outgoing
6649 traffic. On systems supporting this features (currently, only
6650 Linux), this allows one to bind all traffic to the server to
6651 this interface even if it is not the one the system would select
6652 based on routing tables. This should be used with extreme care.
6653 Note that using this option requires root privileges.
6654
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006655 The "source" keyword is useful in complex environments where a specific
6656 address only is allowed to connect to the servers. It may be needed when a
6657 private address must be used through a public gateway for instance, and it is
6658 known that the system cannot determine the adequate source address by itself.
6659
6660 An extension which is available on certain patched Linux kernels may be used
6661 through the "usesrc" optional keyword. It makes it possible to connect to the
6662 servers with an IP address which does not belong to the system itself. This
6663 is called "full transparent proxy mode". For this to work, the destination
6664 servers have to route their traffic back to this address through the machine
6665 running HAProxy, and IP forwarding must generally be enabled on this machine.
6666
6667 In this "full transparent proxy" mode, it is possible to force a specific IP
6668 address to be presented to the servers. This is not much used in fact. A more
6669 common use is to tell HAProxy to present the client's IP address. For this,
6670 there are two methods :
6671
6672 - present the client's IP and port addresses. This is the most transparent
6673 mode, but it can cause problems when IP connection tracking is enabled on
6674 the machine, because a same connection may be seen twice with different
6675 states. However, this solution presents the huge advantage of not
6676 limiting the system to the 64k outgoing address+port couples, because all
6677 of the client ranges may be used.
6678
6679 - present only the client's IP address and select a spare port. This
6680 solution is still quite elegant but slightly less transparent (downstream
6681 firewalls logs will not match upstream's). It also presents the downside
6682 of limiting the number of concurrent connections to the usual 64k ports.
6683 However, since the upstream and downstream ports are different, local IP
6684 connection tracking on the machine will not be upset by the reuse of the
6685 same session.
6686
6687 Note that depending on the transparent proxy technology used, it may be
6688 required to force the source address. In fact, cttproxy version 2 requires an
6689 IP address in <addr> above, and does not support setting of "0.0.0.0" as the
6690 IP address because it creates NAT entries which much match the exact outgoing
6691 address. Tproxy version 4 and some other kernel patches which work in pure
6692 forwarding mode generally will not have this limitation.
6693
6694 This option sets the default source for all servers in the backend. It may
6695 also be specified in a "defaults" section. Finer source address specification
6696 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006697 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006698
6699 Examples :
6700 backend private
6701 # Connect to the servers using our 192.168.1.200 source address
6702 source 192.168.1.200
6703
6704 backend transparent_ssl1
6705 # Connect to the SSL farm from the client's source address
6706 source 192.168.1.200 usesrc clientip
6707
6708 backend transparent_ssl2
6709 # Connect to the SSL farm from the client's source address and port
6710 # not recommended if IP conntrack is present on the local machine.
6711 source 192.168.1.200 usesrc client
6712
6713 backend transparent_ssl3
6714 # Connect to the SSL farm from the client's source address. It
6715 # is more conntrack-friendly.
6716 source 192.168.1.200 usesrc clientip
6717
6718 backend transparent_smtp
6719 # Connect to the SMTP farm from the client's source address/port
6720 # with Tproxy version 4.
6721 source 0.0.0.0 usesrc clientip
6722
Willy Tarreaubce70882009-09-07 11:51:47 +02006723 backend transparent_http
6724 # Connect to the servers using the client's IP as seen by previous
6725 # proxy.
6726 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
6727
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006728 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006729 the Linux kernel on www.balabit.com, the "bind" keyword.
6730
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006731
Willy Tarreau844e3c52008-01-11 16:28:18 +01006732srvtimeout <timeout> (deprecated)
6733 Set the maximum inactivity time on the server side.
6734 May be used in sections : defaults | frontend | listen | backend
6735 yes | no | yes | yes
6736 Arguments :
6737 <timeout> is the timeout value specified in milliseconds by default, but
6738 can be in any other unit if the number is suffixed by the unit,
6739 as explained at the top of this document.
6740
6741 The inactivity timeout applies when the server is expected to acknowledge or
6742 send data. In HTTP mode, this timeout is particularly important to consider
6743 during the first phase of the server's response, when it has to send the
6744 headers, as it directly represents the server's processing time for the
6745 request. To find out what value to put there, it's often good to start with
6746 what would be considered as unacceptable response times, then check the logs
6747 to observe the response time distribution, and adjust the value accordingly.
6748
6749 The value is specified in milliseconds by default, but can be in any other
6750 unit if the number is suffixed by the unit, as specified at the top of this
6751 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
6752 recommended that the client timeout remains equal to the server timeout in
6753 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006754 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01006755 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreaud72758d2010-01-12 10:42:19 +01006756 seconds (eg: 4 or 5 seconds minimum).
Willy Tarreau844e3c52008-01-11 16:28:18 +01006757
6758 This parameter is specific to backends, but can be specified once for all in
6759 "defaults" sections. This is in fact one of the easiest solutions not to
6760 forget about it. An unspecified timeout results in an infinite timeout, which
6761 is not recommended. Such a usage is accepted and works but reports a warning
6762 during startup because it may results in accumulation of expired sessions in
6763 the system if the system's timeouts are not configured either.
6764
6765 This parameter is provided for compatibility but is currently deprecated.
6766 Please use "timeout server" instead.
6767
Willy Tarreauce887fd2012-05-12 12:50:00 +02006768 See also : "timeout server", "timeout tunnel", "timeout client" and
6769 "clitimeout".
Willy Tarreau844e3c52008-01-11 16:28:18 +01006770
6771
Cyril Bonté66c327d2010-10-12 00:14:37 +02006772stats admin { if | unless } <cond>
6773 Enable statistics admin level if/unless a condition is matched
6774 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006775 no | yes | yes | yes
Cyril Bonté66c327d2010-10-12 00:14:37 +02006776
6777 This statement enables the statistics admin level if/unless a condition is
6778 matched.
6779
6780 The admin level allows to enable/disable servers from the web interface. By
6781 default, statistics page is read-only for security reasons.
6782
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006783 Note : Consider not using this feature in multi-process mode (nbproc > 1)
6784 unless you know what you do : memory is not shared between the
6785 processes, which can result in random behaviours.
6786
Cyril Bonté23b39d92011-02-10 22:54:44 +01006787 Currently, the POST request is limited to the buffer size minus the reserved
6788 buffer space, which means that if the list of servers is too long, the
6789 request won't be processed. It is recommended to alter few servers at a
6790 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +02006791
6792 Example :
6793 # statistics admin level only for localhost
6794 backend stats_localhost
6795 stats enable
6796 stats admin if LOCALHOST
6797
6798 Example :
6799 # statistics admin level always enabled because of the authentication
6800 backend stats_auth
6801 stats enable
6802 stats auth admin:AdMiN123
6803 stats admin if TRUE
6804
6805 Example :
6806 # statistics admin level depends on the authenticated user
6807 userlist stats-auth
6808 group admin users admin
6809 user admin insecure-password AdMiN123
6810 group readonly users haproxy
6811 user haproxy insecure-password haproxy
6812
6813 backend stats_auth
6814 stats enable
6815 acl AUTH http_auth(stats-auth)
6816 acl AUTH_ADMIN http_auth_group(stats-auth) admin
6817 stats http-request auth unless AUTH
6818 stats admin if AUTH_ADMIN
6819
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006820 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
6821 "bind-process", section 3.4 about userlists and section 7 about
6822 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +02006823
6824
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006825stats auth <user>:<passwd>
6826 Enable statistics with authentication and grant access to an account
6827 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006828 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006829 Arguments :
6830 <user> is a user name to grant access to
6831
6832 <passwd> is the cleartext password associated to this user
6833
6834 This statement enables statistics with default settings, and restricts access
6835 to declared users only. It may be repeated as many times as necessary to
6836 allow as many users as desired. When a user tries to access the statistics
6837 without a valid account, a "401 Forbidden" response will be returned so that
6838 the browser asks the user to provide a valid user and password. The real
6839 which will be returned to the browser is configurable using "stats realm".
6840
6841 Since the authentication method is HTTP Basic Authentication, the passwords
6842 circulate in cleartext on the network. Thus, it was decided that the
6843 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02006844 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006845
6846 It is also possible to reduce the scope of the proxies which appear in the
6847 report using "stats scope".
6848
6849 Though this statement alone is enough to enable statistics reporting, it is
6850 recommended to set all other settings in order to avoid relying on default
6851 unobvious parameters.
6852
6853 Example :
6854 # public access (limited to this backend only)
6855 backend public_www
6856 server srv1 192.168.0.1:80
6857 stats enable
6858 stats hide-version
6859 stats scope .
6860 stats uri /admin?stats
6861 stats realm Haproxy\ Statistics
6862 stats auth admin1:AdMiN123
6863 stats auth admin2:AdMiN321
6864
6865 # internal monitoring access (unlimited)
6866 backend private_monitoring
6867 stats enable
6868 stats uri /admin?stats
6869 stats refresh 5s
6870
6871 See also : "stats enable", "stats realm", "stats scope", "stats uri"
6872
6873
6874stats enable
6875 Enable statistics reporting with default settings
6876 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006877 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006878 Arguments : none
6879
6880 This statement enables statistics reporting with default settings defined
6881 at build time. Unless stated otherwise, these settings are used :
6882 - stats uri : /haproxy?stats
6883 - stats realm : "HAProxy Statistics"
6884 - stats auth : no authentication
6885 - stats scope : no restriction
6886
6887 Though this statement alone is enough to enable statistics reporting, it is
6888 recommended to set all other settings in order to avoid relying on default
6889 unobvious parameters.
6890
6891 Example :
6892 # public access (limited to this backend only)
6893 backend public_www
6894 server srv1 192.168.0.1:80
6895 stats enable
6896 stats hide-version
6897 stats scope .
6898 stats uri /admin?stats
6899 stats realm Haproxy\ Statistics
6900 stats auth admin1:AdMiN123
6901 stats auth admin2:AdMiN321
6902
6903 # internal monitoring access (unlimited)
6904 backend private_monitoring
6905 stats enable
6906 stats uri /admin?stats
6907 stats refresh 5s
6908
6909 See also : "stats auth", "stats realm", "stats uri"
6910
6911
Willy Tarreaud63335a2010-02-26 12:56:52 +01006912stats hide-version
6913 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02006914 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006915 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01006916 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02006917
Willy Tarreaud63335a2010-02-26 12:56:52 +01006918 By default, the stats page reports some useful status information along with
6919 the statistics. Among them is HAProxy's version. However, it is generally
6920 considered dangerous to report precise version to anyone, as it can help them
6921 target known weaknesses with specific attacks. The "stats hide-version"
6922 statement removes the version from the statistics report. This is recommended
6923 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02006924
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02006925 Though this statement alone is enough to enable statistics reporting, it is
6926 recommended to set all other settings in order to avoid relying on default
6927 unobvious parameters.
6928
Willy Tarreaud63335a2010-02-26 12:56:52 +01006929 Example :
6930 # public access (limited to this backend only)
6931 backend public_www
6932 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02006933 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01006934 stats hide-version
6935 stats scope .
6936 stats uri /admin?stats
6937 stats realm Haproxy\ Statistics
6938 stats auth admin1:AdMiN123
6939 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02006940
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02006941 # internal monitoring access (unlimited)
6942 backend private_monitoring
6943 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01006944 stats uri /admin?stats
6945 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +01006946
Willy Tarreaud63335a2010-02-26 12:56:52 +01006947 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02006948
Willy Tarreau983e01e2010-01-11 18:42:06 +01006949
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02006950stats http-request { allow | deny | auth [realm <realm>] }
6951 [ { if | unless } <condition> ]
6952 Access control for statistics
6953
6954 May be used in sections: defaults | frontend | listen | backend
6955 no | no | yes | yes
6956
6957 As "http-request", these set of options allow to fine control access to
6958 statistics. Each option may be followed by if/unless and acl.
6959 First option with matched condition (or option without condition) is final.
6960 For "deny" a 403 error will be returned, for "allow" normal processing is
6961 performed, for "auth" a 401/407 error code is returned so the client
6962 should be asked to enter a username and password.
6963
6964 There is no fixed limit to the number of http-request statements per
6965 instance.
6966
6967 See also : "http-request", section 3.4 about userlists and section 7
6968 about ACL usage.
6969
6970
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006971stats realm <realm>
6972 Enable statistics and set authentication realm
6973 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006974 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006975 Arguments :
6976 <realm> is the name of the HTTP Basic Authentication realm reported to
6977 the browser. The browser uses it to display it in the pop-up
6978 inviting the user to enter a valid username and password.
6979
6980 The realm is read as a single word, so any spaces in it should be escaped
6981 using a backslash ('\').
6982
6983 This statement is useful only in conjunction with "stats auth" since it is
6984 only related to authentication.
6985
6986 Though this statement alone is enough to enable statistics reporting, it is
6987 recommended to set all other settings in order to avoid relying on default
6988 unobvious parameters.
6989
6990 Example :
6991 # public access (limited to this backend only)
6992 backend public_www
6993 server srv1 192.168.0.1:80
6994 stats enable
6995 stats hide-version
6996 stats scope .
6997 stats uri /admin?stats
6998 stats realm Haproxy\ Statistics
6999 stats auth admin1:AdMiN123
7000 stats auth admin2:AdMiN321
7001
7002 # internal monitoring access (unlimited)
7003 backend private_monitoring
7004 stats enable
7005 stats uri /admin?stats
7006 stats refresh 5s
7007
7008 See also : "stats auth", "stats enable", "stats uri"
7009
7010
7011stats refresh <delay>
7012 Enable statistics with automatic refresh
7013 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007014 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007015 Arguments :
7016 <delay> is the suggested refresh delay, specified in seconds, which will
7017 be returned to the browser consulting the report page. While the
7018 browser is free to apply any delay, it will generally respect it
7019 and refresh the page this every seconds. The refresh interval may
7020 be specified in any other non-default time unit, by suffixing the
7021 unit after the value, as explained at the top of this document.
7022
7023 This statement is useful on monitoring displays with a permanent page
7024 reporting the load balancer's activity. When set, the HTML report page will
7025 include a link "refresh"/"stop refresh" so that the user can select whether
7026 he wants automatic refresh of the page or not.
7027
7028 Though this statement alone is enough to enable statistics reporting, it is
7029 recommended to set all other settings in order to avoid relying on default
7030 unobvious parameters.
7031
7032 Example :
7033 # public access (limited to this backend only)
7034 backend public_www
7035 server srv1 192.168.0.1:80
7036 stats enable
7037 stats hide-version
7038 stats scope .
7039 stats uri /admin?stats
7040 stats realm Haproxy\ Statistics
7041 stats auth admin1:AdMiN123
7042 stats auth admin2:AdMiN321
7043
7044 # internal monitoring access (unlimited)
7045 backend private_monitoring
7046 stats enable
7047 stats uri /admin?stats
7048 stats refresh 5s
7049
7050 See also : "stats auth", "stats enable", "stats realm", "stats uri"
7051
7052
7053stats scope { <name> | "." }
7054 Enable statistics and limit access scope
7055 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007056 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007057 Arguments :
7058 <name> is the name of a listen, frontend or backend section to be
7059 reported. The special name "." (a single dot) designates the
7060 section in which the statement appears.
7061
7062 When this statement is specified, only the sections enumerated with this
7063 statement will appear in the report. All other ones will be hidden. This
7064 statement may appear as many times as needed if multiple sections need to be
7065 reported. Please note that the name checking is performed as simple string
7066 comparisons, and that it is never checked that a give section name really
7067 exists.
7068
7069 Though this statement alone is enough to enable statistics reporting, it is
7070 recommended to set all other settings in order to avoid relying on default
7071 unobvious parameters.
7072
7073 Example :
7074 # public access (limited to this backend only)
7075 backend public_www
7076 server srv1 192.168.0.1:80
7077 stats enable
7078 stats hide-version
7079 stats scope .
7080 stats uri /admin?stats
7081 stats realm Haproxy\ Statistics
7082 stats auth admin1:AdMiN123
7083 stats auth admin2:AdMiN321
7084
7085 # internal monitoring access (unlimited)
7086 backend private_monitoring
7087 stats enable
7088 stats uri /admin?stats
7089 stats refresh 5s
7090
7091 See also : "stats auth", "stats enable", "stats realm", "stats uri"
7092
Willy Tarreaud63335a2010-02-26 12:56:52 +01007093
Willy Tarreauc9705a12010-07-27 20:05:50 +02007094stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +01007095 Enable reporting of a description on the statistics page.
7096 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007097 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01007098
Willy Tarreauc9705a12010-07-27 20:05:50 +02007099 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +01007100 description from global section is automatically used instead.
7101
7102 This statement is useful for users that offer shared services to their
7103 customers, where node or description should be different for each customer.
7104
7105 Though this statement alone is enough to enable statistics reporting, it is
7106 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007107 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +01007108
7109 Example :
7110 # internal monitoring access (unlimited)
7111 backend private_monitoring
7112 stats enable
7113 stats show-desc Master node for Europe, Asia, Africa
7114 stats uri /admin?stats
7115 stats refresh 5s
7116
7117 See also: "show-node", "stats enable", "stats uri" and "description" in
7118 global section.
7119
7120
7121stats show-legends
Willy Tarreaued2119c2014-04-24 22:10:39 +02007122 Enable reporting additional information on the statistics page
7123 May be used in sections : defaults | frontend | listen | backend
7124 yes | yes | yes | yes
7125 Arguments : none
7126
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007127 Enable reporting additional information on the statistics page :
Willy Tarreaud63335a2010-02-26 12:56:52 +01007128 - cap: capabilities (proxy)
7129 - mode: one of tcp, http or health (proxy)
7130 - id: SNMP ID (proxy, socket, server)
7131 - IP (socket, server)
7132 - cookie (backend, server)
7133
7134 Though this statement alone is enough to enable statistics reporting, it is
7135 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007136 unobvious parameters. Default behaviour is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +01007137
7138 See also: "stats enable", "stats uri".
7139
7140
7141stats show-node [ <name> ]
7142 Enable reporting of a host name on the statistics page.
7143 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007144 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01007145 Arguments:
7146 <name> is an optional name to be reported. If unspecified, the
7147 node name from global section is automatically used instead.
7148
7149 This statement is useful for users that offer shared services to their
7150 customers, where node or description might be different on a stats page
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007151 provided for each customer. Default behaviour is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +01007152
7153 Though this statement alone is enough to enable statistics reporting, it is
7154 recommended to set all other settings in order to avoid relying on default
7155 unobvious parameters.
7156
7157 Example:
7158 # internal monitoring access (unlimited)
7159 backend private_monitoring
7160 stats enable
7161 stats show-node Europe-1
7162 stats uri /admin?stats
7163 stats refresh 5s
7164
7165 See also: "show-desc", "stats enable", "stats uri", and "node" in global
7166 section.
7167
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007168
7169stats uri <prefix>
7170 Enable statistics and define the URI prefix to access them
7171 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007172 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007173 Arguments :
7174 <prefix> is the prefix of any URI which will be redirected to stats. This
7175 prefix may contain a question mark ('?') to indicate part of a
7176 query string.
7177
7178 The statistics URI is intercepted on the relayed traffic, so it appears as a
7179 page within the normal application. It is strongly advised to ensure that the
7180 selected URI will never appear in the application, otherwise it will never be
7181 possible to reach it in the application.
7182
7183 The default URI compiled in haproxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007184 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007185 It is generally a good idea to include a question mark in the URI so that
7186 intermediate proxies refrain from caching the results. Also, since any string
7187 beginning with the prefix will be accepted as a stats request, the question
7188 mark helps ensuring that no valid URI will begin with the same words.
7189
7190 It is sometimes very convenient to use "/" as the URI prefix, and put that
7191 statement in a "listen" instance of its own. That makes it easy to dedicate
7192 an address or a port to statistics only.
7193
7194 Though this statement alone is enough to enable statistics reporting, it is
7195 recommended to set all other settings in order to avoid relying on default
7196 unobvious parameters.
7197
7198 Example :
7199 # public access (limited to this backend only)
7200 backend public_www
7201 server srv1 192.168.0.1:80
7202 stats enable
7203 stats hide-version
7204 stats scope .
7205 stats uri /admin?stats
7206 stats realm Haproxy\ Statistics
7207 stats auth admin1:AdMiN123
7208 stats auth admin2:AdMiN321
7209
7210 # internal monitoring access (unlimited)
7211 backend private_monitoring
7212 stats enable
7213 stats uri /admin?stats
7214 stats refresh 5s
7215
7216 See also : "stats auth", "stats enable", "stats realm"
7217
7218
Willy Tarreaud63335a2010-02-26 12:56:52 +01007219stick match <pattern> [table <table>] [{if | unless} <cond>]
7220 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007221 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +01007222 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007223
7224 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02007225 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007226 describes what elements of the incoming request or connection
7227 will be analysed in the hope to find a matching entry in a
7228 stickiness table. This rule is mandatory.
7229
7230 <table> is an optional stickiness table name. If unspecified, the same
7231 backend's table is used. A stickiness table is declared using
7232 the "stick-table" statement.
7233
7234 <cond> is an optional matching condition. It makes it possible to match
7235 on a certain criterion only when other conditions are met (or
7236 not met). For instance, it could be used to match on a source IP
7237 address except when a request passes through a known proxy, in
7238 which case we'd match on a header containing that IP address.
7239
7240 Some protocols or applications require complex stickiness rules and cannot
7241 always simply rely on cookies nor hashing. The "stick match" statement
7242 describes a rule to extract the stickiness criterion from an incoming request
7243 or connection. See section 7 for a complete list of possible patterns and
7244 transformation rules.
7245
7246 The table has to be declared using the "stick-table" statement. It must be of
7247 a type compatible with the pattern. By default it is the one which is present
7248 in the same backend. It is possible to share a table with other backends by
7249 referencing it using the "table" keyword. If another table is referenced,
7250 the server's ID inside the backends are used. By default, all server IDs
7251 start at 1 in each backend, so the server ordering is enough. But in case of
7252 doubt, it is highly recommended to force server IDs using their "id" setting.
7253
7254 It is possible to restrict the conditions where a "stick match" statement
7255 will apply, using "if" or "unless" followed by a condition. See section 7 for
7256 ACL based conditions.
7257
7258 There is no limit on the number of "stick match" statements. The first that
7259 applies and matches will cause the request to be directed to the same server
7260 as was used for the request which created the entry. That way, multiple
7261 matches can be used as fallbacks.
7262
7263 The stick rules are checked after the persistence cookies, so they will not
7264 affect stickiness if a cookie has already been used to select a server. That
7265 way, it becomes very easy to insert cookies and match on IP addresses in
7266 order to maintain stickiness between HTTP and HTTPS.
7267
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007268 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7269 unless you know what you do : memory is not shared between the
7270 processes, which can result in random behaviours.
7271
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007272 Example :
7273 # forward SMTP users to the same server they just used for POP in the
7274 # last 30 minutes
7275 backend pop
7276 mode tcp
7277 balance roundrobin
7278 stick store-request src
7279 stick-table type ip size 200k expire 30m
7280 server s1 192.168.1.1:110
7281 server s2 192.168.1.1:110
7282
7283 backend smtp
7284 mode tcp
7285 balance roundrobin
7286 stick match src table pop
7287 server s1 192.168.1.1:25
7288 server s2 192.168.1.1:25
7289
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007290 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02007291 about ACLs and samples fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007292
7293
7294stick on <pattern> [table <table>] [{if | unless} <condition>]
7295 Define a request pattern to associate a user to a server
7296 May be used in sections : defaults | frontend | listen | backend
7297 no | no | yes | yes
7298
7299 Note : This form is exactly equivalent to "stick match" followed by
7300 "stick store-request", all with the same arguments. Please refer
7301 to both keywords for details. It is only provided as a convenience
7302 for writing more maintainable configurations.
7303
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007304 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7305 unless you know what you do : memory is not shared between the
7306 processes, which can result in random behaviours.
7307
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007308 Examples :
7309 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +01007310 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007311
7312 # ...is strictly equivalent to this one :
7313 stick match src table pop if !localhost
7314 stick store-request src table pop if !localhost
7315
7316
7317 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
7318 # well as HTTP without cookie. Share the same table between both accesses.
7319 backend http
7320 mode http
7321 balance roundrobin
7322 stick on src table https
7323 cookie SRV insert indirect nocache
7324 server s1 192.168.1.1:80 cookie s1
7325 server s2 192.168.1.1:80 cookie s2
7326
7327 backend https
7328 mode tcp
7329 balance roundrobin
7330 stick-table type ip size 200k expire 30m
7331 stick on src
7332 server s1 192.168.1.1:443
7333 server s2 192.168.1.1:443
7334
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007335 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007336
7337
7338stick store-request <pattern> [table <table>] [{if | unless} <condition>]
7339 Define a request pattern used to create an entry in a stickiness table
7340 May be used in sections : defaults | frontend | listen | backend
7341 no | no | yes | yes
7342
7343 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02007344 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007345 describes what elements of the incoming request or connection
7346 will be analysed, extracted and stored in the table once a
7347 server is selected.
7348
7349 <table> is an optional stickiness table name. If unspecified, the same
7350 backend's table is used. A stickiness table is declared using
7351 the "stick-table" statement.
7352
7353 <cond> is an optional storage condition. It makes it possible to store
7354 certain criteria only when some conditions are met (or not met).
7355 For instance, it could be used to store the source IP address
7356 except when the request passes through a known proxy, in which
7357 case we'd store a converted form of a header containing that IP
7358 address.
7359
7360 Some protocols or applications require complex stickiness rules and cannot
7361 always simply rely on cookies nor hashing. The "stick store-request" statement
7362 describes a rule to decide what to extract from the request and when to do
7363 it, in order to store it into a stickiness table for further requests to
7364 match it using the "stick match" statement. Obviously the extracted part must
7365 make sense and have a chance to be matched in a further request. Storing a
7366 client's IP address for instance often makes sense. Storing an ID found in a
7367 URL parameter also makes sense. Storing a source port will almost never make
7368 any sense because it will be randomly matched. See section 7 for a complete
7369 list of possible patterns and transformation rules.
7370
7371 The table has to be declared using the "stick-table" statement. It must be of
7372 a type compatible with the pattern. By default it is the one which is present
7373 in the same backend. It is possible to share a table with other backends by
7374 referencing it using the "table" keyword. If another table is referenced,
7375 the server's ID inside the backends are used. By default, all server IDs
7376 start at 1 in each backend, so the server ordering is enough. But in case of
7377 doubt, it is highly recommended to force server IDs using their "id" setting.
7378
7379 It is possible to restrict the conditions where a "stick store-request"
7380 statement will apply, using "if" or "unless" followed by a condition. This
7381 condition will be evaluated while parsing the request, so any criteria can be
7382 used. See section 7 for ACL based conditions.
7383
7384 There is no limit on the number of "stick store-request" statements, but
7385 there is a limit of 8 simultaneous stores per request or response. This
7386 makes it possible to store up to 8 criteria, all extracted from either the
7387 request or the response, regardless of the number of rules. Only the 8 first
7388 ones which match will be kept. Using this, it is possible to feed multiple
7389 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01007390 another protocol or access method. Using multiple store-request rules with
7391 the same table is possible and may be used to find the best criterion to rely
7392 on, by arranging the rules by decreasing preference order. Only the first
7393 extracted criterion for a given table will be stored. All subsequent store-
7394 request rules referencing the same table will be skipped and their ACLs will
7395 not be evaluated.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007396
7397 The "store-request" rules are evaluated once the server connection has been
7398 established, so that the table will contain the real server that processed
7399 the request.
7400
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007401 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7402 unless you know what you do : memory is not shared between the
7403 processes, which can result in random behaviours.
7404
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007405 Example :
7406 # forward SMTP users to the same server they just used for POP in the
7407 # last 30 minutes
7408 backend pop
7409 mode tcp
7410 balance roundrobin
7411 stick store-request src
7412 stick-table type ip size 200k expire 30m
7413 server s1 192.168.1.1:110
7414 server s2 192.168.1.1:110
7415
7416 backend smtp
7417 mode tcp
7418 balance roundrobin
7419 stick match src table pop
7420 server s1 192.168.1.1:25
7421 server s2 192.168.1.1:25
7422
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007423 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02007424 about ACLs and sample fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007425
7426
Emeric Brun7c6b82e2010-09-24 16:34:28 +02007427stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Emeric Brunf099e792010-09-27 12:05:28 +02007428 size <size> [expire <expire>] [nopurge] [peers <peersect>]
7429 [store <data_type>]*
Godbach64cef792013-12-04 16:08:22 +08007430 Configure the stickiness table for the current section
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007431 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +02007432 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007433
7434 Arguments :
7435 ip a table declared with "type ip" will only store IPv4 addresses.
7436 This form is very compact (about 50 bytes per entry) and allows
7437 very fast entry lookup and stores with almost no overhead. This
7438 is mainly used to store client source IP addresses.
7439
David du Colombier9a6d3c92011-03-17 10:40:24 +01007440 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
7441 This form is very compact (about 60 bytes per entry) and allows
7442 very fast entry lookup and stores with almost no overhead. This
7443 is mainly used to store client source IP addresses.
7444
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007445 integer a table declared with "type integer" will store 32bit integers
7446 which can represent a client identifier found in a request for
7447 instance.
7448
7449 string a table declared with "type string" will store substrings of up
7450 to <len> characters. If the string provided by the pattern
7451 extractor is larger than <len>, it will be truncated before
7452 being stored. During matching, at most <len> characters will be
7453 compared between the string in the table and the extracted
7454 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +02007455 to 32 characters.
7456
7457 binary a table declared with "type binary" will store binary blocks
7458 of <len> bytes. If the block provided by the pattern
7459 extractor is larger than <len>, it will be truncated before
Willy Tarreaube722a22014-06-13 16:31:59 +02007460 being stored. If the block provided by the sample expression
Emeric Brun7c6b82e2010-09-24 16:34:28 +02007461 is shorter than <len>, it will be padded by 0. When not
7462 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007463
7464 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +02007465 "string" type table (See type "string" above). Or the number
7466 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007467 changing this parameter as memory usage will proportionally
7468 increase.
7469
7470 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +01007471 value directly impacts memory usage. Count approximately
7472 50 bytes per entry, plus the size of a string if any. The size
7473 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007474
7475 [nopurge] indicates that we refuse to purge older entries when the table
7476 is full. When not specified and the table is full when haproxy
7477 wants to store an entry in it, it will flush a few of the oldest
7478 entries in order to release some space for the new ones. This is
7479 most often the desired behaviour. In some specific cases, it
7480 be desirable to refuse new entries instead of purging the older
7481 ones. That may be the case when the amount of data to store is
7482 far above the hardware limits and we prefer not to offer access
7483 to new clients than to reject the ones already connected. When
7484 using this parameter, be sure to properly set the "expire"
7485 parameter (see below).
7486
Emeric Brunf099e792010-09-27 12:05:28 +02007487 <peersect> is the name of the peers section to use for replication. Entries
7488 which associate keys to server IDs are kept synchronized with
7489 the remote peers declared in this section. All entries are also
7490 automatically learned from the local peer (old process) during a
7491 soft restart.
7492
Willy Tarreau1abc6732015-05-01 19:21:02 +02007493 NOTE : each peers section may be referenced only by tables
7494 belonging to the same unique process.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007495
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007496 <expire> defines the maximum duration of an entry in the table since it
7497 was last created, refreshed or matched. The expiration delay is
7498 defined using the standard time format, similarly as the various
7499 timeouts. The maximum duration is slightly above 24 days. See
7500 section 2.2 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +02007501 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007502 be removed once full. Be sure not to use the "nopurge" parameter
7503 if not expiration delay is specified.
7504
Willy Tarreau08d5f982010-06-06 13:34:54 +02007505 <data_type> is used to store additional information in the stick-table. This
7506 may be used by ACLs in order to control various criteria related
7507 to the activity of the client matching the stick-table. For each
7508 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +02007509 that the additional data can fit. Several data types may be
7510 stored with an entry. Multiple data types may be specified after
7511 the "store" keyword, as a comma-separated list. Alternatively,
7512 it is possible to repeat the "store" keyword followed by one or
7513 several data types. Except for the "server_id" type which is
7514 automatically detected and enabled, all data types must be
7515 explicitly declared to be stored. If an ACL references a data
7516 type which is not stored, the ACL will simply not match. Some
7517 data types require an argument which must be passed just after
7518 the type between parenthesis. See below for the supported data
7519 types and their arguments.
7520
7521 The data types that can be stored with an entry are the following :
7522 - server_id : this is an integer which holds the numeric ID of the server a
7523 request was assigned to. It is used by the "stick match", "stick store",
7524 and "stick on" rules. It is automatically enabled when referenced.
7525
7526 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
7527 integer which may be used for anything. Most of the time it will be used
7528 to put a special tag on some entries, for instance to note that a
7529 specific behaviour was detected and must be known for future matches.
7530
Willy Tarreauba2ffd12013-05-29 15:54:14 +02007531 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
7532 over a period. It is a positive 32-bit integer integer which may be used
7533 for anything. Just like <gpc0>, it counts events, but instead of keeping
7534 a cumulative count, it maintains the rate at which the counter is
7535 incremented. Most of the time it will be used to measure the frequency of
7536 occurrence of certain events (eg: requests to a specific URL).
7537
Willy Tarreauc9705a12010-07-27 20:05:50 +02007538 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
7539 the absolute number of connections received from clients which matched
7540 this entry. It does not mean the connections were accepted, just that
7541 they were received.
7542
7543 - conn_cur : Current Connections. It is a positive 32-bit integer which
7544 stores the concurrent connection counts for the entry. It is incremented
7545 once an incoming connection matches the entry, and decremented once the
7546 connection leaves. That way it is possible to know at any time the exact
7547 number of concurrent connections for an entry.
7548
7549 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
7550 integer parameter <period> which indicates in milliseconds the length
7551 of the period over which the average is measured. It reports the average
7552 incoming connection rate over that period, in connections per period. The
7553 result is an integer which can be matched using ACLs.
7554
7555 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
7556 the absolute number of sessions received from clients which matched this
7557 entry. A session is a connection that was accepted by the layer 4 rules.
7558
7559 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
7560 integer parameter <period> which indicates in milliseconds the length
7561 of the period over which the average is measured. It reports the average
7562 incoming session rate over that period, in sessions per period. The
7563 result is an integer which can be matched using ACLs.
7564
7565 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
7566 counts the absolute number of HTTP requests received from clients which
7567 matched this entry. It does not matter whether they are valid requests or
7568 not. Note that this is different from sessions when keep-alive is used on
7569 the client side.
7570
7571 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
7572 integer parameter <period> which indicates in milliseconds the length
7573 of the period over which the average is measured. It reports the average
7574 HTTP request rate over that period, in requests per period. The result is
7575 an integer which can be matched using ACLs. It does not matter whether
7576 they are valid requests or not. Note that this is different from sessions
7577 when keep-alive is used on the client side.
7578
7579 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
7580 counts the absolute number of HTTP requests errors induced by clients
7581 which matched this entry. Errors are counted on invalid and truncated
7582 requests, as well as on denied or tarpitted requests, and on failed
7583 authentications. If the server responds with 4xx, then the request is
7584 also counted as an error since it's an error triggered by the client
7585 (eg: vulnerability scan).
7586
7587 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
7588 integer parameter <period> which indicates in milliseconds the length
7589 of the period over which the average is measured. It reports the average
7590 HTTP request error rate over that period, in requests per period (see
7591 http_err_cnt above for what is accounted as an error). The result is an
7592 integer which can be matched using ACLs.
7593
7594 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
7595 integer which counts the cumulated amount of bytes received from clients
7596 which matched this entry. Headers are included in the count. This may be
7597 used to limit abuse of upload features on photo or video servers.
7598
7599 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
7600 integer parameter <period> which indicates in milliseconds the length
7601 of the period over which the average is measured. It reports the average
7602 incoming bytes rate over that period, in bytes per period. It may be used
7603 to detect users which upload too much and too fast. Warning: with large
7604 uploads, it is possible that the amount of uploaded data will be counted
7605 once upon termination, thus causing spikes in the average transfer speed
7606 instead of having a smooth one. This may partially be smoothed with
7607 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
7608 recommended for better fairness.
7609
7610 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
7611 integer which counts the cumulated amount of bytes sent to clients which
7612 matched this entry. Headers are included in the count. This may be used
7613 to limit abuse of bots sucking the whole site.
7614
7615 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
7616 an integer parameter <period> which indicates in milliseconds the length
7617 of the period over which the average is measured. It reports the average
7618 outgoing bytes rate over that period, in bytes per period. It may be used
7619 to detect users which download too much and too fast. Warning: with large
7620 transfers, it is possible that the amount of transferred data will be
7621 counted once upon termination, thus causing spikes in the average
7622 transfer speed instead of having a smooth one. This may partially be
7623 smoothed with "option contstats" though this is not perfect yet. Use of
7624 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +02007625
Willy Tarreauc00cdc22010-06-06 16:48:26 +02007626 There is only one stick-table per proxy. At the moment of writing this doc,
7627 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007628 to be required, simply create a dummy backend with a stick-table in it and
7629 reference it.
7630
7631 It is important to understand that stickiness based on learning information
7632 has some limitations, including the fact that all learned associations are
7633 lost upon restart. In general it can be good as a complement but not always
7634 as an exclusive stickiness.
7635
Willy Tarreauc9705a12010-07-27 20:05:50 +02007636 Last, memory requirements may be important when storing many data types.
7637 Indeed, storing all indicators above at once in each entry requires 116 bytes
7638 per entry, or 116 MB for a 1-million entries table. This is definitely not
7639 something that can be ignored.
7640
7641 Example:
7642 # Keep track of counters of up to 1 million IP addresses over 5 minutes
7643 # and store a general purpose counter and the average connection rate
7644 # computed over a sliding window of 30 seconds.
7645 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
7646
7647 See also : "stick match", "stick on", "stick store-request", section 2.2
David du Colombiera13d1b92011-03-17 10:40:22 +01007648 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007649
7650
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007651stick store-response <pattern> [table <table>] [{if | unless} <condition>]
7652 Define a request pattern used to create an entry in a stickiness table
7653 May be used in sections : defaults | frontend | listen | backend
7654 no | no | yes | yes
7655
7656 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02007657 <pattern> is a sample expression rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007658 describes what elements of the response or connection will
7659 be analysed, extracted and stored in the table once a
7660 server is selected.
7661
7662 <table> is an optional stickiness table name. If unspecified, the same
7663 backend's table is used. A stickiness table is declared using
7664 the "stick-table" statement.
7665
7666 <cond> is an optional storage condition. It makes it possible to store
7667 certain criteria only when some conditions are met (or not met).
7668 For instance, it could be used to store the SSL session ID only
7669 when the response is a SSL server hello.
7670
7671 Some protocols or applications require complex stickiness rules and cannot
7672 always simply rely on cookies nor hashing. The "stick store-response"
7673 statement describes a rule to decide what to extract from the response and
7674 when to do it, in order to store it into a stickiness table for further
7675 requests to match it using the "stick match" statement. Obviously the
7676 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007677 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007678 See section 7 for a complete list of possible patterns and transformation
7679 rules.
7680
7681 The table has to be declared using the "stick-table" statement. It must be of
7682 a type compatible with the pattern. By default it is the one which is present
7683 in the same backend. It is possible to share a table with other backends by
7684 referencing it using the "table" keyword. If another table is referenced,
7685 the server's ID inside the backends are used. By default, all server IDs
7686 start at 1 in each backend, so the server ordering is enough. But in case of
7687 doubt, it is highly recommended to force server IDs using their "id" setting.
7688
7689 It is possible to restrict the conditions where a "stick store-response"
7690 statement will apply, using "if" or "unless" followed by a condition. This
7691 condition will be evaluated while parsing the response, so any criteria can
7692 be used. See section 7 for ACL based conditions.
7693
7694 There is no limit on the number of "stick store-response" statements, but
7695 there is a limit of 8 simultaneous stores per request or response. This
7696 makes it possible to store up to 8 criteria, all extracted from either the
7697 request or the response, regardless of the number of rules. Only the 8 first
7698 ones which match will be kept. Using this, it is possible to feed multiple
7699 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01007700 another protocol or access method. Using multiple store-response rules with
7701 the same table is possible and may be used to find the best criterion to rely
7702 on, by arranging the rules by decreasing preference order. Only the first
7703 extracted criterion for a given table will be stored. All subsequent store-
7704 response rules referencing the same table will be skipped and their ACLs will
7705 not be evaluated. However, even if a store-request rule references a table, a
7706 store-response rule may also use the same table. This means that each table
7707 may learn exactly one element from the request and one element from the
7708 response at once.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007709
7710 The table will contain the real server that processed the request.
7711
7712 Example :
7713 # Learn SSL session ID from both request and response and create affinity.
7714 backend https
7715 mode tcp
7716 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +02007717 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007718 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007719
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007720 acl clienthello req_ssl_hello_type 1
7721 acl serverhello rep_ssl_hello_type 2
7722
7723 # use tcp content accepts to detects ssl client and server hello.
7724 tcp-request inspect-delay 5s
7725 tcp-request content accept if clienthello
7726
7727 # no timeout on response inspect delay by default.
7728 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007729
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007730 # SSL session ID (SSLID) may be present on a client or server hello.
7731 # Its length is coded on 1 byte at offset 43 and its value starts
7732 # at offset 44.
7733
7734 # Match and learn on request if client hello.
7735 stick on payload_lv(43,1) if clienthello
7736
7737 # Learn on response if server hello.
7738 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +02007739
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007740 server s1 192.168.1.1:443
7741 server s2 192.168.1.1:443
7742
7743 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
7744 extraction.
7745
7746
Willy Tarreau938c7fe2014-04-25 14:21:39 +02007747tcp-check connect [params*]
7748 Opens a new connection
7749 May be used in sections: defaults | frontend | listen | backend
7750 no | no | yes | yes
7751
7752 When an application lies on more than a single TCP port or when HAProxy
7753 load-balance many services in a single backend, it makes sense to probe all
7754 the services individually before considering a server as operational.
7755
7756 When there are no TCP port configured on the server line neither server port
7757 directive, then the 'tcp-check connect port <port>' must be the first step
7758 of the sequence.
7759
7760 In a tcp-check ruleset a 'connect' is required, it is also mandatory to start
7761 the ruleset with a 'connect' rule. Purpose is to ensure admin know what they
7762 do.
7763
7764 Parameters :
7765 They are optional and can be used to describe how HAProxy should open and
7766 use the TCP connection.
7767
7768 port if not set, check port or server port is used.
7769 It tells HAProxy where to open the connection to.
7770 <port> must be a valid TCP port source integer, from 1 to 65535.
7771
7772 send-proxy send a PROXY protocol string
7773
7774 ssl opens a ciphered connection
7775
7776 Examples:
7777 # check HTTP and HTTPs services on a server.
7778 # first open port 80 thanks to server line port directive, then
7779 # tcp-check opens port 443, ciphered and run a request on it:
7780 option tcp-check
7781 tcp-check connect
7782 tcp-check send GET\ /\ HTTP/1.0\r\n
7783 tcp-check send Host:\ haproxy.1wt.eu\r\n
7784 tcp-check send \r\n
7785 tcp-check expect rstring (2..|3..)
7786 tcp-check connect port 443 ssl
7787 tcp-check send GET\ /\ HTTP/1.0\r\n
7788 tcp-check send Host:\ haproxy.1wt.eu\r\n
7789 tcp-check send \r\n
7790 tcp-check expect rstring (2..|3..)
7791 server www 10.0.0.1 check port 80
7792
7793 # check both POP and IMAP from a single server:
7794 option tcp-check
7795 tcp-check connect port 110
7796 tcp-check expect string +OK\ POP3\ ready
7797 tcp-check connect port 143
7798 tcp-check expect string *\ OK\ IMAP4\ ready
7799 server mail 10.0.0.1 check
7800
7801 See also : "option tcp-check", "tcp-check send", "tcp-check expect"
7802
7803
7804tcp-check expect [!] <match> <pattern>
7805 Specify data to be collected and analysed during a generic health check
7806 May be used in sections: defaults | frontend | listen | backend
7807 no | no | yes | yes
7808
7809 Arguments :
7810 <match> is a keyword indicating how to look for a specific pattern in the
7811 response. The keyword may be one of "string", "rstring" or
7812 binary.
7813 The keyword may be preceded by an exclamation mark ("!") to negate
7814 the match. Spaces are allowed between the exclamation mark and the
7815 keyword. See below for more details on the supported keywords.
7816
7817 <pattern> is the pattern to look for. It may be a string or a regular
7818 expression. If the pattern contains spaces, they must be escaped
7819 with the usual backslash ('\').
7820 If the match is set to binary, then the pattern must be passed as
7821 a serie of hexadecimal digits in an even number. Each sequence of
7822 two digits will represent a byte. The hexadecimal digits may be
7823 used upper or lower case.
7824
7825
7826 The available matches are intentionally similar to their http-check cousins :
7827
7828 string <string> : test the exact string matches in the response buffer.
7829 A health check response will be considered valid if the
7830 response's buffer contains this exact string. If the
7831 "string" keyword is prefixed with "!", then the response
7832 will be considered invalid if the body contains this
7833 string. This can be used to look for a mandatory pattern
7834 in a protocol response, or to detect a failure when a
7835 specific error appears in a protocol banner.
7836
7837 rstring <regex> : test a regular expression on the response buffer.
7838 A health check response will be considered valid if the
7839 response's buffer matches this expression. If the
7840 "rstring" keyword is prefixed with "!", then the response
7841 will be considered invalid if the body matches the
7842 expression.
7843
7844 binary <hexstring> : test the exact string in its hexadecimal form matches
7845 in the response buffer. A health check response will
7846 be considered valid if the response's buffer contains
7847 this exact hexadecimal string.
7848 Purpose is to match data on binary protocols.
7849
7850 It is important to note that the responses will be limited to a certain size
7851 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
7852 Thus, too large responses may not contain the mandatory pattern when using
7853 "string", "rstring" or binary. If a large response is absolutely required, it
7854 is possible to change the default max size by setting the global variable.
7855 However, it is worth keeping in mind that parsing very large responses can
7856 waste some CPU cycles, especially when regular expressions are used, and that
7857 it is always better to focus the checks on smaller resources. Also, in its
7858 current state, the check will not find any string nor regex past a null
7859 character in the response. Similarly it is not possible to request matching
7860 the null character.
7861
7862 Examples :
7863 # perform a POP check
7864 option tcp-check
7865 tcp-check expect string +OK\ POP3\ ready
7866
7867 # perform an IMAP check
7868 option tcp-check
7869 tcp-check expect string *\ OK\ IMAP4\ ready
7870
7871 # look for the redis master server
7872 option tcp-check
7873 tcp-check send PING\r\n
7874 tcp-check expect +PONG
7875 tcp-check send info\ replication\r\n
7876 tcp-check expect string role:master
7877 tcp-check send QUIT\r\n
7878 tcp-check expect string +OK
7879
7880
7881 See also : "option tcp-check", "tcp-check connect", "tcp-check send",
7882 "tcp-check send-binary", "http-check expect", tune.chksize
7883
7884
7885tcp-check send <data>
7886 Specify a string to be sent as a question during a generic health check
7887 May be used in sections: defaults | frontend | listen | backend
7888 no | no | yes | yes
7889
7890 <data> : the data to be sent as a question during a generic health check
7891 session. For now, <data> must be a string.
7892
7893 Examples :
7894 # look for the redis master server
7895 option tcp-check
7896 tcp-check send info\ replication\r\n
7897 tcp-check expect string role:master
7898
7899 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
7900 "tcp-check send-binary", tune.chksize
7901
7902
7903tcp-check send-binary <hexastring>
7904 Specify an hexa digits string to be sent as a binary question during a raw
7905 tcp health check
7906 May be used in sections: defaults | frontend | listen | backend
7907 no | no | yes | yes
7908
7909 <data> : the data to be sent as a question during a generic health check
7910 session. For now, <data> must be a string.
7911 <hexastring> : test the exact string in its hexadecimal form matches in the
7912 response buffer. A health check response will be considered
7913 valid if the response's buffer contains this exact
7914 hexadecimal string.
7915 Purpose is to send binary data to ask on binary protocols.
7916
7917 Examples :
7918 # redis check in binary
7919 option tcp-check
7920 tcp-check send-binary 50494e470d0a # PING\r\n
7921 tcp-check expect binary 2b504F4e47 # +PONG
7922
7923
7924 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
7925 "tcp-check send", tune.chksize
7926
7927
Willy Tarreaue9656522010-08-17 15:40:09 +02007928tcp-request connection <action> [{if | unless} <condition>]
7929 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +02007930 May be used in sections : defaults | frontend | listen | backend
7931 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +02007932 Arguments :
7933 <action> defines the action to perform if the condition applies. Valid
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007934 actions include : "accept", "reject", "track-sc0", "track-sc1",
7935 "track-sc2", and "expect-proxy". See below for more details.
Willy Tarreau1a687942010-05-23 22:40:30 +02007936
Willy Tarreaue9656522010-08-17 15:40:09 +02007937 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007938
7939 Immediately after acceptance of a new incoming connection, it is possible to
7940 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +02007941 or dropped or have its counters tracked. Those conditions cannot make use of
7942 any data contents because the connection has not been read from yet, and the
7943 buffers are not yet allocated. This is used to selectively and very quickly
7944 accept or drop connections from various sources with a very low overhead. If
7945 some contents need to be inspected in order to take the decision, the
7946 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007947
Willy Tarreaue9656522010-08-17 15:40:09 +02007948 The "tcp-request connection" rules are evaluated in their exact declaration
7949 order. If no rule matches or if there is no rule, the default action is to
7950 accept the incoming connection. There is no specific limit to the number of
7951 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007952
Willy Tarreaua9083d02015-05-08 15:27:59 +02007953 Four types of actions are supported :
Willy Tarreaue9656522010-08-17 15:40:09 +02007954 - accept :
7955 accepts the connection if the condition is true (when used with "if")
7956 or false (when used with "unless"). The first such rule executed ends
7957 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007958
Willy Tarreaue9656522010-08-17 15:40:09 +02007959 - reject :
7960 rejects the connection if the condition is true (when used with "if")
7961 or false (when used with "unless"). The first such rule executed ends
7962 the rules evaluation. Rejected connections do not even become a
7963 session, which is why they are accounted separately for in the stats,
7964 as "denied connections". They are not considered for the session
7965 rate-limit and are not logged either. The reason is that these rules
7966 should only be used to filter extremely high connection rates such as
7967 the ones encountered during a massive DDoS attack. Under these extreme
7968 conditions, the simple action of logging each event would make the
7969 system collapse and would considerably lower the filtering capacity. If
7970 logging is absolutely desired, then "tcp-request content" rules should
7971 be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007972
Willy Tarreau4f0d9192013-06-11 20:40:55 +02007973 - expect-proxy layer4 :
7974 configures the client-facing connection to receive a PROXY protocol
7975 header before any byte is read from the socket. This is equivalent to
7976 having the "accept-proxy" keyword on the "bind" line, except that using
7977 the TCP rule allows the PROXY protocol to be accepted only for certain
7978 IP address ranges using an ACL. This is convenient when multiple layers
7979 of load balancers are passed through by traffic coming from public
7980 hosts.
7981
Willy Tarreau18bf01e2014-06-13 16:18:52 +02007982 - capture <sample> len <length> :
7983 This only applies to "tcp-request content" rules. It captures sample
7984 expression <sample> from the request buffer, and converts it to a
7985 string of at most <len> characters. The resulting string is stored into
7986 the next request "capture" slot, so it will possibly appear next to
7987 some captured HTTP headers. It will then automatically appear in the
7988 logs, and it will be possible to extract it using sample fetch rules to
7989 feed it into headers or anything. The length should be limited given
7990 that this size will be allocated for each capture during the whole
Willy Tarreaua9083d02015-05-08 15:27:59 +02007991 session life. Please check section 7.3 (Fetching samples) and "capture
7992 request header" for more information.
Willy Tarreau18bf01e2014-06-13 16:18:52 +02007993
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007994 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +02007995 enables tracking of sticky counters from current connection. These
Willy Tarreau09448f72014-06-25 18:12:15 +02007996 rules do not stop evaluation and do not change default action. 3 sets
Willy Tarreaue9656522010-08-17 15:40:09 +02007997 of counters may be simultaneously tracked by the same connection. The
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007998 first "track-sc0" rule executed enables tracking of the counters of the
7999 specified table as the first set. The first "track-sc1" rule executed
Willy Tarreaue9656522010-08-17 15:40:09 +02008000 enables tracking of the counters of the specified table as the second
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008001 set. The first "track-sc2" rule executed enables tracking of the
8002 counters of the specified table as the third set. It is a recommended
8003 practice to use the first set of counters for the per-frontend counters
8004 and the second set for the per-backend ones. But this is just a
8005 guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008006
Willy Tarreaue9656522010-08-17 15:40:09 +02008007 These actions take one or two arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008008 <key> is mandatory, and is a sample expression rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +02008009 in section 7.3. It describes what elements of the incoming
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008010 request or connection will be analysed, extracted, combined,
8011 and used to select which table entry to update the counters.
8012 Note that "tcp-request connection" cannot use content-based
8013 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008014
Willy Tarreaue9656522010-08-17 15:40:09 +02008015 <table> is an optional table to be used instead of the default one,
8016 which is the stick-table declared in the current proxy. All
8017 the counters for the matches and updates for the key will
8018 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008019
Willy Tarreaue9656522010-08-17 15:40:09 +02008020 Once a "track-sc*" rule is executed, the key is looked up in the table
8021 and if it is not found, an entry is allocated for it. Then a pointer to
8022 that entry is kept during all the session's life, and this entry's
8023 counters are updated as often as possible, every time the session's
8024 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008025 Counters are only updated for events that happen after the tracking has
8026 been started. For example, connection counters will not be updated when
8027 tracking layer 7 information, since the connection event happens before
8028 layer7 information is extracted.
8029
Willy Tarreaue9656522010-08-17 15:40:09 +02008030 If the entry tracks concurrent connection counters, one connection is
8031 counted for as long as the entry is tracked, and the entry will not
8032 expire during that time. Tracking counters also provides a performance
8033 advantage over just checking the keys, because only one table lookup is
8034 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008035
Willy Tarreaue9656522010-08-17 15:40:09 +02008036 Note that the "if/unless" condition is optional. If no condition is set on
8037 the action, it is simply performed unconditionally. That can be useful for
8038 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008039
Willy Tarreaue9656522010-08-17 15:40:09 +02008040 Example: accept all connections from white-listed hosts, reject too fast
8041 connection without counting them, and track accepted connections.
8042 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008043
Willy Tarreaue9656522010-08-17 15:40:09 +02008044 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008045 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008046 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008047
Willy Tarreaue9656522010-08-17 15:40:09 +02008048 Example: accept all connections from white-listed hosts, count all other
8049 connections and reject too fast ones. This results in abusive ones
8050 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008051
Willy Tarreaue9656522010-08-17 15:40:09 +02008052 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008053 tcp-request connection track-sc0 src
8054 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008055
Willy Tarreau4f0d9192013-06-11 20:40:55 +02008056 Example: enable the PROXY protocol for traffic coming from all known proxies.
8057
8058 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
8059
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008060 See section 7 about ACL usage.
8061
Willy Tarreaue9656522010-08-17 15:40:09 +02008062 See also : "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008063
8064
Willy Tarreaue9656522010-08-17 15:40:09 +02008065tcp-request content <action> [{if | unless} <condition>]
8066 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +02008067 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02008068 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +02008069 Arguments :
8070 <action> defines the action to perform if the condition applies. Valid
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008071 actions include : "accept", "reject", "track-sc0", "track-sc1",
Thierry FOURNIER90da1912015-03-05 11:17:06 +01008072 "track-sc2", "capture" and "lua". See "tcp-request connection"
8073 above for their signification.
Willy Tarreau62644772008-07-16 18:36:06 +02008074
Willy Tarreaue9656522010-08-17 15:40:09 +02008075 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +02008076
Willy Tarreaue9656522010-08-17 15:40:09 +02008077 A request's contents can be analysed at an early stage of request processing
8078 called "TCP content inspection". During this stage, ACL-based rules are
8079 evaluated every time the request contents are updated, until either an
8080 "accept" or a "reject" rule matches, or the TCP request inspection delay
8081 expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +02008082
Willy Tarreaue9656522010-08-17 15:40:09 +02008083 The first difference between these rules and "tcp-request connection" rules
8084 is that "tcp-request content" rules can make use of contents to take a
8085 decision. Most often, these decisions will consider a protocol recognition or
8086 validity. The second difference is that content-based rules can be used in
Willy Tarreauf3338342014-01-28 21:40:28 +01008087 both frontends and backends. In case of HTTP keep-alive with the client, all
8088 tcp-request content rules are evaluated again, so haproxy keeps a record of
8089 what sticky counters were assigned by a "tcp-request connection" versus a
8090 "tcp-request content" rule, and flushes all the content-related ones after
8091 processing an HTTP request, so that they may be evaluated again by the rules
8092 being evaluated again for the next request. This is of particular importance
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008093 when the rule tracks some L7 information or when it is conditioned by an
Willy Tarreauf3338342014-01-28 21:40:28 +01008094 L7-based ACL, since tracking may change between requests.
Willy Tarreau62644772008-07-16 18:36:06 +02008095
Willy Tarreaue9656522010-08-17 15:40:09 +02008096 Content-based rules are evaluated in their exact declaration order. If no
8097 rule matches or if there is no rule, the default action is to accept the
8098 contents. There is no specific limit to the number of rules which may be
8099 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +02008100
Willy Tarreau18bf01e2014-06-13 16:18:52 +02008101 Four types of actions are supported :
8102 - accept : the request is accepted
8103 - reject : the request is rejected and the connection is closed
8104 - capture : the specified sample expression is captured
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008105 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Willy Tarreau62644772008-07-16 18:36:06 +02008106
Willy Tarreaue9656522010-08-17 15:40:09 +02008107 They have the same meaning as their counter-parts in "tcp-request connection"
8108 so please refer to that section for a complete description.
Willy Tarreau62644772008-07-16 18:36:06 +02008109
Willy Tarreauf3338342014-01-28 21:40:28 +01008110 While there is nothing mandatory about it, it is recommended to use the
8111 track-sc0 in "tcp-request connection" rules, track-sc1 for "tcp-request
8112 content" rules in the frontend, and track-sc2 for "tcp-request content"
8113 rules in the backend, because that makes the configuration more readable
8114 and easier to troubleshoot, but this is just a guideline and all counters
8115 may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +02008116
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008117 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +02008118 the action, it is simply performed unconditionally. That can be useful for
8119 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +02008120
Willy Tarreaue9656522010-08-17 15:40:09 +02008121 It is perfectly possible to match layer 7 contents with "tcp-request content"
Willy Tarreauc0239e02012-04-16 14:42:55 +02008122 rules, since HTTP-specific ACL matches are able to preliminarily parse the
8123 contents of a buffer before extracting the required data. If the buffered
8124 contents do not parse as a valid HTTP message, then the ACL does not match.
8125 The parser which is involved there is exactly the same as for all other HTTP
Willy Tarreauf3338342014-01-28 21:40:28 +01008126 processing, so there is no risk of parsing something differently. In an HTTP
8127 backend connected to from an HTTP frontend, it is guaranteed that HTTP
8128 contents will always be immediately present when the rule is evaluated first.
Willy Tarreau62644772008-07-16 18:36:06 +02008129
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008130 Tracking layer7 information is also possible provided that the information
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02008131 are present when the rule is processed. The rule processing engine is able to
8132 wait until the inspect delay expires when the data to be tracked is not yet
8133 available.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008134
Thierry FOURNIER90da1912015-03-05 11:17:06 +01008135 The "lua" keyword is followed by a Lua function name. It is used to run a Lua
8136 function if the action is executed. The single parameter is the name of the
8137 function to run. The prototype of the function is documented in the API
8138 documentation.
8139
Willy Tarreau62644772008-07-16 18:36:06 +02008140 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +02008141 # Accept HTTP requests containing a Host header saying "example.com"
8142 # and reject everything else.
8143 acl is_host_com hdr(Host) -i example.com
8144 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +02008145 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +02008146 tcp-request content reject
8147
8148 Example:
Willy Tarreau62644772008-07-16 18:36:06 +02008149 # reject SMTP connection if client speaks first
8150 tcp-request inspect-delay 30s
8151 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008152 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +02008153
8154 # Forward HTTPS connection only if client speaks
8155 tcp-request inspect-delay 30s
8156 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008157 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +02008158 tcp-request content reject
8159
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008160 Example:
8161 # Track the last IP from X-Forwarded-For
8162 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02008163 tcp-request content track-sc0 hdr(x-forwarded-for,-1)
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008164
8165 Example:
8166 # track request counts per "base" (concatenation of Host+URL)
8167 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02008168 tcp-request content track-sc0 base table req-rate
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008169
Willy Tarreaue9656522010-08-17 15:40:09 +02008170 Example: track per-frontend and per-backend counters, block abusers at the
8171 frontend when the backend detects abuse.
8172
8173 frontend http
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008174 # Use General Purpose Couter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +02008175 # protecting all our sites
8176 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008177 tcp-request connection track-sc0 src
8178 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +02008179 ...
8180 use_backend http_dynamic if { path_end .php }
8181
8182 backend http_dynamic
8183 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008184 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +02008185 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008186 acl click_too_fast sc1_http_req_rate gt 10
8187 acl mark_as_abuser sc0_inc_gpc0 gt 0
8188 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +02008189 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +02008190
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008191 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +02008192
Willy Tarreaue9656522010-08-17 15:40:09 +02008193 See also : "tcp-request connection", "tcp-request inspect-delay"
Willy Tarreau62644772008-07-16 18:36:06 +02008194
8195
8196tcp-request inspect-delay <timeout>
8197 Set the maximum allowed time to wait for data during content inspection
8198 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02008199 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +02008200 Arguments :
8201 <timeout> is the timeout value specified in milliseconds by default, but
8202 can be in any other unit if the number is suffixed by the unit,
8203 as explained at the top of this document.
8204
8205 People using haproxy primarily as a TCP relay are often worried about the
8206 risk of passing any type of protocol to a server without any analysis. In
8207 order to be able to analyze the request contents, we must first withhold
8208 the data then analyze them. This statement simply enables withholding of
8209 data for at most the specified amount of time.
8210
Willy Tarreaufb356202010-08-03 14:02:05 +02008211 TCP content inspection applies very early when a connection reaches a
8212 frontend, then very early when the connection is forwarded to a backend. This
8213 means that a connection may experience a first delay in the frontend and a
8214 second delay in the backend if both have tcp-request rules.
8215
Willy Tarreau62644772008-07-16 18:36:06 +02008216 Note that when performing content inspection, haproxy will evaluate the whole
8217 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008218 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +02008219 a last check is performed upon expiration, this time considering that the
Willy Tarreaud869b242009-03-15 14:43:58 +01008220 contents are definitive. If no delay is set, haproxy will not wait at all
8221 and will immediately apply a verdict based on the available information.
8222 Obviously this is unlikely to be very useful and might even be racy, so such
8223 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +02008224
8225 As soon as a rule matches, the request is released and continues as usual. If
8226 the timeout is reached and no rule matches, the default policy will be to let
8227 it pass through unaffected.
8228
8229 For most protocols, it is enough to set it to a few seconds, as most clients
8230 send the full request immediately upon connection. Add 3 or more seconds to
8231 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +01008232 to use large values, for instance to ensure that the client never talks
Willy Tarreau62644772008-07-16 18:36:06 +02008233 before the server (eg: SMTP), or to wait for a client to talk before passing
8234 data to the server (eg: SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +02008235 least the inspection delay, otherwise it will expire first. If the client
8236 closes the connection or if the buffer is full, the delay immediately expires
8237 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +02008238
Willy Tarreau55165fe2009-05-10 12:02:55 +02008239 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +02008240 "timeout client".
8241
8242
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008243tcp-response content <action> [{if | unless} <condition>]
8244 Perform an action on a session response depending on a layer 4-7 condition
8245 May be used in sections : defaults | frontend | listen | backend
8246 no | no | yes | yes
8247 Arguments :
8248 <action> defines the action to perform if the condition applies. Valid
Thierry FOURNIER90da1912015-03-05 11:17:06 +01008249 actions include : "accept", "close", "reject", "lua".
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008250
8251 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
8252
8253 Response contents can be analysed at an early stage of response processing
8254 called "TCP content inspection". During this stage, ACL-based rules are
8255 evaluated every time the response contents are updated, until either an
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02008256 "accept", "close" or a "reject" rule matches, or a TCP response inspection
8257 delay is set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008258
8259 Most often, these decisions will consider a protocol recognition or validity.
8260
8261 Content-based rules are evaluated in their exact declaration order. If no
8262 rule matches or if there is no rule, the default action is to accept the
8263 contents. There is no specific limit to the number of rules which may be
8264 inserted.
8265
8266 Two types of actions are supported :
8267 - accept :
8268 accepts the response if the condition is true (when used with "if")
8269 or false (when used with "unless"). The first such rule executed ends
8270 the rules evaluation.
8271
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02008272 - close :
8273 immediately closes the connection with the server if the condition is
8274 true (when used with "if"), or false (when used with "unless"). The
8275 first such rule executed ends the rules evaluation. The main purpose of
8276 this action is to force a connection to be finished between a client
8277 and a server after an exchange when the application protocol expects
8278 some long time outs to elapse first. The goal is to eliminate idle
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008279 connections which take significant resources on servers with certain
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02008280 protocols.
8281
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008282 - reject :
8283 rejects the response if the condition is true (when used with "if")
8284 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -04008285 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008286
8287 Note that the "if/unless" condition is optional. If no condition is set on
8288 the action, it is simply performed unconditionally. That can be useful for
8289 for changing the default action to a reject.
8290
Jamie Gloudonaaa21002012-08-25 00:18:33 -04008291 It is perfectly possible to match layer 7 contents with "tcp-response
8292 content" rules, but then it is important to ensure that a full response has
8293 been buffered, otherwise no contents will match. In order to achieve this,
8294 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008295 period.
8296
Thierry FOURNIER90da1912015-03-05 11:17:06 +01008297 The "lua" keyword is followed by a Lua function name. It is used to run a Lua
8298 function if the action is executed. The single parameter is the name of the
8299 function to run. The prototype of the function is documented in the API
8300 documentation.
8301
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008302 See section 7 about ACL usage.
8303
8304 See also : "tcp-request content", "tcp-response inspect-delay"
8305
8306
8307tcp-response inspect-delay <timeout>
8308 Set the maximum allowed time to wait for a response during content inspection
8309 May be used in sections : defaults | frontend | listen | backend
8310 no | no | yes | yes
8311 Arguments :
8312 <timeout> is the timeout value specified in milliseconds by default, but
8313 can be in any other unit if the number is suffixed by the unit,
8314 as explained at the top of this document.
8315
8316 See also : "tcp-response content", "tcp-request inspect-delay".
8317
8318
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01008319timeout check <timeout>
8320 Set additional check timeout, but only after a connection has been already
8321 established.
8322
8323 May be used in sections: defaults | frontend | listen | backend
8324 yes | no | yes | yes
8325 Arguments:
8326 <timeout> is the timeout value specified in milliseconds by default, but
8327 can be in any other unit if the number is suffixed by the unit,
8328 as explained at the top of this document.
8329
8330 If set, haproxy uses min("timeout connect", "inter") as a connect timeout
8331 for check and "timeout check" as an additional read timeout. The "min" is
8332 used so that people running with *very* long "timeout connect" (eg. those
8333 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +01008334 (Please also note that there is no valid reason to have such long connect
8335 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
8336 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01008337
8338 If "timeout check" is not set haproxy uses "inter" for complete check
8339 timeout (connect + read) exactly like all <1.3.15 version.
8340
8341 In most cases check request is much simpler and faster to handle than normal
8342 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +01008343 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01008344
8345 This parameter is specific to backends, but can be specified once for all in
8346 "defaults" sections. This is in fact one of the easiest solutions not to
8347 forget about it.
8348
Willy Tarreau41a340d2008-01-22 12:25:31 +01008349 See also: "timeout connect", "timeout queue", "timeout server",
8350 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01008351
8352
Willy Tarreau0ba27502007-12-24 16:55:16 +01008353timeout client <timeout>
8354timeout clitimeout <timeout> (deprecated)
8355 Set the maximum inactivity time on the client side.
8356 May be used in sections : defaults | frontend | listen | backend
8357 yes | yes | yes | no
8358 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01008359 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01008360 can be in any other unit if the number is suffixed by the unit,
8361 as explained at the top of this document.
8362
8363 The inactivity timeout applies when the client is expected to acknowledge or
8364 send data. In HTTP mode, this timeout is particularly important to consider
8365 during the first phase, when the client sends the request, and during the
8366 response while it is reading data sent by the server. The value is specified
8367 in milliseconds by default, but can be in any other unit if the number is
8368 suffixed by the unit, as specified at the top of this document. In TCP mode
8369 (and to a lesser extent, in HTTP mode), it is highly recommended that the
8370 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01008371 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01008372 losses by specifying timeouts that are slightly above multiples of 3 seconds
Willy Tarreauce887fd2012-05-12 12:50:00 +02008373 (eg: 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
8374 sessions (eg: WebSocket and HTTP), it's worth considering "timeout tunnel",
Willy Tarreau05cdd962014-05-10 14:30:07 +02008375 which overrides "timeout client" and "timeout server" for tunnels, as well as
8376 "timeout client-fin" for half-closed connections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01008377
8378 This parameter is specific to frontends, but can be specified once for all in
8379 "defaults" sections. This is in fact one of the easiest solutions not to
8380 forget about it. An unspecified timeout results in an infinite timeout, which
8381 is not recommended. Such a usage is accepted and works but reports a warning
8382 during startup because it may results in accumulation of expired sessions in
8383 the system if the system's timeouts are not configured either.
8384
8385 This parameter replaces the old, deprecated "clitimeout". It is recommended
8386 to use it to write new configurations. The form "timeout clitimeout" is
8387 provided only by backwards compatibility but its use is strongly discouraged.
8388
Willy Tarreauce887fd2012-05-12 12:50:00 +02008389 See also : "clitimeout", "timeout server", "timeout tunnel".
Willy Tarreau0ba27502007-12-24 16:55:16 +01008390
8391
Willy Tarreau05cdd962014-05-10 14:30:07 +02008392timeout client-fin <timeout>
8393 Set the inactivity timeout on the client side for half-closed connections.
8394 May be used in sections : defaults | frontend | listen | backend
8395 yes | yes | yes | no
8396 Arguments :
8397 <timeout> is the timeout value specified in milliseconds by default, but
8398 can be in any other unit if the number is suffixed by the unit,
8399 as explained at the top of this document.
8400
8401 The inactivity timeout applies when the client is expected to acknowledge or
8402 send data while one direction is already shut down. This timeout is different
8403 from "timeout client" in that it only applies to connections which are closed
8404 in one direction. This is particularly useful to avoid keeping connections in
8405 FIN_WAIT state for too long when clients do not disconnect cleanly. This
8406 problem is particularly common long connections such as RDP or WebSocket.
8407 Note that this timeout can override "timeout tunnel" when a connection shuts
8408 down in one direction.
8409
8410 This parameter is specific to frontends, but can be specified once for all in
8411 "defaults" sections. By default it is not set, so half-closed connections
8412 will use the other timeouts (timeout.client or timeout.tunnel).
8413
8414 See also : "timeout client", "timeout server-fin", and "timeout tunnel".
8415
8416
Willy Tarreau0ba27502007-12-24 16:55:16 +01008417timeout connect <timeout>
8418timeout contimeout <timeout> (deprecated)
8419 Set the maximum time to wait for a connection attempt to a server to succeed.
8420 May be used in sections : defaults | frontend | listen | backend
8421 yes | no | yes | yes
8422 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01008423 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01008424 can be in any other unit if the number is suffixed by the unit,
8425 as explained at the top of this document.
8426
8427 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01008428 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01008429 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01008430 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01008431 connect timeout also presets both queue and tarpit timeouts to the same value
8432 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +01008433
8434 This parameter is specific to backends, but can be specified once for all in
8435 "defaults" sections. This is in fact one of the easiest solutions not to
8436 forget about it. An unspecified timeout results in an infinite timeout, which
8437 is not recommended. Such a usage is accepted and works but reports a warning
8438 during startup because it may results in accumulation of failed sessions in
8439 the system if the system's timeouts are not configured either.
8440
8441 This parameter replaces the old, deprecated "contimeout". It is recommended
8442 to use it to write new configurations. The form "timeout contimeout" is
8443 provided only by backwards compatibility but its use is strongly discouraged.
8444
Willy Tarreau41a340d2008-01-22 12:25:31 +01008445 See also: "timeout check", "timeout queue", "timeout server", "contimeout",
8446 "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +01008447
8448
Willy Tarreaub16a5742010-01-10 14:46:16 +01008449timeout http-keep-alive <timeout>
8450 Set the maximum allowed time to wait for a new HTTP request to appear
8451 May be used in sections : defaults | frontend | listen | backend
8452 yes | yes | yes | yes
8453 Arguments :
8454 <timeout> is the timeout value specified in milliseconds by default, but
8455 can be in any other unit if the number is suffixed by the unit,
8456 as explained at the top of this document.
8457
8458 By default, the time to wait for a new request in case of keep-alive is set
8459 by "timeout http-request". However this is not always convenient because some
8460 people want very short keep-alive timeouts in order to release connections
8461 faster, and others prefer to have larger ones but still have short timeouts
8462 once the request has started to present itself.
8463
8464 The "http-keep-alive" timeout covers these needs. It will define how long to
8465 wait for a new HTTP request to start coming after a response was sent. Once
8466 the first byte of request has been seen, the "http-request" timeout is used
8467 to wait for the complete request to come. Note that empty lines prior to a
8468 new request do not refresh the timeout and are not counted as a new request.
8469
8470 There is also another difference between the two timeouts : when a connection
8471 expires during timeout http-keep-alive, no error is returned, the connection
8472 just closes. If the connection expires in "http-request" while waiting for a
8473 connection to complete, a HTTP 408 error is returned.
8474
8475 In general it is optimal to set this value to a few tens to hundreds of
8476 milliseconds, to allow users to fetch all objects of a page at once but
8477 without waiting for further clicks. Also, if set to a very small value (eg:
8478 1 millisecond) it will probably only accept pipelined requests but not the
8479 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +02008480 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +01008481
8482 If this parameter is not set, the "http-request" timeout applies, and if both
8483 are not set, "timeout client" still applies at the lower level. It should be
8484 set in the frontend to take effect, unless the frontend is in TCP mode, in
8485 which case the HTTP backend's timeout will be used.
8486
8487 See also : "timeout http-request", "timeout client".
8488
8489
Willy Tarreau036fae02008-01-06 13:24:40 +01008490timeout http-request <timeout>
8491 Set the maximum allowed time to wait for a complete HTTP request
8492 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +02008493 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +01008494 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01008495 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +01008496 can be in any other unit if the number is suffixed by the unit,
8497 as explained at the top of this document.
8498
8499 In order to offer DoS protection, it may be required to lower the maximum
8500 accepted time to receive a complete HTTP request without affecting the client
8501 timeout. This helps protecting against established connections on which
8502 nothing is sent. The client timeout cannot offer a good protection against
8503 this abuse because it is an inactivity timeout, which means that if the
8504 attacker sends one character every now and then, the timeout will not
8505 trigger. With the HTTP request timeout, no matter what speed the client
Willy Tarreau2705a612014-05-23 17:38:34 +02008506 types, the request will be aborted if it does not complete in time. When the
8507 timeout expires, an HTTP 408 response is sent to the client to inform it
8508 about the problem, and the connection is closed. The logs will report
8509 termination codes "cR". Some recent browsers are having problems with this
8510 standard, well-documented behaviour, so it might be needed to hide the 408
Willy Tarreau0f228a02015-05-01 15:37:53 +02008511 code using "option http-ignore-probes" or "errorfile 408 /dev/null". See
8512 more details in the explanations of the "cR" termination code in section 8.5.
Willy Tarreau036fae02008-01-06 13:24:40 +01008513
8514 Note that this timeout only applies to the header part of the request, and
8515 not to any data. As soon as the empty line is received, this timeout is not
Willy Tarreaub16a5742010-01-10 14:46:16 +01008516 used anymore. It is used again on keep-alive connections to wait for a second
8517 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +01008518
8519 Generally it is enough to set it to a few seconds, as most clients send the
8520 full request immediately upon connection. Add 3 or more seconds to cover TCP
8521 retransmits but that's all. Setting it to very low values (eg: 50 ms) will
8522 generally work on local networks as long as there are no packet losses. This
8523 will prevent people from sending bare HTTP requests using telnet.
8524
8525 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +02008526 chunk of the incoming request. It should be set in the frontend to take
8527 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
8528 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +01008529
Willy Tarreau0f228a02015-05-01 15:37:53 +02008530 See also : "errorfile", "http-ignore-probes", "timeout http-keep-alive", and
8531 "timeout client".
Willy Tarreau036fae02008-01-06 13:24:40 +01008532
Willy Tarreau844e3c52008-01-11 16:28:18 +01008533
8534timeout queue <timeout>
8535 Set the maximum time to wait in the queue for a connection slot to be free
8536 May be used in sections : defaults | frontend | listen | backend
8537 yes | no | yes | yes
8538 Arguments :
8539 <timeout> is the timeout value specified in milliseconds by default, but
8540 can be in any other unit if the number is suffixed by the unit,
8541 as explained at the top of this document.
8542
8543 When a server's maxconn is reached, connections are left pending in a queue
8544 which may be server-specific or global to the backend. In order not to wait
8545 indefinitely, a timeout is applied to requests pending in the queue. If the
8546 timeout is reached, it is considered that the request will almost never be
8547 served, so it is dropped and a 503 error is returned to the client.
8548
8549 The "timeout queue" statement allows to fix the maximum time for a request to
8550 be left pending in a queue. If unspecified, the same value as the backend's
8551 connection timeout ("timeout connect") is used, for backwards compatibility
8552 with older versions with no "timeout queue" parameter.
8553
8554 See also : "timeout connect", "contimeout".
8555
8556
8557timeout server <timeout>
8558timeout srvtimeout <timeout> (deprecated)
8559 Set the maximum inactivity time on the server side.
8560 May be used in sections : defaults | frontend | listen | backend
8561 yes | no | yes | yes
8562 Arguments :
8563 <timeout> is the timeout value specified in milliseconds by default, but
8564 can be in any other unit if the number is suffixed by the unit,
8565 as explained at the top of this document.
8566
8567 The inactivity timeout applies when the server is expected to acknowledge or
8568 send data. In HTTP mode, this timeout is particularly important to consider
8569 during the first phase of the server's response, when it has to send the
8570 headers, as it directly represents the server's processing time for the
8571 request. To find out what value to put there, it's often good to start with
8572 what would be considered as unacceptable response times, then check the logs
8573 to observe the response time distribution, and adjust the value accordingly.
8574
8575 The value is specified in milliseconds by default, but can be in any other
8576 unit if the number is suffixed by the unit, as specified at the top of this
8577 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
8578 recommended that the client timeout remains equal to the server timeout in
8579 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01008580 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01008581 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreauce887fd2012-05-12 12:50:00 +02008582 seconds (eg: 4 or 5 seconds minimum). If some long-lived sessions are mixed
8583 with short-lived sessions (eg: WebSocket and HTTP), it's worth considering
8584 "timeout tunnel", which overrides "timeout client" and "timeout server" for
8585 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +01008586
8587 This parameter is specific to backends, but can be specified once for all in
8588 "defaults" sections. This is in fact one of the easiest solutions not to
8589 forget about it. An unspecified timeout results in an infinite timeout, which
8590 is not recommended. Such a usage is accepted and works but reports a warning
8591 during startup because it may results in accumulation of expired sessions in
8592 the system if the system's timeouts are not configured either.
8593
8594 This parameter replaces the old, deprecated "srvtimeout". It is recommended
8595 to use it to write new configurations. The form "timeout srvtimeout" is
8596 provided only by backwards compatibility but its use is strongly discouraged.
8597
Willy Tarreauce887fd2012-05-12 12:50:00 +02008598 See also : "srvtimeout", "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +01008599
Willy Tarreau05cdd962014-05-10 14:30:07 +02008600
8601timeout server-fin <timeout>
8602 Set the inactivity timeout on the server side for half-closed connections.
8603 May be used in sections : defaults | frontend | listen | backend
8604 yes | no | yes | yes
8605 Arguments :
8606 <timeout> is the timeout value specified in milliseconds by default, but
8607 can be in any other unit if the number is suffixed by the unit,
8608 as explained at the top of this document.
8609
8610 The inactivity timeout applies when the server is expected to acknowledge or
8611 send data while one direction is already shut down. This timeout is different
8612 from "timeout server" in that it only applies to connections which are closed
8613 in one direction. This is particularly useful to avoid keeping connections in
8614 FIN_WAIT state for too long when a remote server does not disconnect cleanly.
8615 This problem is particularly common long connections such as RDP or WebSocket.
8616 Note that this timeout can override "timeout tunnel" when a connection shuts
8617 down in one direction. This setting was provided for completeness, but in most
8618 situations, it should not be needed.
8619
8620 This parameter is specific to backends, but can be specified once for all in
8621 "defaults" sections. By default it is not set, so half-closed connections
8622 will use the other timeouts (timeout.server or timeout.tunnel).
8623
8624 See also : "timeout client-fin", "timeout server", and "timeout tunnel".
8625
Willy Tarreau844e3c52008-01-11 16:28:18 +01008626
8627timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +01008628 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +01008629 May be used in sections : defaults | frontend | listen | backend
8630 yes | yes | yes | yes
8631 Arguments :
8632 <timeout> is the tarpit duration specified in milliseconds by default, but
8633 can be in any other unit if the number is suffixed by the unit,
8634 as explained at the top of this document.
8635
8636 When a connection is tarpitted using "reqtarpit", it is maintained open with
8637 no activity for a certain amount of time, then closed. "timeout tarpit"
8638 defines how long it will be maintained open.
8639
8640 The value is specified in milliseconds by default, but can be in any other
8641 unit if the number is suffixed by the unit, as specified at the top of this
8642 document. If unspecified, the same value as the backend's connection timeout
8643 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +01008644 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +01008645
8646 See also : "timeout connect", "contimeout".
8647
8648
Willy Tarreauce887fd2012-05-12 12:50:00 +02008649timeout tunnel <timeout>
8650 Set the maximum inactivity time on the client and server side for tunnels.
8651 May be used in sections : defaults | frontend | listen | backend
8652 yes | no | yes | yes
8653 Arguments :
8654 <timeout> is the timeout value specified in milliseconds by default, but
8655 can be in any other unit if the number is suffixed by the unit,
8656 as explained at the top of this document.
8657
Jamie Gloudonaaa21002012-08-25 00:18:33 -04008658 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +02008659 between a client and a server, and the connection remains inactive in both
8660 directions. This timeout supersedes both the client and server timeouts once
8661 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
8662 analyser remains attached to either connection (eg: tcp content rules are
8663 accepted). In HTTP, this timeout is used when a connection is upgraded (eg:
8664 when switching to the WebSocket protocol, or forwarding a CONNECT request
8665 to a proxy), or after the first response when no keepalive/close option is
8666 specified.
8667
Willy Tarreau05cdd962014-05-10 14:30:07 +02008668 Since this timeout is usually used in conjunction with long-lived connections,
8669 it usually is a good idea to also set "timeout client-fin" to handle the
8670 situation where a client suddenly disappears from the net and does not
8671 acknowledge a close, or sends a shutdown and does not acknowledge pending
8672 data anymore. This can happen in lossy networks where firewalls are present,
8673 and is detected by the presence of large amounts of sessions in a FIN_WAIT
8674 state.
8675
Willy Tarreauce887fd2012-05-12 12:50:00 +02008676 The value is specified in milliseconds by default, but can be in any other
8677 unit if the number is suffixed by the unit, as specified at the top of this
8678 document. Whatever the expected normal idle time, it is a good practice to
8679 cover at least one or several TCP packet losses by specifying timeouts that
8680 are slightly above multiples of 3 seconds (eg: 4 or 5 seconds minimum).
8681
8682 This parameter is specific to backends, but can be specified once for all in
8683 "defaults" sections. This is in fact one of the easiest solutions not to
8684 forget about it.
8685
8686 Example :
8687 defaults http
8688 option http-server-close
8689 timeout connect 5s
8690 timeout client 30s
Willy Tarreau05cdd962014-05-10 14:30:07 +02008691 timeout client-fin 30s
Willy Tarreauce887fd2012-05-12 12:50:00 +02008692 timeout server 30s
8693 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
8694
Willy Tarreau05cdd962014-05-10 14:30:07 +02008695 See also : "timeout client", "timeout client-fin", "timeout server".
Willy Tarreauce887fd2012-05-12 12:50:00 +02008696
8697
Willy Tarreau844e3c52008-01-11 16:28:18 +01008698transparent (deprecated)
8699 Enable client-side transparent proxying
8700 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01008701 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01008702 Arguments : none
8703
8704 This keyword was introduced in order to provide layer 7 persistence to layer
8705 3 load balancers. The idea is to use the OS's ability to redirect an incoming
8706 connection for a remote address to a local process (here HAProxy), and let
8707 this process know what address was initially requested. When this option is
8708 used, sessions without cookies will be forwarded to the original destination
8709 IP address of the incoming request (which should match that of another
8710 equipment), while requests with cookies will still be forwarded to the
8711 appropriate server.
8712
8713 The "transparent" keyword is deprecated, use "option transparent" instead.
8714
8715 Note that contrary to a common belief, this option does NOT make HAProxy
8716 present the client's IP to the server when establishing the connection.
8717
Willy Tarreau844e3c52008-01-11 16:28:18 +01008718 See also: "option transparent"
8719
William Lallemanda73203e2012-03-12 12:48:57 +01008720unique-id-format <string>
8721 Generate a unique ID for each request.
8722 May be used in sections : defaults | frontend | listen | backend
8723 yes | yes | yes | no
8724 Arguments :
8725 <string> is a log-format string.
8726
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008727 This keyword creates a ID for each request using the custom log format. A
8728 unique ID is useful to trace a request passing through many components of
8729 a complex infrastructure. The newly created ID may also be logged using the
8730 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +01008731
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008732 The format should be composed from elements that are guaranteed to be
8733 unique when combined together. For instance, if multiple haproxy instances
8734 are involved, it might be important to include the node name. It is often
8735 needed to log the incoming connection's source and destination addresses
8736 and ports. Note that since multiple requests may be performed over the same
8737 connection, including a request counter may help differentiate them.
8738 Similarly, a timestamp may protect against a rollover of the counter.
8739 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +01008740
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008741 It is recommended to use hexadecimal notation for many fields since it
8742 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +01008743
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008744 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01008745
Julien Vehentf21be322014-03-07 08:27:34 -05008746 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +01008747
8748 will generate:
8749
8750 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
8751
8752 See also: "unique-id-header"
8753
8754unique-id-header <name>
8755 Add a unique ID header in the HTTP request.
8756 May be used in sections : defaults | frontend | listen | backend
8757 yes | yes | yes | no
8758 Arguments :
8759 <name> is the name of the header.
8760
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008761 Add a unique-id header in the HTTP request sent to the server, using the
8762 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +01008763
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008764 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01008765
Julien Vehentf21be322014-03-07 08:27:34 -05008766 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +01008767 unique-id-header X-Unique-ID
8768
8769 will generate:
8770
8771 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
8772
8773 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +01008774
Willy Tarreauf51658d2014-04-23 01:21:56 +02008775use_backend <backend> [{if | unless} <condition>]
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02008776 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +01008777 May be used in sections : defaults | frontend | listen | backend
8778 no | yes | yes | no
8779 Arguments :
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01008780 <backend> is the name of a valid backend or "listen" section, or a
8781 "log-format" string resolving to a backend name.
Willy Tarreau844e3c52008-01-11 16:28:18 +01008782
Willy Tarreauf51658d2014-04-23 01:21:56 +02008783 <condition> is a condition composed of ACLs, as described in section 7. If
8784 it is omitted, the rule is unconditionally applied.
Willy Tarreau844e3c52008-01-11 16:28:18 +01008785
8786 When doing content-switching, connections arrive on a frontend and are then
8787 dispatched to various backends depending on a number of conditions. The
8788 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02008789 "use_backend" keyword. While it is normally used with HTTP processing, it can
8790 also be used in pure TCP, either without content using stateless ACLs (eg:
8791 source address validation) or combined with a "tcp-request" rule to wait for
8792 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +01008793
8794 There may be as many "use_backend" rules as desired. All of these rules are
8795 evaluated in their declaration order, and the first one which matches will
8796 assign the backend.
8797
8798 In the first form, the backend will be used if the condition is met. In the
8799 second form, the backend will be used if the condition is not met. If no
8800 condition is valid, the backend defined with "default_backend" will be used.
8801 If no default backend is defined, either the servers in the same section are
8802 used (in case of a "listen" section) or, in case of a frontend, no server is
8803 used and a 503 service unavailable response is returned.
8804
Willy Tarreau51aecc72009-07-12 09:47:04 +02008805 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008806 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +02008807 and backend processing will immediately follow, or the backend will wait for
8808 a complete HTTP request to get in. This feature is useful when a frontend
8809 must decode several protocols on a unique port, one of them being HTTP.
8810
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01008811 When <backend> is a simple name, it is resolved at configuration time, and an
8812 error is reported if the specified backend does not exist. If <backend> is
8813 a log-format string instead, no check may be done at configuration time, so
8814 the backend name is resolved dynamically at run time. If the resulting
8815 backend name does not correspond to any valid backend, no other rule is
8816 evaluated, and the default_backend directive is applied instead. Note that
8817 when using dynamic backend names, it is highly recommended to use a prefix
8818 that no other backend uses in order to ensure that an unauthorized backend
8819 cannot be forced from the request.
8820
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008821 It is worth mentioning that "use_backend" rules with an explicit name are
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01008822 used to detect the association between frontends and backends to compute the
8823 backend's "fullconn" setting. This cannot be done for dynamic names.
8824
8825 See also: "default_backend", "tcp-request", "fullconn", "log-format", and
8826 section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +01008827
Willy Tarreau036fae02008-01-06 13:24:40 +01008828
Willy Tarreau4a5cade2012-04-05 21:09:48 +02008829use-server <server> if <condition>
8830use-server <server> unless <condition>
8831 Only use a specific server if/unless an ACL-based condition is matched.
8832 May be used in sections : defaults | frontend | listen | backend
8833 no | no | yes | yes
8834 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008835 <server> is the name of a valid server in the same backend section.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02008836
8837 <condition> is a condition composed of ACLs, as described in section 7.
8838
8839 By default, connections which arrive to a backend are load-balanced across
8840 the available servers according to the configured algorithm, unless a
8841 persistence mechanism such as a cookie is used and found in the request.
8842
8843 Sometimes it is desirable to forward a particular request to a specific
8844 server without having to declare a dedicated backend for this server. This
8845 can be achieved using the "use-server" rules. These rules are evaluated after
8846 the "redirect" rules and before evaluating cookies, and they have precedence
8847 on them. There may be as many "use-server" rules as desired. All of these
8848 rules are evaluated in their declaration order, and the first one which
8849 matches will assign the server.
8850
8851 If a rule designates a server which is down, and "option persist" is not used
8852 and no force-persist rule was validated, it is ignored and evaluation goes on
8853 with the next rules until one matches.
8854
8855 In the first form, the server will be used if the condition is met. In the
8856 second form, the server will be used if the condition is not met. If no
8857 condition is valid, the processing continues and the server will be assigned
8858 according to other persistence mechanisms.
8859
8860 Note that even if a rule is matched, cookie processing is still performed but
8861 does not assign the server. This allows prefixed cookies to have their prefix
8862 stripped.
8863
8864 The "use-server" statement works both in HTTP and TCP mode. This makes it
8865 suitable for use with content-based inspection. For instance, a server could
8866 be selected in a farm according to the TLS SNI field. And if these servers
8867 have their weight set to zero, they will not be used for other traffic.
8868
8869 Example :
8870 # intercept incoming TLS requests based on the SNI field
8871 use-server www if { req_ssl_sni -i www.example.com }
8872 server www 192.168.0.1:443 weight 0
8873 use-server mail if { req_ssl_sni -i mail.example.com }
8874 server mail 192.168.0.1:587 weight 0
8875 use-server imap if { req_ssl_sni -i imap.example.com }
8876 server mail 192.168.0.1:993 weight 0
8877 # all the rest is forwarded to this server
8878 server default 192.168.0.2:443 check
8879
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008880 See also: "use_backend", section 5 about server and section 7 about ACLs.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02008881
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008882
88835. Bind and Server options
8884--------------------------
8885
8886The "bind", "server" and "default-server" keywords support a number of settings
8887depending on some build options and on the system HAProxy was built on. These
8888settings generally each consist in one word sometimes followed by a value,
8889written on the same line as the "bind" or "server" line. All these options are
8890described in this section.
8891
8892
88935.1. Bind options
8894-----------------
8895
8896The "bind" keyword supports a certain number of settings which are all passed
8897as arguments on the same line. The order in which those arguments appear makes
8898no importance, provided that they appear after the bind address. All of these
8899parameters are optional. Some of them consist in a single words (booleans),
8900while other ones expect a value after them. In this case, the value must be
8901provided immediately after the setting name.
8902
8903The currently supported settings are the following ones.
8904
8905accept-proxy
8906 Enforces the use of the PROXY protocol over any connection accepted by any of
Willy Tarreau77992672014-06-14 11:06:17 +02008907 the sockets declared on the same line. Versions 1 and 2 of the PROXY protocol
8908 are supported and correctly detected. The PROXY protocol dictates the layer
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008909 3/4 addresses of the incoming connection to be used everywhere an address is
8910 used, with the only exception of "tcp-request connection" rules which will
8911 only see the real connection address. Logs will reflect the addresses
8912 indicated in the protocol, unless it is violated, in which case the real
8913 address will still be used. This keyword combined with support from external
8914 components can be used as an efficient and reliable alternative to the
8915 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +02008916 usable. See also "tcp-request connection expect-proxy" for a finer-grained
8917 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008918
Willy Tarreauab861d32013-04-02 02:30:41 +02008919alpn <protocols>
8920 This enables the TLS ALPN extension and advertises the specified protocol
8921 list as supported on top of ALPN. The protocol list consists in a comma-
8922 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
8923 quotes). This requires that the SSL library is build with support for TLS
8924 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
8925 initial NPN extension.
8926
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008927backlog <backlog>
8928 Sets the socket's backlog to this value. If unspecified, the frontend's
8929 backlog is used instead, which generally defaults to the maxconn value.
8930
Emeric Brun7fb34422012-09-28 15:26:15 +02008931ecdhe <named curve>
8932 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +01008933 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
8934 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +02008935
Emeric Brunfd33a262012-10-11 16:28:27 +02008936ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +02008937 This setting is only available when support for OpenSSL was built in. It
8938 designates a PEM file from which to load CA certificates used to verify
8939 client's certificate.
8940
Emeric Brunb6dc9342012-09-28 17:55:37 +02008941ca-ignore-err [all|<errorID>,...]
8942 This setting is only available when support for OpenSSL was built in.
8943 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
8944 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
8945 error is ignored.
8946
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008947ciphers <ciphers>
8948 This setting is only available when support for OpenSSL was built in. It sets
8949 the string describing the list of cipher algorithms ("cipher suite") that are
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008950 negotiated during the SSL/TLS handshake. The format of the string is defined
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008951 in "man 1 ciphers" from OpenSSL man pages, and can be for instance a string
8952 such as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes).
8953
Emeric Brunfd33a262012-10-11 16:28:27 +02008954crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +02008955 This setting is only available when support for OpenSSL was built in. It
8956 designates a PEM file from which to load certificate revocation list used
8957 to verify client's certificate.
8958
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008959crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +00008960 This setting is only available when support for OpenSSL was built in. It
8961 designates a PEM file containing both the required certificates and any
8962 associated private keys. This file can be built by concatenating multiple
8963 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
8964 requires an intermediate certificate, this can also be concatenated into this
8965 file.
8966
8967 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
8968 are loaded.
8969
8970 If a directory name is used instead of a PEM file, then all files found in
Cyril Bonté3180f7b2015-01-25 00:16:08 +01008971 that directory will be loaded in alphabetic order unless their name ends with
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +01008972 '.issuer', '.ocsp' or '.sctl' (reserved extensions). This directive may be
8973 specified multiple times in order to load certificates from multiple files or
8974 directories. The certificates will be presented to clients who provide a
8975 valid TLS Server Name Indication field matching one of their CN or alt
8976 subjects. Wildcards are supported, where a wildcard character '*' is used
8977 instead of the first hostname component (eg: *.example.org matches
8978 www.example.org but not www.sub.example.org).
Alex Davies0fbf0162013-03-02 16:04:50 +00008979
8980 If no SNI is provided by the client or if the SSL library does not support
8981 TLS extensions, or if the client provides an SNI hostname which does not
8982 match any certificate, then the first loaded certificate will be presented.
8983 This means that when loading certificates from a directory, it is highly
Cyril Bonté3180f7b2015-01-25 00:16:08 +01008984 recommended to load the default one first as a file or to ensure that it will
8985 always be the first one in the directory.
Alex Davies0fbf0162013-03-02 16:04:50 +00008986
Emeric Brune032bfa2012-09-28 13:01:45 +02008987 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008988
Alex Davies0fbf0162013-03-02 16:04:50 +00008989 Some CAs (such as Godaddy) offer a drop down list of server types that do not
8990 include HAProxy when obtaining a certificate. If this happens be sure to
Godbach8bf60a12014-04-21 21:42:41 +08008991 choose a webserver that the CA believes requires an intermediate CA (for
Alex Davies0fbf0162013-03-02 16:04:50 +00008992 Godaddy, selection Apache Tomcat will get the correct bundle, but many
8993 others, e.g. nginx, result in a wrong bundle that will not work for some
8994 clients).
8995
Emeric Brun4147b2e2014-06-16 18:36:30 +02008996 For each PEM file, haproxy checks for the presence of file at the same path
8997 suffixed by ".ocsp". If such file is found, support for the TLS Certificate
8998 Status Request extension (also known as "OCSP stapling") is automatically
8999 enabled. The content of this file is optional. If not empty, it must contain
9000 a valid OCSP Response in DER format. In order to be valid an OCSP Response
9001 must comply with the following rules: it has to indicate a good status,
9002 it has to be a single response for the certificate of the PEM file, and it
9003 has to be valid at the moment of addition. If these rules are not respected
9004 the OCSP Response is ignored and a warning is emitted. In order to identify
9005 which certificate an OCSP Response applies to, the issuer's certificate is
9006 necessary. If the issuer's certificate is not found in the PEM file, it will
9007 be loaded from a file at the same path as the PEM file suffixed by ".issuer"
9008 if it exists otherwise it will fail with an error.
9009
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +01009010 For each PEM file, haproxy also checks for the presence of file at the same
9011 path suffixed by ".sctl". If such file is found, support for Certificate
9012 Transparency (RFC6962) TLS extension is enabled. The file must contain a
9013 valid Signed Certificate Timestamp List, as described in RFC. File is parsed
9014 to check basic syntax, but no signatures are verified.
9015
Emeric Brunb6dc9342012-09-28 17:55:37 +02009016crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +00009017 This setting is only available when support for OpenSSL was built in. Sets a
9018 comma separated list of errorIDs to ignore during verify at depth == 0. If
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009019 set to 'all', all errors are ignored. SSL handshake is not aborted if an error
Alex Davies0fbf0162013-03-02 16:04:50 +00009020 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +02009021
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009022crt-list <file>
9023 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02009024 designates a list of PEM file with an optional list of SNI filter per
9025 certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009026
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02009027 <crtfile> [[!]<snifilter> ...]
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009028
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02009029 Wildcards are supported in the SNI filter. Negative filter are also supported,
9030 only useful in combination with a wildcard filter to exclude a particular SNI.
9031 The certificates will be presented to clients who provide a valid TLS Server
9032 Name Indication field matching one of the SNI filters. If no SNI filter is
9033 specified, the CN and alt subjects are used. This directive may be specified
9034 multiple times. See the "crt" option for more information. The default
9035 certificate is still needed to meet OpenSSL expectations. If it is not used,
9036 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009037
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009038defer-accept
9039 Is an optional keyword which is supported only on certain Linux kernels. It
9040 states that a connection will only be accepted once some data arrive on it,
9041 or at worst after the first retransmit. This should be used only on protocols
9042 for which the client talks first (eg: HTTP). It can slightly improve
9043 performance by ensuring that most of the request is already available when
9044 the connection is accepted. On the other hand, it will not be able to detect
9045 connections which don't talk. It is important to note that this option is
9046 broken in all kernels up to 2.6.31, as the connection is never accepted until
9047 the client talks. This can cause issues with front firewalls which would see
9048 an established connection while the proxy will only see it in SYN_RECV. This
9049 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
9050
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009051force-sslv3
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009052 This option enforces use of SSLv3 only on SSL connections instantiated from
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009053 this listener. SSLv3 is generally less expensive than the TLS counterparts
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009054 for high connection rates. This option is also available on global statement
9055 "ssl-default-bind-options". See also "no-tlsv*" and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009056
9057force-tlsv10
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009058 This option enforces use of TLSv1.0 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009059 this listener. This option is also available on global statement
9060 "ssl-default-bind-options". See also "no-tlsv*" and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009061
9062force-tlsv11
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009063 This option enforces use of TLSv1.1 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009064 this listener. This option is also available on global statement
9065 "ssl-default-bind-options". See also "no-tlsv*", and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009066
9067force-tlsv12
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009068 This option enforces use of TLSv1.2 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009069 this listener. This option is also available on global statement
9070 "ssl-default-bind-options". See also "no-tlsv*", and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009071
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009072gid <gid>
9073 Sets the group of the UNIX sockets to the designated system gid. It can also
9074 be set by default in the global section's "unix-bind" statement. Note that
9075 some platforms simply ignore this. This setting is equivalent to the "group"
9076 setting except that the group ID is used instead of its name. This setting is
9077 ignored by non UNIX sockets.
9078
9079group <group>
9080 Sets the group of the UNIX sockets to the designated system group. It can
9081 also be set by default in the global section's "unix-bind" statement. Note
9082 that some platforms simply ignore this. This setting is equivalent to the
9083 "gid" setting except that the group name is used instead of its gid. This
9084 setting is ignored by non UNIX sockets.
9085
9086id <id>
9087 Fixes the socket ID. By default, socket IDs are automatically assigned, but
9088 sometimes it is more convenient to fix them to ease monitoring. This value
9089 must be strictly positive and unique within the listener/frontend. This
9090 option can only be used when defining only a single socket.
9091
9092interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +01009093 Restricts the socket to a specific interface. When specified, only packets
9094 received from that particular interface are processed by the socket. This is
9095 currently only supported on Linux. The interface must be a primary system
9096 interface, not an aliased interface. It is also possible to bind multiple
9097 frontends to the same address if they are bound to different interfaces. Note
9098 that binding to a network interface requires root privileges. This parameter
9099 is only compatible with TCPv4/TCPv6 sockets.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009100
Willy Tarreauabb175f2012-09-24 12:43:26 +02009101level <level>
9102 This setting is used with the stats sockets only to restrict the nature of
9103 the commands that can be issued on the socket. It is ignored by other
9104 sockets. <level> can be one of :
9105 - "user" is the least privileged level ; only non-sensitive stats can be
9106 read, and no change is allowed. It would make sense on systems where it
9107 is not easy to restrict access to the socket.
9108 - "operator" is the default level and fits most common uses. All data can
9109 be read, and only non-sensitive changes are permitted (eg: clear max
9110 counters).
9111 - "admin" should be used with care, as everything is permitted (eg: clear
9112 all counters).
9113
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009114maxconn <maxconn>
9115 Limits the sockets to this number of concurrent connections. Extraneous
9116 connections will remain in the system's backlog until a connection is
9117 released. If unspecified, the limit will be the same as the frontend's
9118 maxconn. Note that in case of port ranges or multiple addresses, the same
9119 value will be applied to each socket. This setting enables different
9120 limitations on expensive sockets, for instance SSL entries which may easily
9121 eat all memory.
9122
9123mode <mode>
9124 Sets the octal mode used to define access permissions on the UNIX socket. It
9125 can also be set by default in the global section's "unix-bind" statement.
9126 Note that some platforms simply ignore this. This setting is ignored by non
9127 UNIX sockets.
9128
9129mss <maxseg>
9130 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
9131 connections. This can be used to force a lower MSS for certain specific
9132 ports, for instance for connections passing through a VPN. Note that this
9133 relies on a kernel feature which is theoretically supported under Linux but
9134 was buggy in all versions prior to 2.6.28. It may or may not work on other
9135 operating systems. It may also not change the advertised value but change the
9136 effective size of outgoing segments. The commonly advertised value for TCPv4
9137 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
9138 positive, it will be used as the advertised MSS. If it is negative, it will
9139 indicate by how much to reduce the incoming connection's advertised MSS for
9140 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
9141
9142name <name>
9143 Sets an optional name for these sockets, which will be reported on the stats
9144 page.
9145
9146nice <nice>
9147 Sets the 'niceness' of connections initiated from the socket. Value must be
9148 in the range -1024..1024 inclusive, and defaults to zero. Positive values
9149 means that such connections are more friendly to others and easily offer
9150 their place in the scheduler. On the opposite, negative values mean that
9151 connections want to run with a higher priority than others. The difference
9152 only happens under high loads when the system is close to saturation.
9153 Negative values are appropriate for low-latency or administration services,
9154 and high values are generally recommended for CPU intensive tasks such as SSL
9155 processing or bulk transfers which are less sensible to latency. For example,
9156 it may make sense to use a positive value for an SMTP socket and a negative
9157 one for an RDP socket.
9158
Emeric Brun9b3009b2012-10-05 11:55:06 +02009159no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009160 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009161 disables support for SSLv3 on any sockets instantiated from the listener when
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009162 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009163 be enabled using any configuration option. This option is also available on
9164 global statement "ssl-default-bind-options". See also "force-tls*",
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009165 and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009166
Emeric Brun90ad8722012-10-02 14:00:59 +02009167no-tls-tickets
9168 This setting is only available when support for OpenSSL was built in. It
9169 disables the stateless session resumption (RFC 5077 TLS Ticket
9170 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009171 session resumption is more expensive in CPU usage. This option is also
9172 available on global statement "ssl-default-bind-options".
Emeric Brun90ad8722012-10-02 14:00:59 +02009173
Emeric Brun9b3009b2012-10-05 11:55:06 +02009174no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009175 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009176 disables support for TLSv1.0 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009177 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009178 cannot be enabled using any configuration option. This option is also
9179 available on global statement "ssl-default-bind-options". See also
9180 "force-tlsv*", and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009181
Emeric Brun9b3009b2012-10-05 11:55:06 +02009182no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +02009183 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009184 disables support for TLSv1.1 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009185 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009186 cannot be enabled using any configuration option. This option is also
9187 available on global statement "ssl-default-bind-options". See also
9188 "force-tlsv*", and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +02009189
Emeric Brun9b3009b2012-10-05 11:55:06 +02009190no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +02009191 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009192 disables support for TLSv1.2 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009193 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009194 cannot be enabled using any configuration option. This option is also
9195 available on global statement "ssl-default-bind-options". See also
9196 "force-tlsv*", and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +02009197
Willy Tarreau6c9a3d52012-10-18 18:57:14 +02009198npn <protocols>
9199 This enables the NPN TLS extension and advertises the specified protocol list
9200 as supported on top of NPN. The protocol list consists in a comma-delimited
9201 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
9202 This requires that the SSL library is build with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +02009203 enabled (check with haproxy -vv). Note that the NPN extension has been
9204 replaced with the ALPN extension (see the "alpn" keyword).
Willy Tarreau6c9a3d52012-10-18 18:57:14 +02009205
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02009206process [ all | odd | even | <number 1-64>[-<number 1-64>] ]
9207 This restricts the list of processes on which this listener is allowed to
9208 run. It does not enforce any process but eliminates those which do not match.
9209 If the frontend uses a "bind-process" setting, the intersection between the
9210 two is applied. If in the end the listener is not allowed to run on any
9211 remaining process, a warning is emitted, and the listener will either run on
9212 the first process of the listener if a single process was specified, or on
9213 all of its processes if multiple processes were specified. For the unlikely
Willy Tarreauae302532014-05-07 19:22:24 +02009214 case where several ranges are needed, this directive may be repeated. The
9215 main purpose of this directive is to be used with the stats sockets and have
9216 one different socket per process. The second purpose is to have multiple bind
9217 lines sharing the same IP:port but not the same process in a listener, so
9218 that the system can distribute the incoming connections into multiple queues
9219 and allow a smoother inter-process load balancing. Currently Linux 3.9 and
9220 above is known for supporting this. See also "bind-process" and "nbproc".
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02009221
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009222ssl
9223 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009224 enables SSL deciphering on connections instantiated from this listener. A
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009225 certificate is necessary (see "crt" above). All contents in the buffers will
9226 appear in clear text, so that ACLs and HTTP processing will only have access
9227 to deciphered contents.
9228
Emmanuel Hocdet65623372013-01-24 17:17:15 +01009229strict-sni
9230 This setting is only available when support for OpenSSL was built in. The
9231 SSL/TLS negotiation is allow only if the client provided an SNI which match
9232 a certificate. The default certificate is not used.
9233 See the "crt" option for more information.
9234
Willy Tarreau2af207a2015-02-04 00:45:58 +01009235tcp-ut <delay>
9236 Sets the TCP User Timeout for all incoming connections instanciated from this
9237 listening socket. This option is available on Linux since version 2.6.37. It
9238 allows haproxy to configure a timeout for sockets which contain data not
9239 receiving an acknoledgement for the configured delay. This is especially
9240 useful on long-lived connections experiencing long idle periods such as
9241 remote terminals or database connection pools, where the client and server
9242 timeouts must remain high to allow a long period of idle, but where it is
9243 important to detect that the client has disappeared in order to release all
9244 resources associated with its connection (and the server's session). The
9245 argument is a delay expressed in milliseconds by default. This only works
9246 for regular TCP connections, and is ignored for other protocols.
9247
Willy Tarreau1c862c52012-10-05 16:21:00 +02009248tfo
Lukas Tribus0defb902013-02-13 23:35:39 +01009249 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +02009250 enables TCP Fast Open on the listening socket, which means that clients which
9251 support this feature will be able to send a request and receive a response
9252 during the 3-way handshake starting from second connection, thus saving one
9253 round-trip after the first connection. This only makes sense with protocols
9254 that use high connection rates and where each round trip matters. This can
9255 possibly cause issues with many firewalls which do not accept data on SYN
9256 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +02009257 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
9258 need to build HAProxy with USE_TFO=1 if your libc doesn't define
9259 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +02009260
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +01009261tls-ticket-keys <keyfile>
9262 Sets the TLS ticket keys file to load the keys from. The keys need to be 48
9263 bytes long, encoded with base64 (ex. openssl rand -base64 48). Number of keys
9264 is specified by the TLS_TICKETS_NO build option (default 3) and at least as
9265 many keys need to be present in the file. Last TLS_TICKETS_NO keys will be
9266 used for decryption and the penultimate one for encryption. This enables easy
9267 key rotation by just appending new key to the file and reloading the process.
9268 Keys must be periodically rotated (ex. every 12h) or Perfect Forward Secrecy
9269 is compromised. It is also a good idea to keep the keys off any permanent
9270 storage such as hard drives (hint: use tmpfs and don't swap those files).
9271 Lifetime hint can be changed using tune.ssl.timeout.
9272
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009273transparent
9274 Is an optional keyword which is supported only on certain Linux kernels. It
9275 indicates that the addresses will be bound even if they do not belong to the
9276 local machine, and that packets targeting any of these addresses will be
9277 intercepted just as if the addresses were locally configured. This normally
9278 requires that IP forwarding is enabled. Caution! do not use this with the
9279 default address '*', as it would redirect any traffic for the specified port.
9280 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
9281 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
9282 kernel version. Some distribution kernels include backports of the feature,
9283 so check for support with your vendor.
9284
Willy Tarreau77e3af92012-11-24 15:07:23 +01009285v4v6
9286 Is an optional keyword which is supported only on most recent systems
9287 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
9288 and IPv6 when it uses the default address. Doing so is sometimes necessary
9289 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009290 sockets, and is overridden by the "v6only" option.
Willy Tarreau77e3af92012-11-24 15:07:23 +01009291
Willy Tarreau9b6700f2012-11-24 11:55:28 +01009292v6only
9293 Is an optional keyword which is supported only on most recent systems
9294 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
9295 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +01009296 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
9297 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +01009298
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009299uid <uid>
9300 Sets the owner of the UNIX sockets to the designated system uid. It can also
9301 be set by default in the global section's "unix-bind" statement. Note that
9302 some platforms simply ignore this. This setting is equivalent to the "user"
9303 setting except that the user numeric ID is used instead of its name. This
9304 setting is ignored by non UNIX sockets.
9305
9306user <user>
9307 Sets the owner of the UNIX sockets to the designated system user. It can also
9308 be set by default in the global section's "unix-bind" statement. Note that
9309 some platforms simply ignore this. This setting is equivalent to the "uid"
9310 setting except that the user name is used instead of its uid. This setting is
9311 ignored by non UNIX sockets.
9312
Emeric Brun1a073b42012-09-28 17:07:34 +02009313verify [none|optional|required]
9314 This setting is only available when support for OpenSSL was built in. If set
9315 to 'none', client certificate is not requested. This is the default. In other
9316 cases, a client certificate is requested. If the client does not provide a
9317 certificate after the request and if 'verify' is set to 'required', then the
9318 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +02009319 certificate provided by the client is always verified using CAs from
9320 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
9321 is aborted, regardless of the 'verify' option, unless the error code exactly
9322 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02009323
Willy Tarreaub6205fd2012-09-24 12:27:33 +020093245.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +01009325------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02009326
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01009327The "server" and "default-server" keywords support a certain number of settings
9328which are all passed as arguments on the server line. The order in which those
9329arguments appear does not count, and they are all optional. Some of those
9330settings are single words (booleans) while others expect one or several values
9331after them. In this case, the values must immediately follow the setting name.
9332Except default-server, all those settings must be specified after the server's
9333address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +02009334
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009335 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01009336 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +02009337
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009338The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +01009339
Willy Tarreauceb4ac92012-04-28 00:41:46 +02009340addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009341 Using the "addr" parameter, it becomes possible to use a different IP address
9342 to send health-checks. On some servers, it may be desirable to dedicate an IP
9343 address to specific component able to perform complex tests which are more
9344 suitable to health-checks than the application. This parameter is ignored if
9345 the "check" parameter is not set. See also the "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +02009346
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009347 Supported in default-server: No
9348
Simon Hormand60d6912013-11-25 10:46:36 +09009349agent-check
9350 Enable an auxiliary agent check which is run independently of a regular
Willy Tarreau81f5d942013-12-09 20:51:51 +01009351 health check. An agent health check is performed by making a TCP connection
9352 to the port set by the "agent-port" parameter and reading an ASCII string.
9353 The string is made of a series of words delimited by spaces, tabs or commas
9354 in any order, optionally terminated by '\r' and/or '\n', each consisting of :
Simon Hormand60d6912013-11-25 10:46:36 +09009355
Willy Tarreau81f5d942013-12-09 20:51:51 +01009356 - An ASCII representation of a positive integer percentage, e.g. "75%".
Simon Hormand60d6912013-11-25 10:46:36 +09009357 Values in this format will set the weight proportional to the initial
Willy Tarreauc5af3a62014-10-07 15:27:33 +02009358 weight of a server as configured when haproxy starts. Note that a zero
9359 weight is reported on the stats page as "DRAIN" since it has the same
9360 effect on the server (it's removed from the LB farm).
Simon Hormand60d6912013-11-25 10:46:36 +09009361
Willy Tarreau81f5d942013-12-09 20:51:51 +01009362 - The word "ready". This will turn the server's administrative state to the
9363 READY mode, thus cancelling any DRAIN or MAINT state
Simon Hormand60d6912013-11-25 10:46:36 +09009364
Willy Tarreau81f5d942013-12-09 20:51:51 +01009365 - The word "drain". This will turn the server's administrative state to the
9366 DRAIN mode, thus it will not accept any new connections other than those
9367 that are accepted via persistence.
Simon Hormand60d6912013-11-25 10:46:36 +09009368
Willy Tarreau81f5d942013-12-09 20:51:51 +01009369 - The word "maint". This will turn the server's administrative state to the
9370 MAINT mode, thus it will not accept any new connections at all, and health
9371 checks will be stopped.
Simon Hormand60d6912013-11-25 10:46:36 +09009372
Willy Tarreau81f5d942013-12-09 20:51:51 +01009373 - The words "down", "failed", or "stopped", optionally followed by a
9374 description string after a sharp ('#'). All of these mark the server's
9375 operating state as DOWN, but since the word itself is reported on the stats
9376 page, the difference allows an administrator to know if the situation was
9377 expected or not : the service may intentionally be stopped, may appear up
9378 but fail some validity tests, or may be seen as down (eg: missing process,
9379 or port not responding).
Simon Hormand60d6912013-11-25 10:46:36 +09009380
Willy Tarreau81f5d942013-12-09 20:51:51 +01009381 - The word "up" sets back the server's operating state as UP if health checks
9382 also report that the service is accessible.
Simon Hormand60d6912013-11-25 10:46:36 +09009383
Willy Tarreau81f5d942013-12-09 20:51:51 +01009384 Parameters which are not advertised by the agent are not changed. For
9385 example, an agent might be designed to monitor CPU usage and only report a
9386 relative weight and never interact with the operating status. Similarly, an
9387 agent could be designed as an end-user interface with 3 radio buttons
9388 allowing an administrator to change only the administrative state. However,
9389 it is important to consider that only the agent may revert its own actions,
9390 so if a server is set to DRAIN mode or to DOWN state using the agent, the
9391 agent must implement the other equivalent actions to bring the service into
9392 operations again.
Simon Hormand60d6912013-11-25 10:46:36 +09009393
Simon Horman2f1f9552013-11-25 10:46:37 +09009394 Failure to connect to the agent is not considered an error as connectivity
9395 is tested by the regular health check which is enabled by the "check"
Willy Tarreau81f5d942013-12-09 20:51:51 +01009396 parameter. Warning though, it is not a good idea to stop an agent after it
9397 reports "down", since only an agent reporting "up" will be able to turn the
9398 server up again. Note that the CLI on the Unix stats socket is also able to
9399 force an agent's result in order to workaround a bogus agent if needed.
Simon Horman2f1f9552013-11-25 10:46:37 +09009400
Willy Tarreau81f5d942013-12-09 20:51:51 +01009401 Requires the "agent-port" parameter to be set. See also the "agent-inter"
9402 parameter.
Simon Hormand60d6912013-11-25 10:46:36 +09009403
9404 Supported in default-server: No
9405
9406agent-inter <delay>
9407 The "agent-inter" parameter sets the interval between two agent checks
9408 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
9409
9410 Just as with every other time-based parameter, it may be entered in any
9411 other explicit unit among { us, ms, s, m, h, d }. The "agent-inter"
9412 parameter also serves as a timeout for agent checks "timeout check" is
9413 not set. In order to reduce "resonance" effects when multiple servers are
9414 hosted on the same hardware, the agent and health checks of all servers
9415 are started with a small time offset between them. It is also possible to
9416 add some random noise in the agent and health checks interval using the
9417 global "spread-checks" keyword. This makes sense for instance when a lot
9418 of backends use the same servers.
9419
9420 See also the "agent-check" and "agent-port" parameters.
9421
9422 Supported in default-server: Yes
9423
9424agent-port <port>
9425 The "agent-port" parameter sets the TCP port used for agent checks.
9426
9427 See also the "agent-check" and "agent-inter" parameters.
9428
9429 Supported in default-server: Yes
9430
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009431backup
9432 When "backup" is present on a server line, the server is only used in load
9433 balancing when all other non-backup servers are unavailable. Requests coming
9434 with a persistence cookie referencing the server will always be served
9435 though. By default, only the first operational backup server is used, unless
9436 the "allbackups" option is set in the backend. See also the "allbackups"
9437 option.
Willy Tarreau6a06a402007-07-15 20:15:28 +02009438
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009439 Supported in default-server: No
9440
Emeric Brunef42d922012-10-11 16:11:36 +02009441ca-file <cafile>
9442 This setting is only available when support for OpenSSL was built in. It
9443 designates a PEM file from which to load CA certificates used to verify
9444 server's certificate.
9445
9446 Supported in default-server: No
9447
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009448check
9449 This option enables health checks on the server. By default, a server is
Patrick Mézardb7aeec62012-01-22 16:01:22 +01009450 always considered available. If "check" is set, the server is available when
9451 accepting periodic TCP connections, to ensure that it is really able to serve
9452 requests. The default address and port to send the tests to are those of the
9453 server, and the default source is the same as the one defined in the
9454 backend. It is possible to change the address using the "addr" parameter, the
9455 port using the "port" parameter, the source address using the "source"
9456 address, and the interval and timers using the "inter", "rise" and "fall"
Simon Hormanafc47ee2013-11-25 10:46:35 +09009457 parameters. The request method is define in the backend using the "httpchk",
9458 "smtpchk", "mysql-check", "pgsql-check" and "ssl-hello-chk" options. Please
9459 refer to those options and parameters for more information.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009460
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009461 Supported in default-server: No
9462
Willy Tarreau6c16adc2012-10-05 00:04:16 +02009463check-send-proxy
9464 This option forces emission of a PROXY protocol line with outgoing health
9465 checks, regardless of whether the server uses send-proxy or not for the
9466 normal traffic. By default, the PROXY protocol is enabled for health checks
9467 if it is already enabled for normal traffic and if no "port" nor "addr"
9468 directive is present. However, if such a directive is present, the
9469 "check-send-proxy" option needs to be used to force the use of the
9470 protocol. See also the "send-proxy" option for more information.
9471
9472 Supported in default-server: No
9473
Willy Tarreau763a95b2012-10-04 23:15:39 +02009474check-ssl
9475 This option forces encryption of all health checks over SSL, regardless of
9476 whether the server uses SSL or not for the normal traffic. This is generally
9477 used when an explicit "port" or "addr" directive is specified and SSL health
9478 checks are not inherited. It is important to understand that this option
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009479 inserts an SSL transport layer below the checks, so that a simple TCP connect
Willy Tarreau763a95b2012-10-04 23:15:39 +02009480 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
9481 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
9482 All SSL settings are common to health checks and traffic (eg: ciphers).
9483 See the "ssl" option for more information.
9484
9485 Supported in default-server: No
9486
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009487ciphers <ciphers>
9488 This option sets the string describing the list of cipher algorithms that is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009489 is negotiated during the SSL/TLS handshake with the server. The format of the
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009490 string is defined in "man 1 ciphers". When SSL is used to communicate with
9491 servers on the local network, it is common to see a weaker set of algorithms
9492 than what is used over the internet. Doing so reduces CPU usage on both the
9493 server and haproxy while still keeping it compatible with deployed software.
9494 Some algorithms such as RC4-SHA1 are reasonably cheap. If no security at all
9495 is needed and just connectivity, using DES can be appropriate.
9496
Willy Tarreau763a95b2012-10-04 23:15:39 +02009497 Supported in default-server: No
9498
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009499cookie <value>
9500 The "cookie" parameter sets the cookie value assigned to the server to
9501 <value>. This value will be checked in incoming requests, and the first
9502 operational server possessing the same value will be selected. In return, in
9503 cookie insertion or rewrite modes, this value will be assigned to the cookie
9504 sent to the client. There is nothing wrong in having several servers sharing
9505 the same cookie value, and it is in fact somewhat common between normal and
9506 backup servers. See also the "cookie" keyword in backend section.
9507
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009508 Supported in default-server: No
9509
Emeric Brunef42d922012-10-11 16:11:36 +02009510crl-file <crlfile>
9511 This setting is only available when support for OpenSSL was built in. It
9512 designates a PEM file from which to load certificate revocation list used
9513 to verify server's certificate.
9514
9515 Supported in default-server: No
9516
Emeric Bruna7aa3092012-10-26 12:58:00 +02009517crt <cert>
9518 This setting is only available when support for OpenSSL was built in.
9519 It designates a PEM file from which to load both a certificate and the
9520 associated private key. This file can be built by concatenating both PEM
9521 files into one. This certificate will be sent if the server send a client
9522 certificate request.
9523
9524 Supported in default-server: No
9525
Willy Tarreau96839092010-03-29 10:02:24 +02009526disabled
9527 The "disabled" keyword starts the server in the "disabled" state. That means
9528 that it is marked down in maintenance mode, and no connection other than the
9529 ones allowed by persist mode will reach it. It is very well suited to setup
9530 new servers, because normal traffic will never reach them, while it is still
9531 possible to test the service by making use of the force-persist mechanism.
9532
9533 Supported in default-server: No
9534
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009535error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +01009536 If health observing is enabled, the "error-limit" parameter specifies the
9537 number of consecutive errors that triggers event selected by the "on-error"
9538 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009539
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009540 Supported in default-server: Yes
9541
9542 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009543
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009544fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009545 The "fall" parameter states that a server will be considered as dead after
9546 <count> consecutive unsuccessful health checks. This value defaults to 3 if
9547 unspecified. See also the "check", "inter" and "rise" parameters.
9548
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009549 Supported in default-server: Yes
9550
Emeric Brun8694b9a2012-10-05 14:39:07 +02009551force-sslv3
9552 This option enforces use of SSLv3 only when SSL is used to communicate with
9553 the server. SSLv3 is generally less expensive than the TLS counterparts for
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009554 high connection rates. This option is also available on global statement
9555 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +02009556
9557 Supported in default-server: No
9558
9559force-tlsv10
9560 This option enforces use of TLSv1.0 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009561 the server. This option is also available on global statement
9562 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +02009563
9564 Supported in default-server: No
9565
9566force-tlsv11
9567 This option enforces use of TLSv1.1 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009568 the server. This option is also available on global statement
9569 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +02009570
9571 Supported in default-server: No
9572
9573force-tlsv12
9574 This option enforces use of TLSv1.2 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009575 the server. This option is also available on global statement
9576 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +02009577
9578 Supported in default-server: No
9579
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009580id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02009581 Set a persistent ID for the server. This ID must be positive and unique for
9582 the proxy. An unused ID will automatically be assigned if unset. The first
9583 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009584
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009585 Supported in default-server: No
9586
9587inter <delay>
9588fastinter <delay>
9589downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009590 The "inter" parameter sets the interval between two consecutive health checks
9591 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
9592 It is also possible to use "fastinter" and "downinter" to optimize delays
9593 between checks depending on the server state :
9594
9595 Server state | Interval used
9596 ---------------------------------+-----------------------------------------
9597 UP 100% (non-transitional) | "inter"
9598 ---------------------------------+-----------------------------------------
9599 Transitionally UP (going down), |
9600 Transitionally DOWN (going up), | "fastinter" if set, "inter" otherwise.
9601 or yet unchecked. |
9602 ---------------------------------+-----------------------------------------
9603 DOWN 100% (non-transitional) | "downinter" if set, "inter" otherwise.
9604 ---------------------------------+-----------------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +01009605
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009606 Just as with every other time-based parameter, they can be entered in any
9607 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
9608 serves as a timeout for health checks sent to servers if "timeout check" is
9609 not set. In order to reduce "resonance" effects when multiple servers are
Simon Hormand60d6912013-11-25 10:46:36 +09009610 hosted on the same hardware, the agent and health checks of all servers
9611 are started with a small time offset between them. It is also possible to
9612 add some random noise in the agent and health checks interval using the
9613 global "spread-checks" keyword. This makes sense for instance when a lot
9614 of backends use the same servers.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009615
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009616 Supported in default-server: Yes
9617
9618maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009619 The "maxconn" parameter specifies the maximal number of concurrent
9620 connections that will be sent to this server. If the number of incoming
9621 concurrent requests goes higher than this value, they will be queued, waiting
9622 for a connection to be released. This parameter is very important as it can
9623 save fragile servers from going down under extreme loads. If a "minconn"
9624 parameter is specified, the limit becomes dynamic. The default value is "0"
9625 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
9626 the backend's "fullconn" keyword.
9627
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009628 Supported in default-server: Yes
9629
9630maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009631 The "maxqueue" parameter specifies the maximal number of connections which
9632 will wait in the queue for this server. If this limit is reached, next
9633 requests will be redispatched to other servers instead of indefinitely
9634 waiting to be served. This will break persistence but may allow people to
9635 quickly re-log in when the server they try to connect to is dying. The
9636 default value is "0" which means the queue is unlimited. See also the
9637 "maxconn" and "minconn" parameters.
9638
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009639 Supported in default-server: Yes
9640
9641minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009642 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
9643 limit following the backend's load. The server will always accept at least
9644 <minconn> connections, never more than <maxconn>, and the limit will be on
9645 the ramp between both values when the backend has less than <fullconn>
9646 concurrent connections. This makes it possible to limit the load on the
9647 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009648 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009649 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009650
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009651 Supported in default-server: Yes
9652
Willy Tarreau2a3fb1c2015-02-05 16:47:07 +01009653no-ssl-reuse
9654 This option disables SSL session reuse when SSL is used to communicate with
9655 the server. It will force the server to perform a full handshake for every
9656 new connection. It's probably only useful for benchmarking, troubleshooting,
9657 and for paranoid users.
9658
9659 Supported in default-server: No
9660
Emeric Brun9b3009b2012-10-05 11:55:06 +02009661no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009662 This option disables support for SSLv3 when SSL is used to communicate with
9663 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emeric Brun8694b9a2012-10-05 14:39:07 +02009664 using any configuration option. See also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009665
Willy Tarreau763a95b2012-10-04 23:15:39 +02009666 Supported in default-server: No
9667
Emeric Brunf9c5c472012-10-11 15:28:34 +02009668no-tls-tickets
9669 This setting is only available when support for OpenSSL was built in. It
9670 disables the stateless session resumption (RFC 5077 TLS Ticket
9671 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009672 session resumption is more expensive in CPU usage for servers. This option
9673 is also available on global statement "ssl-default-server-options".
Emeric Brunf9c5c472012-10-11 15:28:34 +02009674
9675 Supported in default-server: No
9676
Emeric Brun9b3009b2012-10-05 11:55:06 +02009677no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +02009678 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +02009679 the server. Note that SSLv2 is disabled in the code and cannot be enabled
9680 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009681 often makes sense to disable it when communicating with local servers. This
9682 option is also available on global statement "ssl-default-server-options".
9683 See also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +02009684
Willy Tarreau763a95b2012-10-04 23:15:39 +02009685 Supported in default-server: No
9686
Emeric Brun9b3009b2012-10-05 11:55:06 +02009687no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +02009688 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +02009689 the server. Note that SSLv2 is disabled in the code and cannot be enabled
9690 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009691 often makes sense to disable it when communicating with local servers. This
9692 option is also available on global statement "ssl-default-server-options".
9693 See also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +02009694
Willy Tarreau763a95b2012-10-04 23:15:39 +02009695 Supported in default-server: No
9696
Emeric Brun9b3009b2012-10-05 11:55:06 +02009697no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +02009698 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009699 the server. Note that SSLv2 is disabled in the code and cannot be enabled
9700 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009701 often makes sense to disable it when communicating with local servers. This
9702 option is also available on global statement "ssl-default-server-options".
9703 See also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009704
Willy Tarreau763a95b2012-10-04 23:15:39 +02009705 Supported in default-server: No
9706
Simon Hormanfa461682011-06-25 09:39:49 +09009707non-stick
9708 Never add connections allocated to this sever to a stick-table.
9709 This may be used in conjunction with backup to ensure that
9710 stick-table persistence is disabled for backup servers.
9711
Willy Tarreau763a95b2012-10-04 23:15:39 +02009712 Supported in default-server: No
9713
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009714observe <mode>
9715 This option enables health adjusting based on observing communication with
9716 the server. By default this functionality is disabled and enabling it also
9717 requires to enable health checks. There are two supported modes: "layer4" and
9718 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
9719 significant. In layer7, which is only allowed for http proxies, responses
9720 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +01009721 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009722
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009723 Supported in default-server: No
9724
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009725 See also the "check", "on-error" and "error-limit".
9726
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009727on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009728 Select what should happen when enough consecutive errors are detected.
9729 Currently, four modes are available:
9730 - fastinter: force fastinter
9731 - fail-check: simulate a failed check, also forces fastinter (default)
9732 - sudden-death: simulate a pre-fatal failed health check, one more failed
9733 check will mark a server down, forces fastinter
9734 - mark-down: mark the server immediately down and force fastinter
9735
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009736 Supported in default-server: Yes
9737
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009738 See also the "check", "observe" and "error-limit".
9739
Simon Hormane0d1bfb2011-06-21 14:34:58 +09009740on-marked-down <action>
9741 Modify what occurs when a server is marked down.
9742 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -07009743 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
9744 all connections to the server are immediately terminated when the server
9745 goes down. It might be used if the health check detects more complex cases
9746 than a simple connection status, and long timeouts would cause the service
9747 to remain unresponsive for too long a time. For instance, a health check
9748 might detect that a database is stuck and that there's no chance to reuse
9749 existing connections anymore. Connections killed this way are logged with
9750 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +09009751
9752 Actions are disabled by default
9753
9754 Supported in default-server: Yes
9755
Justin Karnegeseb2c24a2012-05-24 15:28:52 -07009756on-marked-up <action>
9757 Modify what occurs when a server is marked up.
9758 Currently one action is available:
9759 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
9760 done only if the server is not in backup state and if it is not disabled
9761 (it must have an effective weight > 0). This can be used sometimes to force
9762 an active server to take all the traffic back after recovery when dealing
9763 with long sessions (eg: LDAP, SQL, ...). Doing this can cause more trouble
9764 than it tries to solve (eg: incomplete transactions), so use this feature
9765 with extreme care. Sessions killed because a server comes up are logged
9766 with an 'U' termination code (for "Up").
9767
9768 Actions are disabled by default
9769
9770 Supported in default-server: Yes
9771
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009772port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009773 Using the "port" parameter, it becomes possible to use a different port to
9774 send health-checks. On some servers, it may be desirable to dedicate a port
9775 to a specific component able to perform complex tests which are more suitable
9776 to health-checks than the application. It is common to run a simple script in
9777 inetd for instance. This parameter is ignored if the "check" parameter is not
9778 set. See also the "addr" parameter.
9779
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009780 Supported in default-server: Yes
9781
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009782redir <prefix>
9783 The "redir" parameter enables the redirection mode for all GET and HEAD
9784 requests addressing this server. This means that instead of having HAProxy
9785 forward the request to the server, it will send an "HTTP 302" response with
9786 the "Location" header composed of this prefix immediately followed by the
9787 requested URI beginning at the leading '/' of the path component. That means
9788 that no trailing slash should be used after <prefix>. All invalid requests
9789 will be rejected, and all non-GET or HEAD requests will be normally served by
9790 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009791 mangling nor cookie insertion is possible in the response. However, cookies in
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009792 requests are still analysed, making this solution completely usable to direct
9793 users to a remote location in case of local disaster. Main use consists in
9794 increasing bandwidth for static servers by having the clients directly
9795 connect to them. Note: never use a relative location here, it would cause a
9796 loop between the client and HAProxy!
9797
9798 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
9799
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009800 Supported in default-server: No
9801
9802rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009803 The "rise" parameter states that a server will be considered as operational
9804 after <count> consecutive successful health checks. This value defaults to 2
9805 if unspecified. See also the "check", "inter" and "fall" parameters.
9806
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009807 Supported in default-server: Yes
9808
Willy Tarreau5ab04ec2011-03-20 10:32:26 +01009809send-proxy
9810 The "send-proxy" parameter enforces use of the PROXY protocol over any
9811 connection established to this server. The PROXY protocol informs the other
9812 end about the layer 3/4 addresses of the incoming connection, so that it can
9813 know the client's address or the public address it accessed to, whatever the
9814 upper layer protocol. For connections accepted by an "accept-proxy" listener,
9815 the advertised address will be used. Only TCPv4 and TCPv6 address families
9816 are supported. Other families such as Unix sockets, will report an UNKNOWN
9817 family. Servers using this option can fully be chained to another instance of
9818 haproxy listening with an "accept-proxy" setting. This setting must not be
Willy Tarreau6c16adc2012-10-05 00:04:16 +02009819 used if the server isn't aware of the protocol. When health checks are sent
9820 to the server, the PROXY protocol is automatically used when this option is
9821 set, unless there is an explicit "port" or "addr" directive, in which case an
9822 explicit "check-send-proxy" directive would also be needed to use the PROXY
9823 protocol. See also the "accept-proxy" option of the "bind" keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +01009824
9825 Supported in default-server: No
9826
David Safb76832014-05-08 23:42:08 -04009827send-proxy-v2
9828 The "send-proxy-v2" parameter enforces use of the PROXY protocol version 2
9829 over any connection established to this server. The PROXY protocol informs
9830 the other end about the layer 3/4 addresses of the incoming connection, so
9831 that it can know the client's address or the public address it accessed to,
9832 whatever the upper layer protocol. This setting must not be used if the
9833 server isn't aware of this version of the protocol. See also the "send-proxy"
9834 option of the "bind" keyword.
9835
9836 Supported in default-server: No
9837
9838send-proxy-v2-ssl
9839 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
9840 2 over any connection established to this server. The PROXY protocol informs
9841 the other end about the layer 3/4 addresses of the incoming connection, so
9842 that it can know the client's address or the public address it accessed to,
9843 whatever the upper layer protocol. In addition, the SSL information extension
9844 of the PROXY protocol is added to the PROXY protocol header. This setting
9845 must not be used if the server isn't aware of this version of the protocol.
9846 See also the "send-proxy-v2" option of the "bind" keyword.
9847
9848 Supported in default-server: No
9849
9850send-proxy-v2-ssl-cn
9851 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
9852 2 over any connection established to this server. The PROXY protocol informs
9853 the other end about the layer 3/4 addresses of the incoming connection, so
9854 that it can know the client's address or the public address it accessed to,
9855 whatever the upper layer protocol. In addition, the SSL information extension
9856 of the PROXY protocol, along along with the Common Name from the subject of
9857 the client certificate (if any), is added to the PROXY protocol header. This
9858 setting must not be used if the server isn't aware of this version of the
9859 protocol. See also the "send-proxy-v2" option of the "bind" keyword.
9860
9861 Supported in default-server: No
9862
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009863slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009864 The "slowstart" parameter for a server accepts a value in milliseconds which
9865 indicates after how long a server which has just come back up will run at
9866 full speed. Just as with every other time-based parameter, it can be entered
9867 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
9868 linearly from 0 to 100% during this time. The limitation applies to two
9869 parameters :
9870
9871 - maxconn: the number of connections accepted by the server will grow from 1
9872 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
9873
9874 - weight: when the backend uses a dynamic weighted algorithm, the weight
9875 grows linearly from 1 to 100%. In this case, the weight is updated at every
9876 health-check. For this reason, it is important that the "inter" parameter
9877 is smaller than the "slowstart", in order to maximize the number of steps.
9878
9879 The slowstart never applies when haproxy starts, otherwise it would cause
9880 trouble to running servers. It only applies when a server has been previously
9881 seen as failed.
9882
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009883 Supported in default-server: Yes
9884
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02009885source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02009886source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02009887source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009888 The "source" parameter sets the source address which will be used when
9889 connecting to the server. It follows the exact same parameters and principle
9890 as the backend "source" keyword, except that it only applies to the server
9891 referencing it. Please consult the "source" keyword for details.
9892
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02009893 Additionally, the "source" statement on a server line allows one to specify a
9894 source port range by indicating the lower and higher bounds delimited by a
9895 dash ('-'). Some operating systems might require a valid IP address when a
9896 source port range is specified. It is permitted to have the same IP/range for
9897 several servers. Doing so makes it possible to bypass the maximum of 64k
9898 total concurrent connections. The limit will then reach 64k connections per
9899 server.
9900
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009901 Supported in default-server: No
9902
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009903ssl
Willy Tarreau44f65392013-06-25 07:56:20 +02009904 This option enables SSL ciphering on outgoing connections to the server. It
9905 is critical to verify server certificates using "verify" when using SSL to
9906 connect to servers, otherwise the communication is prone to trivial man in
9907 the-middle attacks rendering SSL useless. When this option is used, health
9908 checks are automatically sent in SSL too unless there is a "port" or an
9909 "addr" directive indicating the check should be sent to a different location.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009910 See the "check-ssl" option to force SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +02009911
9912 Supported in default-server: No
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009913
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009914track [<proxy>/]<server>
Willy Tarreau32091232014-05-16 13:52:00 +02009915 This option enables ability to set the current state of the server by tracking
9916 another one. It is possible to track a server which itself tracks another
9917 server, provided that at the end of the chain, a server has health checks
9918 enabled. If <proxy> is omitted the current one is used. If disable-on-404 is
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009919 used, it has to be enabled on both proxies.
9920
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009921 Supported in default-server: No
9922
Emeric Brunef42d922012-10-11 16:11:36 +02009923verify [none|required]
9924 This setting is only available when support for OpenSSL was built in. If set
Emeric Brun850efd52014-01-29 12:24:34 +01009925 to 'none', server certificate is not verified. In the other case, The
9926 certificate provided by the server is verified using CAs from 'ca-file'
9927 and optional CRLs from 'crl-file'. If 'ssl_server_verify' is not specified
9928 in global section, this is the default. On verify failure the handshake
Willy Tarreau44f65392013-06-25 07:56:20 +02009929 is aborted. It is critically important to verify server certificates when
9930 using SSL to connect to servers, otherwise the communication is prone to
9931 trivial man-in-the-middle attacks rendering SSL totally useless.
Emeric Brunef42d922012-10-11 16:11:36 +02009932
9933 Supported in default-server: No
9934
Evan Broderbe554312013-06-27 00:05:25 -07009935verifyhost <hostname>
9936 This setting is only available when support for OpenSSL was built in, and
9937 only takes effect if 'verify required' is also specified. When set, the
9938 hostnames in the subject and subjectAlternateNames of the certificate
9939 provided by the server are checked. If none of the hostnames in the
9940 certificate match the specified hostname, the handshake is aborted. The
9941 hostnames in the server-provided certificate may include wildcards.
9942
9943 Supported in default-server: No
9944
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009945weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009946 The "weight" parameter is used to adjust the server's weight relative to
9947 other servers. All servers will receive a load proportional to their weight
9948 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +02009949 load. The default weight is 1, and the maximal value is 256. A value of 0
9950 means the server will not participate in load-balancing but will still accept
9951 persistent connections. If this parameter is used to distribute the load
9952 according to server's capacity, it is recommended to start with values which
9953 can both grow and shrink, for instance between 10 and 100 to leave enough
9954 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009955
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009956 Supported in default-server: Yes
9957
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009958
99596. HTTP header manipulation
9960---------------------------
9961
9962In HTTP mode, it is possible to rewrite, add or delete some of the request and
9963response headers based on regular expressions. It is also possible to block a
9964request or a response if a particular header matches a regular expression,
9965which is enough to stop most elementary protocol attacks, and to protect
Willy Tarreau70dffda2014-01-30 03:07:23 +01009966against information leak from the internal network.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009967
Willy Tarreau70dffda2014-01-30 03:07:23 +01009968If HAProxy encounters an "Informational Response" (status code 1xx), it is able
9969to process all rsp* rules which can allow, deny, rewrite or delete a header,
9970but it will refuse to add a header to any such messages as this is not
9971HTTP-compliant. The reason for still processing headers in such responses is to
9972stop and/or fix any possible information leak which may happen, for instance
9973because another downstream equipment would unconditionally add a header, or if
9974a server name appears there. When such messages are seen, normal processing
9975still occurs on the next non-informational messages.
Willy Tarreau816b9792009-09-15 21:25:21 +02009976
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009977This section covers common usage of the following keywords, described in detail
9978in section 4.2 :
9979
9980 - reqadd <string>
9981 - reqallow <search>
9982 - reqiallow <search>
9983 - reqdel <search>
9984 - reqidel <search>
9985 - reqdeny <search>
9986 - reqideny <search>
9987 - reqpass <search>
9988 - reqipass <search>
9989 - reqrep <search> <replace>
9990 - reqirep <search> <replace>
9991 - reqtarpit <search>
9992 - reqitarpit <search>
9993 - rspadd <string>
9994 - rspdel <search>
9995 - rspidel <search>
9996 - rspdeny <search>
9997 - rspideny <search>
9998 - rsprep <search> <replace>
9999 - rspirep <search> <replace>
10000
10001With all these keywords, the same conventions are used. The <search> parameter
10002is a POSIX extended regular expression (regex) which supports grouping through
10003parenthesis (without the backslash). Spaces and other delimiters must be
10004prefixed with a backslash ('\') to avoid confusion with a field delimiter.
10005Other characters may be prefixed with a backslash to change their meaning :
10006
10007 \t for a tab
10008 \r for a carriage return (CR)
10009 \n for a new line (LF)
10010 \ to mark a space and differentiate it from a delimiter
10011 \# to mark a sharp and differentiate it from a comment
10012 \\ to use a backslash in a regex
10013 \\\\ to use a backslash in the text (*2 for regex, *2 for haproxy)
10014 \xXX to write the ASCII hex code XX as in the C language
10015
10016The <replace> parameter contains the string to be used to replace the largest
10017portion of text matching the regex. It can make use of the special characters
10018above, and can reference a substring which is delimited by parenthesis in the
10019regex, by writing a backslash ('\') immediately followed by one digit from 0 to
100209 indicating the group position (0 designating the entire line). This practice
10021is very common to users of the "sed" program.
10022
10023The <string> parameter represents the string which will systematically be added
10024after the last header line. It can also use special character sequences above.
10025
10026Notes related to these keywords :
10027---------------------------------
10028 - these keywords are not always convenient to allow/deny based on header
10029 contents. It is strongly recommended to use ACLs with the "block" keyword
10030 instead, resulting in far more flexible and manageable rules.
10031
10032 - lines are always considered as a whole. It is not possible to reference
10033 a header name only or a value only. This is important because of the way
10034 headers are written (notably the number of spaces after the colon).
10035
10036 - the first line is always considered as a header, which makes it possible to
10037 rewrite or filter HTTP requests URIs or response codes, but in turn makes
10038 it harder to distinguish between headers and request line. The regex prefix
10039 ^[^\ \t]*[\ \t] matches any HTTP method followed by a space, and the prefix
10040 ^[^ \t:]*: matches any header name followed by a colon.
10041
10042 - for performances reasons, the number of characters added to a request or to
10043 a response is limited at build time to values between 1 and 4 kB. This
10044 should normally be far more than enough for most usages. If it is too short
10045 on occasional usages, it is possible to gain some space by removing some
10046 useless headers before adding new ones.
10047
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010048 - keywords beginning with "reqi" and "rspi" are the same as their counterpart
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010049 without the 'i' letter except that they ignore case when matching patterns.
10050
10051 - when a request passes through a frontend then a backend, all req* rules
10052 from the frontend will be evaluated, then all req* rules from the backend
10053 will be evaluated. The reverse path is applied to responses.
10054
10055 - req* statements are applied after "block" statements, so that "block" is
10056 always the first one, but before "use_backend" in order to permit rewriting
Willy Tarreaud72758d2010-01-12 10:42:19 +010010057 before switching.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010058
10059
Willy Tarreau74ca5042013-06-11 23:12:07 +0200100607. Using ACLs and fetching samples
10061----------------------------------
10062
10063Haproxy is capable of extracting data from request or response streams, from
10064client or server information, from tables, environmental information etc...
10065The action of extracting such data is called fetching a sample. Once retrieved,
10066these samples may be used for various purposes such as a key to a stick-table,
10067but most common usages consist in matching them against predefined constant
10068data called patterns.
10069
10070
100717.1. ACL basics
10072---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010073
10074The use of Access Control Lists (ACL) provides a flexible solution to perform
10075content switching and generally to take decisions based on content extracted
10076from the request, the response or any environmental status. The principle is
10077simple :
10078
Willy Tarreau74ca5042013-06-11 23:12:07 +020010079 - extract a data sample from a stream, table or the environment
Willy Tarreaue6b11e42013-11-26 19:02:32 +010010080 - optionally apply some format conversion to the extracted sample
Willy Tarreau74ca5042013-06-11 23:12:07 +020010081 - apply one or multiple pattern matching methods on this sample
10082 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010083
Willy Tarreau74ca5042013-06-11 23:12:07 +020010084The actions generally consist in blocking a request, selecting a backend, or
10085adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010086
10087In order to define a test, the "acl" keyword is used. The syntax is :
10088
Willy Tarreau74ca5042013-06-11 23:12:07 +020010089 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010090
10091This creates a new ACL <aclname> or completes an existing one with new tests.
10092Those tests apply to the portion of request/response specified in <criterion>
10093and may be adjusted with optional flags [flags]. Some criteria also support
Willy Tarreaue6b11e42013-11-26 19:02:32 +010010094an operator which may be specified before the set of values. Optionally some
10095conversion operators may be applied to the sample, and they will be specified
10096as a comma-delimited list of keywords just after the first keyword. The values
10097are of the type supported by the criterion, and are separated by spaces.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010098
10099ACL names must be formed from upper and lower case letters, digits, '-' (dash),
10100'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
10101which means that "my_acl" and "My_Acl" are two different ACLs.
10102
10103There is no enforced limit to the number of ACLs. The unused ones do not affect
10104performance, they just consume a small amount of memory.
10105
Willy Tarreau74ca5042013-06-11 23:12:07 +020010106The criterion generally is the name of a sample fetch method, or one of its ACL
10107specific declinations. The default test method is implied by the output type of
10108this sample fetch method. The ACL declinations can describe alternate matching
Willy Tarreaue6b11e42013-11-26 19:02:32 +010010109methods of a same sample fetch method. The sample fetch methods are the only
10110ones supporting a conversion.
Willy Tarreau74ca5042013-06-11 23:12:07 +020010111
10112Sample fetch methods return data which can be of the following types :
10113 - boolean
10114 - integer (signed or unsigned)
10115 - IPv4 or IPv6 address
10116 - string
10117 - data block
10118
Willy Tarreaue6b11e42013-11-26 19:02:32 +010010119Converters transform any of these data into any of these. For example, some
10120converters might convert a string to a lower-case string while other ones
10121would turn a string to an IPv4 address, or apply a netmask to an IP address.
10122The resulting sample is of the type of the last converter applied to the list,
10123which defaults to the type of the sample fetch method.
10124
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020010125Each sample or converter returns data of a specific type, specified with its
10126keyword in this documentation. When an ACL is declared using a standard sample
10127fetch method, certain types automatically involved a default matching method
10128which are summarized in the table below :
10129
10130 +---------------------+-----------------+
10131 | Sample or converter | Default |
10132 | output type | matching method |
10133 +---------------------+-----------------+
10134 | boolean | bool |
10135 +---------------------+-----------------+
10136 | integer | int |
10137 +---------------------+-----------------+
10138 | ip | ip |
10139 +---------------------+-----------------+
10140 | string | str |
10141 +---------------------+-----------------+
10142 | binary | none, use "-m" |
10143 +---------------------+-----------------+
10144
10145Note that in order to match a binary samples, it is mandatory to specify a
10146matching method, see below.
10147
Willy Tarreau74ca5042013-06-11 23:12:07 +020010148The ACL engine can match these types against patterns of the following types :
10149 - boolean
10150 - integer or integer range
10151 - IP address / network
10152 - string (exact, substring, suffix, prefix, subdir, domain)
10153 - regular expression
10154 - hex block
10155
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010156The following ACL flags are currently supported :
10157
Willy Tarreau2b5285d2010-05-09 23:45:24 +020010158 -i : ignore case during matching of all subsequent patterns.
10159 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +020010160 -m : use a specific pattern matching method
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010010161 -n : forbid the DNS resolutions
Thierry FOURNIER9860c412014-01-29 14:23:29 +010010162 -M : load the file pointed by -f like a map file.
Thierry FOURNIER3534d882014-01-20 17:01:44 +010010163 -u : force the unique id of the ACL
Willy Tarreau6a06a402007-07-15 20:15:28 +020010164 -- : force end of flags. Useful when a string looks like one of the flags.
10165
Willy Tarreau74ca5042013-06-11 23:12:07 +020010166The "-f" flag is followed by the name of a file from which all lines will be
10167read as individual values. It is even possible to pass multiple "-f" arguments
10168if the patterns are to be loaded from multiple files. Empty lines as well as
10169lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
10170will be stripped. If it is absolutely necessary to insert a valid pattern
10171beginning with a sharp, just prefix it with a space so that it is not taken for
10172a comment. Depending on the data type and match method, haproxy may load the
10173lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
10174exact string matching. In this case, duplicates will automatically be removed.
10175
Thierry FOURNIER9860c412014-01-29 14:23:29 +010010176The "-M" flag allows an ACL to use a map file. If this flag is set, the file is
10177parsed as two column file. The first column contains the patterns used by the
10178ACL, and the second column contain the samples. The sample can be used later by
10179a map. This can be useful in some rare cases where an ACL would just be used to
10180check for the existence of a pattern in a map before a mapping is applied.
10181
Thierry FOURNIER3534d882014-01-20 17:01:44 +010010182The "-u" flag forces the unique id of the ACL. This unique id is used with the
10183socket interface to identify ACL and dynamically change its values. Note that a
10184file is always identified by its name even if an id is set.
10185
Willy Tarreau74ca5042013-06-11 23:12:07 +020010186Also, note that the "-i" flag applies to subsequent entries and not to entries
10187loaded from files preceding it. For instance :
10188
10189 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
10190
10191In this example, each line of "exact-ua.lst" will be exactly matched against
10192the "user-agent" header of the request. Then each line of "generic-ua" will be
10193case-insensitively matched. Then the word "test" will be insensitively matched
10194as well.
10195
10196The "-m" flag is used to select a specific pattern matching method on the input
10197sample. All ACL-specific criteria imply a pattern matching method and generally
10198do not need this flag. However, this flag is useful with generic sample fetch
10199methods to describe how they're going to be matched against the patterns. This
10200is required for sample fetches which return data type for which there is no
10201obvious matching method (eg: string or binary). When "-m" is specified and
10202followed by a pattern matching method name, this method is used instead of the
10203default one for the criterion. This makes it possible to match contents in ways
10204that were not initially planned, or with sample fetch methods which return a
10205string. The matching method also affects the way the patterns are parsed.
10206
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010010207The "-n" flag forbids the dns resolutions. It is used with the load of ip files.
10208By default, if the parser cannot parse ip address it considers that the parsed
10209string is maybe a domain name and try dns resolution. The flag "-n" disable this
10210resolution. It is useful for detecting malformed ip lists. Note that if the DNS
10211server is not reachable, the haproxy configuration parsing may last many minutes
10212waiting fir the timeout. During this time no error messages are displayed. The
10213flag "-n" disable this behavior. Note also that during the runtime, this
10214function is disabled for the dynamic acl modifications.
10215
Willy Tarreau74ca5042013-06-11 23:12:07 +020010216There are some restrictions however. Not all methods can be used with all
10217sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
10218be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +020010219
10220 - "found" : only check if the requested sample could be found in the stream,
10221 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +020010222 to pass any pattern to avoid confusion. This matching method is
10223 particularly useful to detect presence of certain contents such
10224 as headers, cookies, etc... even if they are empty and without
10225 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +020010226
10227 - "bool" : check the value as a boolean. It can only be applied to fetches
10228 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +020010229 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +020010230
10231 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +020010232 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +020010233
10234 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +020010235 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +020010236
10237 - "bin" : match the contents against an hexadecimal string representing a
10238 binary sequence. This may be used with binary or string samples.
10239
10240 - "len" : match the sample's length as an integer. This may be used with
10241 binary or string samples.
10242
Willy Tarreau74ca5042013-06-11 23:12:07 +020010243 - "str" : exact match : match the contents against a string. This may be
10244 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020010245
Willy Tarreau74ca5042013-06-11 23:12:07 +020010246 - "sub" : substring match : check that the contents contain at least one of
10247 the provided string patterns. This may be used with binary or
10248 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020010249
Willy Tarreau74ca5042013-06-11 23:12:07 +020010250 - "reg" : regex match : match the contents against a list of regular
10251 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020010252
Willy Tarreau74ca5042013-06-11 23:12:07 +020010253 - "beg" : prefix match : check that the contents begin like the provided
10254 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020010255
Willy Tarreau74ca5042013-06-11 23:12:07 +020010256 - "end" : suffix match : check that the contents end like the provided
10257 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020010258
Willy Tarreau74ca5042013-06-11 23:12:07 +020010259 - "dir" : subdir match : check that a slash-delimited portion of the
10260 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +020010261 This may be used with binary or string samples.
10262
Willy Tarreau74ca5042013-06-11 23:12:07 +020010263 - "dom" : domain match : check that a dot-delimited portion of the contents
10264 exactly match one of the provided string patterns. This may be
10265 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020010266
10267For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
10268request, it is possible to do :
10269
10270 acl jsess_present cook(JSESSIONID) -m found
10271
10272In order to apply a regular expression on the 500 first bytes of data in the
10273buffer, one would use the following acl :
10274
10275 acl script_tag payload(0,500) -m reg -i <script>
10276
Willy Tarreaue6b11e42013-11-26 19:02:32 +010010277On systems where the regex library is much slower when using "-i", it is
10278possible to convert the sample to lowercase before matching, like this :
10279
10280 acl script_tag payload(0,500),lower -m reg <script>
10281
Willy Tarreau74ca5042013-06-11 23:12:07 +020010282All ACL-specific criteria imply a default matching method. Most often, these
10283criteria are composed by concatenating the name of the original sample fetch
10284method and the matching method. For example, "hdr_beg" applies the "beg" match
10285to samples retrieved using the "hdr" fetch method. Since all ACL-specific
10286criteria rely on a sample fetch method, it is always possible instead to use
10287the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +020010288
Willy Tarreau74ca5042013-06-11 23:12:07 +020010289If an alternate match is specified using "-m" on an ACL-specific criterion,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010290the matching method is simply applied to the underlying sample fetch method.
10291For example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +020010292
Willy Tarreau74ca5042013-06-11 23:12:07 +020010293 acl short_form hdr_beg(host) www.
10294 acl alternate1 hdr_beg(host) -m beg www.
10295 acl alternate2 hdr_dom(host) -m beg www.
10296 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +020010297
Willy Tarreau2b5285d2010-05-09 23:45:24 +020010298
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020010299The table below summarizes the compatibility matrix between sample or converter
10300types and the pattern types to fetch against. It indicates for each compatible
10301combination the name of the matching method to be used, surrounded with angle
10302brackets ">" and "<" when the method is the default one and will work by
10303default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +010010304
Willy Tarreau74ca5042013-06-11 23:12:07 +020010305 +-------------------------------------------------+
10306 | Input sample type |
10307 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020010308 | pattern type | boolean | integer | ip | string | binary |
Willy Tarreau74ca5042013-06-11 23:12:07 +020010309 +----------------------+---------+---------+---------+---------+---------+
10310 | none (presence only) | found | found | found | found | found |
10311 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020010312 | none (boolean value) |> bool <| bool | | bool | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020010313 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020010314 | integer (value) | int |> int <| int | int | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020010315 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010010316 | integer (length) | len | len | len | len | len |
Willy Tarreau74ca5042013-06-11 23:12:07 +020010317 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020010318 | IP address | | |> ip <| ip | ip |
Willy Tarreau74ca5042013-06-11 23:12:07 +020010319 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020010320 | exact string | str | str | str |> str <| str |
Willy Tarreau74ca5042013-06-11 23:12:07 +020010321 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010010322 | prefix | beg | beg | beg | beg | beg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020010323 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010010324 | suffix | end | end | end | end | end |
Willy Tarreau74ca5042013-06-11 23:12:07 +020010325 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010010326 | substring | sub | sub | sub | sub | sub |
Willy Tarreau74ca5042013-06-11 23:12:07 +020010327 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010010328 | subdir | dir | dir | dir | dir | dir |
Willy Tarreau74ca5042013-06-11 23:12:07 +020010329 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010010330 | domain | dom | dom | dom | dom | dom |
Willy Tarreau74ca5042013-06-11 23:12:07 +020010331 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010010332 | regex | reg | reg | reg | reg | reg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020010333 +----------------------+---------+---------+---------+---------+---------+
10334 | hex block | | | | bin | bin |
10335 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +020010336
10337
Willy Tarreau74ca5042013-06-11 23:12:07 +0200103387.1.1. Matching booleans
10339------------------------
10340
10341In order to match a boolean, no value is needed and all values are ignored.
10342Boolean matching is used by default for all fetch methods of type "boolean".
10343When boolean matching is used, the fetched value is returned as-is, which means
10344that a boolean "true" will always match and a boolean "false" will never match.
10345
10346Boolean matching may also be enforced using "-m bool" on fetch methods which
10347return an integer value. Then, integer value 0 is converted to the boolean
10348"false" and all other values are converted to "true".
10349
Willy Tarreau6a06a402007-07-15 20:15:28 +020010350
Willy Tarreau74ca5042013-06-11 23:12:07 +0200103517.1.2. Matching integers
10352------------------------
10353
10354Integer matching applies by default to integer fetch methods. It can also be
10355enforced on boolean fetches using "-m int". In this case, "false" is converted
10356to the integer 0, and "true" is converted to the integer 1.
10357
10358Integer matching also supports integer ranges and operators. Note that integer
10359matching only applies to positive values. A range is a value expressed with a
10360lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010361
10362For instance, "1024:65535" is a valid range to represent a range of
10363unprivileged ports, and "1024:" would also work. "0:1023" is a valid
10364representation of privileged ports, and ":1023" would also work.
10365
Willy Tarreau62644772008-07-16 18:36:06 +020010366As a special case, some ACL functions support decimal numbers which are in fact
10367two integers separated by a dot. This is used with some version checks for
10368instance. All integer properties apply to those decimal numbers, including
10369ranges and operators.
10370
Willy Tarreau6a06a402007-07-15 20:15:28 +020010371For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +010010372operators with ranges does not make much sense and is strongly discouraged.
10373Similarly, it does not make much sense to perform order comparisons with a set
10374of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010375
Willy Tarreau0ba27502007-12-24 16:55:16 +010010376Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +020010377
10378 eq : true if the tested value equals at least one value
10379 ge : true if the tested value is greater than or equal to at least one value
10380 gt : true if the tested value is greater than at least one value
10381 le : true if the tested value is less than or equal to at least one value
10382 lt : true if the tested value is less than at least one value
10383
Willy Tarreau0ba27502007-12-24 16:55:16 +010010384For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +020010385
10386 acl negative-length hdr_val(content-length) lt 0
10387
Willy Tarreau62644772008-07-16 18:36:06 +020010388This one matches SSL versions between 3.0 and 3.1 (inclusive) :
10389
10390 acl sslv3 req_ssl_ver 3:3.1
10391
Willy Tarreau6a06a402007-07-15 20:15:28 +020010392
Willy Tarreau74ca5042013-06-11 23:12:07 +0200103937.1.3. Matching strings
10394-----------------------
10395
10396String matching applies to string or binary fetch methods, and exists in 6
10397different forms :
10398
10399 - exact match (-m str) : the extracted string must exactly match the
10400 patterns ;
10401
10402 - substring match (-m sub) : the patterns are looked up inside the
10403 extracted string, and the ACL matches if any of them is found inside ;
10404
10405 - prefix match (-m beg) : the patterns are compared with the beginning of
10406 the extracted string, and the ACL matches if any of them matches.
10407
10408 - suffix match (-m end) : the patterns are compared with the end of the
10409 extracted string, and the ACL matches if any of them matches.
10410
10411 - subdir match (-m sub) : the patterns are looked up inside the extracted
10412 string, delimited with slashes ("/"), and the ACL matches if any of them
10413 matches.
10414
10415 - domain match (-m dom) : the patterns are looked up inside the extracted
10416 string, delimited with dots ("."), and the ACL matches if any of them
10417 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010418
10419String matching applies to verbatim strings as they are passed, with the
10420exception of the backslash ("\") which makes it possible to escape some
10421characters such as the space. If the "-i" flag is passed before the first
10422string, then the matching will be performed ignoring the case. In order
10423to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +010010424before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020010425
10426
Willy Tarreau74ca5042013-06-11 23:12:07 +0200104277.1.4. Matching regular expressions (regexes)
10428---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020010429
10430Just like with string matching, regex matching applies to verbatim strings as
10431they are passed, with the exception of the backslash ("\") which makes it
10432possible to escape some characters such as the space. If the "-i" flag is
10433passed before the first regex, then the matching will be performed ignoring
10434the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +010010435the "--" flag before the first string. Same principle applies of course to
10436match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020010437
10438
Willy Tarreau74ca5042013-06-11 23:12:07 +0200104397.1.5. Matching arbitrary data blocks
10440-------------------------------------
10441
10442It is possible to match some extracted samples against a binary block which may
10443not safely be represented as a string. For this, the patterns must be passed as
10444a series of hexadecimal digits in an even number, when the match method is set
10445to binary. Each sequence of two digits will represent a byte. The hexadecimal
10446digits may be used upper or lower case.
10447
10448Example :
10449 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
10450 acl hello payload(0,6) -m bin 48656c6c6f0a
10451
10452
104537.1.6. Matching IPv4 and IPv6 addresses
10454---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020010455
10456IPv4 addresses values can be specified either as plain addresses or with a
10457netmask appended, in which case the IPv4 address matches whenever it is
10458within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010459host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +010010460difficult to read and debug configurations. If hostnames are used, you should
10461at least ensure that they are present in /etc/hosts so that the configuration
10462does not depend on any random DNS match at the moment the configuration is
10463parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010464
Willy Tarreauceb4ac92012-04-28 00:41:46 +020010465IPv6 may be entered in their usual form, with or without a netmask appended.
10466Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
10467trouble with randomly resolved IP addresses, host names are never allowed in
10468IPv6 patterns.
10469
10470HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
10471following situations :
10472 - tested address is IPv4, pattern address is IPv4, the match applies
10473 in IPv4 using the supplied mask if any.
10474 - tested address is IPv6, pattern address is IPv6, the match applies
10475 in IPv6 using the supplied mask if any.
10476 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
10477 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
10478 ::IPV4 or ::ffff:IPV4, otherwise it fails.
10479 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
10480 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
10481 applied in IPv6 using the supplied IPv6 mask.
10482
Willy Tarreau74ca5042013-06-11 23:12:07 +020010483
104847.2. Using ACLs to form conditions
10485----------------------------------
10486
10487Some actions are only performed upon a valid condition. A condition is a
10488combination of ACLs with operators. 3 operators are supported :
10489
10490 - AND (implicit)
10491 - OR (explicit with the "or" keyword or the "||" operator)
10492 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +020010493
Willy Tarreau74ca5042013-06-11 23:12:07 +020010494A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +020010495
Willy Tarreau74ca5042013-06-11 23:12:07 +020010496 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +020010497
Willy Tarreau74ca5042013-06-11 23:12:07 +020010498Such conditions are generally used after an "if" or "unless" statement,
10499indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +020010500
Willy Tarreau74ca5042013-06-11 23:12:07 +020010501For instance, to block HTTP requests to the "*" URL with methods other than
10502"OPTIONS", as well as POST requests without content-length, and GET or HEAD
10503requests with a content-length greater than 0, and finally every request which
10504is not either GET/HEAD/POST/OPTIONS !
10505
10506 acl missing_cl hdr_cnt(Content-length) eq 0
10507 block if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
10508 block if METH_GET HTTP_CONTENT
10509 block unless METH_GET or METH_POST or METH_OPTIONS
10510
10511To select a different backend for requests to static contents on the "www" site
10512and to every request on the "img", "video", "download" and "ftp" hosts :
10513
10514 acl url_static path_beg /static /images /img /css
10515 acl url_static path_end .gif .png .jpg .css .js
10516 acl host_www hdr_beg(host) -i www
10517 acl host_static hdr_beg(host) -i img. video. download. ftp.
10518
10519 # now use backend "static" for all static-only hosts, and for static urls
10520 # of host "www". Use backend "www" for the rest.
10521 use_backend static if host_static or host_www url_static
10522 use_backend www if host_www
10523
10524It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
10525expressions that are built on the fly without needing to be declared. They must
10526be enclosed between braces, with a space before and after each brace (because
10527the braces must be seen as independent words). Example :
10528
10529 The following rule :
10530
10531 acl missing_cl hdr_cnt(Content-length) eq 0
10532 block if METH_POST missing_cl
10533
10534 Can also be written that way :
10535
10536 block if METH_POST { hdr_cnt(Content-length) eq 0 }
10537
10538It is generally not recommended to use this construct because it's a lot easier
10539to leave errors in the configuration when written that way. However, for very
10540simple rules matching only one source IP address for instance, it can make more
10541sense to use them than to declare ACLs with random names. Another example of
10542good use is the following :
10543
10544 With named ACLs :
10545
10546 acl site_dead nbsrv(dynamic) lt 2
10547 acl site_dead nbsrv(static) lt 2
10548 monitor fail if site_dead
10549
10550 With anonymous ACLs :
10551
10552 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
10553
10554See section 4.2 for detailed help on the "block" and "use_backend" keywords.
10555
10556
105577.3. Fetching samples
10558---------------------
10559
10560Historically, sample fetch methods were only used to retrieve data to match
10561against patterns using ACLs. With the arrival of stick-tables, a new class of
10562sample fetch methods was created, most often sharing the same syntax as their
10563ACL counterpart. These sample fetch methods are also known as "fetches". As
10564of now, ACLs and fetches have converged. All ACL fetch methods have been made
10565available as fetch methods, and ACLs may use any sample fetch method as well.
10566
10567This section details all available sample fetch methods and their output type.
10568Some sample fetch methods have deprecated aliases that are used to maintain
10569compatibility with existing configurations. They are then explicitly marked as
10570deprecated and should not be used in new setups.
10571
10572The ACL derivatives are also indicated when available, with their respective
10573matching methods. These ones all have a well defined default pattern matching
10574method, so it is never necessary (though allowed) to pass the "-m" option to
10575indicate how the sample will be matched using ACLs.
10576
10577As indicated in the sample type versus matching compatibility matrix above,
10578when using a generic sample fetch method in an ACL, the "-m" option is
10579mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
10580the same keyword exists as an ACL keyword and as a standard fetch method, the
10581ACL engine will automatically pick the ACL-only one by default.
10582
10583Some of these keywords support one or multiple mandatory arguments, and one or
10584multiple optional arguments. These arguments are strongly typed and are checked
10585when the configuration is parsed so that there is no risk of running with an
10586incorrect argument (eg: an unresolved backend name). Fetch function arguments
10587are passed between parenthesis and are delimited by commas. When an argument
10588is optional, it will be indicated below between square brackets ('[ ]'). When
10589all arguments are optional, the parenthesis may be omitted.
10590
10591Thus, the syntax of a standard sample fetch method is one of the following :
10592 - name
10593 - name(arg1)
10594 - name(arg1,arg2)
10595
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010596
105977.3.1. Converters
10598-----------------
10599
Willy Tarreaue6b11e42013-11-26 19:02:32 +010010600Sample fetch methods may be combined with transformations to be applied on top
10601of the fetched sample (also called "converters"). These combinations form what
10602is called "sample expressions" and the result is a "sample". Initially this
10603was only supported by "stick on" and "stick store-request" directives but this
10604has now be extended to all places where samples may be used (acls, log-format,
10605unique-id-format, add-header, ...).
10606
10607These transformations are enumerated as a series of specific keywords after the
10608sample fetch method. These keywords may equally be appended immediately after
10609the fetch keyword's argument, delimited by a comma. These keywords can also
10610support some arguments (eg: a netmask) which must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010611
Willy Tarreau97707872015-01-27 15:12:13 +010010612A certain category of converters are bitwise and arithmetic operators which
10613support performing basic operations on integers. Some bitwise operations are
10614supported (and, or, xor, cpl) and some arithmetic operations are supported
10615(add, sub, mul, div, mod, neg). Some comparators are provided (odd, even, not,
10616bool) which make it possible to report a match without having to write an ACL.
10617
Willy Tarreau74ca5042013-06-11 23:12:07 +020010618The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +010010619
Willy Tarreau97707872015-01-27 15:12:13 +010010620add(<value>)
10621 Adds <value> to the input value of type unsigned integer, and returns the
10622 result as an unsigned integer.
10623
10624and(<value>)
10625 Performs a bitwise "AND" between <value> and the input value of type unsigned
10626 integer, and returns the result as an unsigned integer.
10627
Emeric Brun53d1a982014-04-30 18:21:37 +020010628base64
10629 Converts a binary input sample to a base64 string. It is used to log or
10630 transfer binary content in a way that can be reliably transferred (eg:
10631 an SSL ID can be copied in a header).
10632
Willy Tarreau97707872015-01-27 15:12:13 +010010633bool
10634 Returns a boolean TRUE if the input value of type unsigned integer is
10635 non-null, otherwise returns FALSE. Used in conjunction with and(), it can be
10636 used to report true/false for bit testing on input values (eg: verify the
10637 presence of a flag).
10638
Emeric Brun54c4ac82014-11-03 15:32:43 +010010639bytes(<offset>[,<length>])
10640 Extracts some bytes from an input binary sample. The result is a binary
10641 sample starting at an offset (in bytes) of the original sample and
10642 optionnaly truncated at the given length.
10643
Willy Tarreau97707872015-01-27 15:12:13 +010010644cpl
10645 Takes the input value of type unsigned integer, applies a twos-complement
10646 (flips all bits) and returns the result as an unsigned integer.
10647
Willy Tarreau80599772015-01-20 19:35:24 +010010648crc32([<avalanche>])
10649 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32
10650 hash function. Optionally, it is possible to apply a full avalanche hash
10651 function to the output if the optional <avalanche> argument equals 1. This
10652 converter uses the same functions as used by the various hash-based load
10653 balancing algorithms, so it will provide exactly the same results. It is
10654 provided for compatibility with other software which want a CRC32 to be
10655 computed on some input keys, so it follows the most common implementation as
10656 found in Ethernet, Gzip, PNG, etc... It is slower than the other algorithms
10657 but may provide a better or at least less predictable distribution. It must
10658 not be used for security purposes as a 32-bit hash is trivial to break. See
10659 also "djb2", "sdbm", "wt6" and the "hash-type" directive.
10660
Willy Tarreau97707872015-01-27 15:12:13 +010010661div(<value>)
10662 Divides the input value of type unsigned integer by <value>, and returns the
10663 result as an unsigned integer. If <value> is null, the largest unsigned
10664 integer is returned (typically 2^32-1).
10665
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020010666djb2([<avalanche>])
10667 Hashes a binary input sample into an unsigned 32-bit quantity using the DJB2
10668 hash function. Optionally, it is possible to apply a full avalanche hash
10669 function to the output if the optional <avalanche> argument equals 1. This
10670 converter uses the same functions as used by the various hash-based load
10671 balancing algorithms, so it will provide exactly the same results. It is
10672 mostly intended for debugging, but can be used as a stick-table entry to
10673 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010010674 32-bit hash is trivial to break. See also "crc32", "sdbm", "wt6" and the
10675 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020010676
Willy Tarreau97707872015-01-27 15:12:13 +010010677even
10678 Returns a boolean TRUE if the input value of type unsigned integer is even
10679 otherwise returns FALSE. It is functionally equivalent to "not,and(1),bool".
10680
Emeric Brunf399b0d2014-11-03 17:07:03 +010010681field(<index>,<delimiters>)
10682 Extracts the substring at the given index considering given delimiters from
10683 an input string. Indexes start at 1 and delimiters are a string formatted
10684 list of chars.
10685
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010686hex
10687 Converts a binary input sample to an hex string containing two hex digits per
10688 input byte. It is used to log or transfer hex dumps of some binary input data
10689 in a way that can be reliably transferred (eg: an SSL ID can be copied in a
10690 header).
Thierry FOURNIER2f49d6d2014-03-12 15:01:52 +010010691
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010692http_date([<offset>])
10693 Converts an integer supposed to contain a date since epoch to a string
10694 representing this date in a format suitable for use in HTTP header fields. If
10695 an offset value is specified, then it is a number of seconds that is added to
10696 the date before the conversion is operated. This is particularly useful to
10697 emit Date header fields, Expires values in responses when combined with a
10698 positive offset, or Last-Modified values when the offset is negative.
Willy Tarreau74ca5042013-06-11 23:12:07 +020010699
Willy Tarreaud9f316a2014-07-10 14:03:38 +020010700in_table(<table>)
10701 Uses the string representation of the input sample to perform a look up in
10702 the specified table. If the key is not found in the table, a boolean false
10703 is returned. Otherwise a boolean true is returned. This can be used to verify
10704 the presence of a certain key in a table tracking some elements (eg: whether
10705 or not a source IP address or an Authorization header was already seen).
10706
Willy Tarreauffcb2e42014-07-10 16:29:08 +020010707ipmask(<mask>)
10708 Apply a mask to an IPv4 address, and use the result for lookups and storage.
10709 This can be used to make all hosts within a certain mask to share the same
10710 table entries and as such use the same server. The mask can be passed in
10711 dotted form (eg: 255.255.255.0) or in CIDR form (eg: 24).
10712
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020010713json([<input-code>])
10714 Escapes the input string and produces an ASCII ouput string ready to use as a
10715 JSON string. The converter tries to decode the input string according to the
10716 <input-code> parameter. It can be "ascii", "utf8", "utf8s", "utf8"" or
10717 "utf8ps". The "ascii" decoder never fails. The "utf8" decoder detects 3 types
10718 of errors:
10719 - bad UTF-8 sequence (lone continuation byte, bad number of continuation
10720 bytes, ...)
10721 - invalid range (the decoded value is within a UTF-8 prohibited range),
10722 - code overlong (the value is encoded with more bytes than necessary).
10723
10724 The UTF-8 JSON encoding can produce a "too long value" error when the UTF-8
10725 character is greater than 0xffff because the JSON string escape specification
10726 only authorizes 4 hex digits for the value encoding. The UTF-8 decoder exists
10727 in 4 variants designated by a combination of two suffix letters : "p" for
10728 "permissive" and "s" for "silently ignore". The behaviors of the decoders
10729 are :
10730 - "ascii" : never fails ;
10731 - "utf8" : fails on any detected errors ;
10732 - "utf8s" : never fails, but removes characters corresponding to errors ;
10733 - "utf8p" : accepts and fixes the overlong errors, but fails on any other
10734 error ;
10735 - "utf8ps" : never fails, accepts and fixes the overlong errors, but removes
10736 characters corresponding to the other errors.
10737
10738 This converter is particularly useful for building properly escaped JSON for
10739 logging to servers which consume JSON-formated traffic logs.
10740
10741 Example:
10742 capture request header user-agent len 150
10743 capture request header Host len 15
10744 log-format {"ip":"%[src]","user-agent":"%[capture.req.hdr(1),json]"}
10745
10746 Input request from client 127.0.0.1:
10747 GET / HTTP/1.0
10748 User-Agent: Very "Ugly" UA 1/2
10749
10750 Output log:
10751 {"ip":"127.0.0.1","user-agent":"Very \"Ugly\" UA 1\/2"}
10752
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010753language(<value>[,<default>])
10754 Returns the value with the highest q-factor from a list as extracted from the
10755 "accept-language" header using "req.fhdr". Values with no q-factor have a
10756 q-factor of 1. Values with a q-factor of 0 are dropped. Only values which
10757 belong to the list of semi-colon delimited <values> will be considered. The
10758 argument <value> syntax is "lang[;lang[;lang[;...]]]". If no value matches the
10759 given list and a default value is provided, it is returned. Note that language
10760 names may have a variant after a dash ('-'). If this variant is present in the
10761 list, it will be matched, but if it is not, only the base language is checked.
10762 The match is case-sensitive, and the output string is always one of those
10763 provided in arguments. The ordering of arguments is meaningless, only the
10764 ordering of the values in the request counts, as the first value among
10765 multiple sharing the same q-factor is used.
Thierry FOURNIERad903512014-04-11 17:51:01 +020010766
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010767 Example :
Thierry FOURNIERad903512014-04-11 17:51:01 +020010768
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010769 # this configuration switches to the backend matching a
10770 # given language based on the request :
Thierry FOURNIERad903512014-04-11 17:51:01 +020010771
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010772 acl es req.fhdr(accept-language),language(es;fr;en) -m str es
10773 acl fr req.fhdr(accept-language),language(es;fr;en) -m str fr
10774 acl en req.fhdr(accept-language),language(es;fr;en) -m str en
10775 use_backend spanish if es
10776 use_backend french if fr
10777 use_backend english if en
10778 default_backend choose_your_language
Thierry FOURNIERad903512014-04-11 17:51:01 +020010779
Willy Tarreauffcb2e42014-07-10 16:29:08 +020010780lower
10781 Convert a string sample to lower case. This can only be placed after a string
10782 sample fetch function or after a transformation keyword returning a string
10783 type. The result is of type string.
10784
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020010785ltime(<format>[,<offset>])
10786 Converts an integer supposed to contain a date since epoch to a string
10787 representing this date in local time using a format defined by the <format>
10788 string using strftime(3). The purpose is to allow any date format to be used
10789 in logs. An optional <offset> in seconds may be applied to the input date
10790 (positive or negative). See the strftime() man page for the format supported
10791 by your operating system. See also the utime converter.
10792
10793 Example :
10794
10795 # Emit two colons, one with the local time and another with ip:port
10796 # Eg: 20140710162350 127.0.0.1:57325
10797 log-format %[date,ltime(%Y%m%d%H%M%S)]\ %ci:%cp
10798
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010799map(<map_file>[,<default_value>])
10800map_<match_type>(<map_file>[,<default_value>])
10801map_<match_type>_<output_type>(<map_file>[,<default_value>])
10802 Search the input value from <map_file> using the <match_type> matching method,
10803 and return the associated value converted to the type <output_type>. If the
10804 input value cannot be found in the <map_file>, the converter returns the
10805 <default_value>. If the <default_value> is not set, the converter fails and
10806 acts as if no input value could be fetched. If the <match_type> is not set, it
10807 defaults to "str". Likewise, if the <output_type> is not set, it defaults to
10808 "str". For convenience, the "map" keyword is an alias for "map_str" and maps a
10809 string to another string.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010010810
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010811 It is important to avoid overlapping between the keys : IP addresses and
10812 strings are stored in trees, so the first of the finest match will be used.
10813 Other keys are stored in lists, so the first matching occurrence will be used.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010010814
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010815 The following array contains the list of all map functions avalaible sorted by
10816 input type, match type and output type.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010010817
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010818 input type | match method | output type str | output type int | output type ip
10819 -----------+--------------+-----------------+-----------------+---------------
10820 str | str | map_str | map_str_int | map_str_ip
10821 -----------+--------------+-----------------+-----------------+---------------
Willy Tarreau787a4c02014-05-10 07:55:30 +020010822 str | beg | map_beg | map_beg_int | map_end_ip
10823 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010824 str | sub | map_sub | map_sub_int | map_sub_ip
10825 -----------+--------------+-----------------+-----------------+---------------
10826 str | dir | map_dir | map_dir_int | map_dir_ip
10827 -----------+--------------+-----------------+-----------------+---------------
10828 str | dom | map_dom | map_dom_int | map_dom_ip
10829 -----------+--------------+-----------------+-----------------+---------------
10830 str | end | map_end | map_end_int | map_end_ip
10831 -----------+--------------+-----------------+-----------------+---------------
10832 str | reg | map_reg | map_reg_int | map_reg_ip
10833 -----------+--------------+-----------------+-----------------+---------------
10834 int | int | map_int | map_int_int | map_int_ip
10835 -----------+--------------+-----------------+-----------------+---------------
10836 ip | ip | map_ip | map_ip_int | map_ip_ip
10837 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010010838
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010839 The file contains one key + value per line. Lines which start with '#' are
10840 ignored, just like empty lines. Leading tabs and spaces are stripped. The key
10841 is then the first "word" (series of non-space/tabs characters), and the value
10842 is what follows this series of space/tab till the end of the line excluding
10843 trailing spaces/tabs.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010010844
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010845 Example :
10846
10847 # this is a comment and is ignored
10848 2.22.246.0/23 United Kingdom \n
10849 <-><-----------><--><------------><---->
10850 | | | | `- trailing spaces ignored
10851 | | | `---------- value
10852 | | `-------------------- middle spaces ignored
10853 | `---------------------------- key
10854 `------------------------------------ leading spaces ignored
10855
Willy Tarreau97707872015-01-27 15:12:13 +010010856mod(<value>)
10857 Divides the input value of type unsigned integer by <value>, and returns the
10858 remainder as an unsigned integer. If <value> is null, then zero is returned.
10859
10860mul(<value>)
10861 Multiplies the input value of type unsigned integer by <value>, and returns
10862 the product as an unsigned integer. In case of overflow, the higher bits are
10863 lost, leading to seemingly strange values.
10864
10865neg
10866 Takes the input value of type unsigned integer, computes the opposite value,
10867 and returns the remainder as an unsigned integer. 0 is identity. This
10868 operator is provided for reversed subtracts : in order to subtract the input
10869 from a constant, simply perform a "neg,add(value)".
10870
10871not
10872 Returns a boolean FALSE if the input value of type unsigned integer is
10873 non-null, otherwise returns TRUE. Used in conjunction with and(), it can be
10874 used to report true/false for bit testing on input values (eg: verify the
10875 absence of a flag).
10876
10877odd
10878 Returns a boolean TRUE if the input value of type unsigned integer is odd
10879 otherwise returns FALSE. It is functionally equivalent to "and(1),bool".
10880
10881or(<value>)
10882 Performs a bitwise "OR" between <value> and the input value of type unsigned
10883 integer, and returns the result as an unsigned integer.
10884
Willy Tarreauc4dc3502015-01-23 20:39:28 +010010885regsub(<regex>,<subst>[,<flags>])
Willy Tarreau7eda8492015-01-20 19:47:06 +010010886 Applies a regex-based substitution to the input string. It does the same
10887 operation as the well-known "sed" utility with "s/<regex>/<subst>/". By
10888 default it will replace in the input string the first occurrence of the
10889 largest part matching the regular expression <regex> with the substitution
10890 string <subst>. It is possible to replace all occurrences instead by adding
10891 the flag "g" in the third argument <flags>. It is also possible to make the
10892 regex case insensitive by adding the flag "i" in <flags>. Since <flags> is a
10893 string, it is made up from the concatenation of all desired flags. Thus if
10894 both "i" and "g" are desired, using "gi" or "ig" will have the same effect.
10895 It is important to note that due to the current limitations of the
10896 configuration parser, some characters such as closing parenthesis or comma
10897 are not possible to use in the arguments. The first use of this converter is
10898 to replace certain characters or sequence of characters with other ones.
10899
10900 Example :
10901
10902 # de-duplicate "/" in header "x-path".
10903 # input: x-path: /////a///b/c/xzxyz/
10904 # output: x-path: /a/b/c/xzxyz/
10905 http-request set-header x-path %[hdr(x-path),regsub(/+,/,g)]
10906
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020010907sdbm([<avalanche>])
10908 Hashes a binary input sample into an unsigned 32-bit quantity using the SDBM
10909 hash function. Optionally, it is possible to apply a full avalanche hash
10910 function to the output if the optional <avalanche> argument equals 1. This
10911 converter uses the same functions as used by the various hash-based load
10912 balancing algorithms, so it will provide exactly the same results. It is
10913 mostly intended for debugging, but can be used as a stick-table entry to
10914 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010010915 32-bit hash is trivial to break. See also "crc32", "djb2", "wt6" and the
10916 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020010917
Willy Tarreau97707872015-01-27 15:12:13 +010010918sub(<value>)
10919 Subtracts <value> from the input value of type unsigned integer, and returns
10920 the result as an unsigned integer. Note: in order to subtract the input from
10921 a constant, simply perform a "neg,add(value)".
10922
Willy Tarreaud9f316a2014-07-10 14:03:38 +020010923table_bytes_in_rate(<table>)
10924 Uses the string representation of the input sample to perform a look up in
10925 the specified table. If the key is not found in the table, integer value zero
10926 is returned. Otherwise the converter returns the average client-to-server
10927 bytes rate associated with the input sample in the designated table, measured
10928 in amount of bytes over the period configured in the table. See also the
10929 sc_bytes_in_rate sample fetch keyword.
10930
10931
10932table_bytes_out_rate(<table>)
10933 Uses the string representation of the input sample to perform a look up in
10934 the specified table. If the key is not found in the table, integer value zero
10935 is returned. Otherwise the converter returns the average server-to-client
10936 bytes rate associated with the input sample in the designated table, measured
10937 in amount of bytes over the period configured in the table. See also the
10938 sc_bytes_out_rate sample fetch keyword.
10939
10940table_conn_cnt(<table>)
10941 Uses the string representation of the input sample to perform a look up in
10942 the specified table. If the key is not found in the table, integer value zero
10943 is returned. Otherwise the converter returns the cumulated amount of incoming
10944 connections associated with the input sample in the designated table. See
10945 also the sc_conn_cnt sample fetch keyword.
10946
10947table_conn_cur(<table>)
10948 Uses the string representation of the input sample to perform a look up in
10949 the specified table. If the key is not found in the table, integer value zero
10950 is returned. Otherwise the converter returns the current amount of concurrent
10951 tracked connections associated with the input sample in the designated table.
10952 See also the sc_conn_cur sample fetch keyword.
10953
10954table_conn_rate(<table>)
10955 Uses the string representation of the input sample to perform a look up in
10956 the specified table. If the key is not found in the table, integer value zero
10957 is returned. Otherwise the converter returns the average incoming connection
10958 rate associated with the input sample in the designated table. See also the
10959 sc_conn_rate sample fetch keyword.
10960
10961table_gpc0(<table>)
10962 Uses the string representation of the input sample to perform a look up in
10963 the specified table. If the key is not found in the table, integer value zero
10964 is returned. Otherwise the converter returns the current value of the first
10965 general purpose counter associated with the input sample in the designated
10966 table. See also the sc_get_gpc0 sample fetch keyword.
10967
10968table_gpc0_rate(<table>)
10969 Uses the string representation of the input sample to perform a look up in
10970 the specified table. If the key is not found in the table, integer value zero
10971 is returned. Otherwise the converter returns the frequency which the gpc0
10972 counter was incremented over the configured period in the table, associated
10973 with the input sample in the designated table. See also the sc_get_gpc0_rate
10974 sample fetch keyword.
10975
10976table_http_err_cnt(<table>)
10977 Uses the string representation of the input sample to perform a look up in
10978 the specified table. If the key is not found in the table, integer value zero
10979 is returned. Otherwise the converter returns the cumulated amount of HTTP
10980 errors associated with the input sample in the designated table. See also the
10981 sc_http_err_cnt sample fetch keyword.
10982
10983table_http_err_rate(<table>)
10984 Uses the string representation of the input sample to perform a look up in
10985 the specified table. If the key is not found in the table, integer value zero
10986 is returned. Otherwise the average rate of HTTP errors associated with the
10987 input sample in the designated table, measured in amount of errors over the
10988 period configured in the table. See also the sc_http_err_rate sample fetch
10989 keyword.
10990
10991table_http_req_cnt(<table>)
10992 Uses the string representation of the input sample to perform a look up in
10993 the specified table. If the key is not found in the table, integer value zero
10994 is returned. Otherwise the converter returns the cumulated amount of HTTP
10995 requests associated with the input sample in the designated table. See also
10996 the sc_http_req_cnt sample fetch keyword.
10997
10998table_http_req_rate(<table>)
10999 Uses the string representation of the input sample to perform a look up in
11000 the specified table. If the key is not found in the table, integer value zero
11001 is returned. Otherwise the average rate of HTTP requests associated with the
11002 input sample in the designated table, measured in amount of requests over the
11003 period configured in the table. See also the sc_http_req_rate sample fetch
11004 keyword.
11005
11006table_kbytes_in(<table>)
11007 Uses the string representation of the input sample to perform a look up in
11008 the specified table. If the key is not found in the table, integer value zero
11009 is returned. Otherwise the converter returns the cumulated amount of client-
11010 to-server data associated with the input sample in the designated table,
11011 measured in kilobytes. The test is currently performed on 32-bit integers,
11012 which limits values to 4 terabytes. See also the sc_kbytes_in sample fetch
11013 keyword.
11014
11015table_kbytes_out(<table>)
11016 Uses the string representation of the input sample to perform a look up in
11017 the specified table. If the key is not found in the table, integer value zero
11018 is returned. Otherwise the converter returns the cumulated amount of server-
11019 to-client data associated with the input sample in the designated table,
11020 measured in kilobytes. The test is currently performed on 32-bit integers,
11021 which limits values to 4 terabytes. See also the sc_kbytes_out sample fetch
11022 keyword.
11023
11024table_server_id(<table>)
11025 Uses the string representation of the input sample to perform a look up in
11026 the specified table. If the key is not found in the table, integer value zero
11027 is returned. Otherwise the converter returns the server ID associated with
11028 the input sample in the designated table. A server ID is associated to a
11029 sample by a "stick" rule when a connection to a server succeeds. A server ID
11030 zero means that no server is associated with this key.
11031
11032table_sess_cnt(<table>)
11033 Uses the string representation of the input sample to perform a look up in
11034 the specified table. If the key is not found in the table, integer value zero
11035 is returned. Otherwise the converter returns the cumulated amount of incoming
11036 sessions associated with the input sample in the designated table. Note that
11037 a session here refers to an incoming connection being accepted by the
11038 "tcp-request connection" rulesets. See also the sc_sess_cnt sample fetch
11039 keyword.
11040
11041table_sess_rate(<table>)
11042 Uses the string representation of the input sample to perform a look up in
11043 the specified table. If the key is not found in the table, integer value zero
11044 is returned. Otherwise the converter returns the average incoming session
11045 rate associated with the input sample in the designated table. Note that a
11046 session here refers to an incoming connection being accepted by the
11047 "tcp-request connection" rulesets. See also the sc_sess_rate sample fetch
11048 keyword.
11049
11050table_trackers(<table>)
11051 Uses the string representation of the input sample to perform a look up in
11052 the specified table. If the key is not found in the table, integer value zero
11053 is returned. Otherwise the converter returns the current amount of concurrent
11054 connections tracking the same key as the input sample in the designated
11055 table. It differs from table_conn_cur in that it does not rely on any stored
11056 information but on the table's reference count (the "use" value which is
11057 returned by "show table" on the CLI). This may sometimes be more suited for
11058 layer7 tracking. It can be used to tell a server how many concurrent
11059 connections there are from a given address for example. See also the
11060 sc_trackers sample fetch keyword.
11061
Willy Tarreauffcb2e42014-07-10 16:29:08 +020011062upper
11063 Convert a string sample to upper case. This can only be placed after a string
11064 sample fetch function or after a transformation keyword returning a string
11065 type. The result is of type string.
11066
Thierry FOURNIER82ff3c92015-05-07 15:46:20 +020011067url_dec
11068 Takes an url-encoded string provided as input and returns the decoded
11069 version as output. The input and the output are of type string.
11070
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020011071utime(<format>[,<offset>])
11072 Converts an integer supposed to contain a date since epoch to a string
11073 representing this date in UTC time using a format defined by the <format>
11074 string using strftime(3). The purpose is to allow any date format to be used
11075 in logs. An optional <offset> in seconds may be applied to the input date
11076 (positive or negative). See the strftime() man page for the format supported
11077 by your operating system. See also the ltime converter.
11078
11079 Example :
11080
11081 # Emit two colons, one with the UTC time and another with ip:port
11082 # Eg: 20140710162350 127.0.0.1:57325
11083 log-format %[date,utime(%Y%m%d%H%M%S)]\ %ci:%cp
11084
Emeric Brunc9a0f6d2014-11-25 14:09:01 +010011085word(<index>,<delimiters>)
11086 Extracts the nth word considering given delimiters from an input string.
11087 Indexes start at 1 and delimiters are a string formatted list of chars.
11088
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020011089wt6([<avalanche>])
11090 Hashes a binary input sample into an unsigned 32-bit quantity using the WT6
11091 hash function. Optionally, it is possible to apply a full avalanche hash
11092 function to the output if the optional <avalanche> argument equals 1. This
11093 converter uses the same functions as used by the various hash-based load
11094 balancing algorithms, so it will provide exactly the same results. It is
11095 mostly intended for debugging, but can be used as a stick-table entry to
11096 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010011097 32-bit hash is trivial to break. See also "crc32", "djb2", "sdbm", and the
11098 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020011099
Willy Tarreau97707872015-01-27 15:12:13 +010011100xor(<value>)
11101 Performs a bitwise "XOR" (exclusive OR) between <value> and the input value
11102 of type unsigned integer, and returns the result as an unsigned integer.
11103
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010011104
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200111057.3.2. Fetching samples from internal states
Willy Tarreau74ca5042013-06-11 23:12:07 +020011106--------------------------------------------
11107
11108A first set of sample fetch methods applies to internal information which does
11109not even relate to any client information. These ones are sometimes used with
11110"monitor-fail" directives to report an internal status to external watchers.
11111The sample fetch methods described in this section are usable anywhere.
11112
11113always_false : boolean
11114 Always returns the boolean "false" value. It may be used with ACLs as a
11115 temporary replacement for another one when adjusting configurations.
11116
11117always_true : boolean
11118 Always returns the boolean "true" value. It may be used with ACLs as a
11119 temporary replacement for another one when adjusting configurations.
11120
11121avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010011122 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +020011123 divided by the number of active servers. The current backend is used if no
11124 backend is specified. This is very similar to "queue" except that the size of
11125 the farm is considered, in order to give a more accurate measurement of the
11126 time it may take for a new connection to be processed. The main usage is with
11127 ACL to return a sorry page to new users when it becomes certain they will get
11128 a degraded service, or to pass to the backend servers in a header so that
11129 they decide to work in degraded mode or to disable some functions to speed up
11130 the processing a bit. Note that in the event there would not be any active
11131 server anymore, twice the number of queued connections would be considered as
11132 the measured value. This is a fair estimate, as we expect one server to get
11133 back soon anyway, but we still prefer to send new traffic to another backend
11134 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
11135 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +010011136
Willy Tarreau74ca5042013-06-11 23:12:07 +020011137be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +020011138 Applies to the number of currently established connections on the backend,
11139 possibly including the connection being evaluated. If no backend name is
11140 specified, the current one is used. But it is also possible to check another
11141 backend. It can be used to use a specific farm when the nominal one is full.
11142 See also the "fe_conn", "queue" and "be_sess_rate" criteria.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011143
Willy Tarreau74ca5042013-06-11 23:12:07 +020011144be_sess_rate([<backend>]) : integer
11145 Returns an integer value corresponding to the sessions creation rate on the
11146 backend, in number of new sessions per second. This is used with ACLs to
11147 switch to an alternate backend when an expensive or fragile one reaches too
11148 high a session rate, or to limit abuse of service (eg. prevent sucking of an
11149 online dictionary). It can also be useful to add this element to logs using a
11150 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011151
11152 Example :
11153 # Redirect to an error page if the dictionary is requested too often
11154 backend dynamic
11155 mode http
11156 acl being_scanned be_sess_rate gt 100
11157 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +010011158
Willy Tarreau74ca5042013-06-11 23:12:07 +020011159connslots([<backend>]) : integer
11160 Returns an integer value corresponding to the number of connection slots
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011161 still available in the backend, by totaling the maximum amount of
Willy Tarreau74ca5042013-06-11 23:12:07 +020011162 connections on all servers and the maximum queue size. This is probably only
11163 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -050011164
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080011165 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +020011166 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080011167 usage; see "use_backend" keyword) can be redirected to a different backend.
11168
Willy Tarreau55165fe2009-05-10 12:02:55 +020011169 'connslots' = number of available server connection slots, + number of
11170 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080011171
Willy Tarreaua36af912009-10-10 12:02:45 +020011172 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +020011173 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +020011174 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +020011175 you want to be able to differentiate between different backends, and their
11176 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +020011177 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +020011178 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080011179
Willy Tarreau55165fe2009-05-10 12:02:55 +020011180 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
11181 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +020011182 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +020011183 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080011184
Willy Tarreau6236d3a2013-07-25 14:28:25 +020011185date([<offset>]) : integer
11186 Returns the current date as the epoch (number of seconds since 01/01/1970).
11187 If an offset value is specified, then it is a number of seconds that is added
11188 to the current date before returning the value. This is particularly useful
11189 to compute relative dates, as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +020011190 It is useful combined with the http_date converter.
11191
11192 Example :
11193
11194 # set an expires header to now+1 hour in every response
11195 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +020011196
Willy Tarreau595ec542013-06-12 21:34:28 +020011197env(<name>) : string
11198 Returns a string containing the value of environment variable <name>. As a
11199 reminder, environment variables are per-process and are sampled when the
11200 process starts. This can be useful to pass some information to a next hop
11201 server, or with ACLs to take specific action when the process is started a
11202 certain way.
11203
11204 Examples :
11205 # Pass the Via header to next hop with the local hostname in it
11206 http-request add-header Via 1.1\ %[env(HOSTNAME)]
11207
11208 # reject cookie-less requests when the STOP environment variable is set
11209 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
11210
Willy Tarreau74ca5042013-06-11 23:12:07 +020011211fe_conn([<frontend>]) : integer
11212 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +010011213 possibly including the connection being evaluated. If no frontend name is
11214 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020011215 frontend. It can be used to return a sorry page before hard-blocking, or to
11216 use a specific backend to drain new requests when the farm is considered
11217 full. This is mostly used with ACLs but can also be used to pass some
11218 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
11219 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +020011220
Willy Tarreau74ca5042013-06-11 23:12:07 +020011221fe_sess_rate([<frontend>]) : integer
11222 Returns an integer value corresponding to the sessions creation rate on the
11223 frontend, in number of new sessions per second. This is used with ACLs to
11224 limit the incoming session rate to an acceptable range in order to prevent
11225 abuse of service at the earliest moment, for example when combined with other
11226 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
11227 down below the limit. It can also be useful to add this element to logs using
11228 a log-format directive. See also the "rate-limit sessions" directive for use
11229 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010011230
11231 Example :
11232 # This frontend limits incoming mails to 10/s with a max of 100
11233 # concurrent connections. We accept any connection below 10/s, and
11234 # force excess clients to wait for 100 ms. Since clients are limited to
11235 # 100 max, there cannot be more than 10 incoming mails per second.
11236 frontend mail
11237 bind :25
11238 mode tcp
11239 maxconn 100
11240 acl too_fast fe_sess_rate ge 10
11241 tcp-request inspect-delay 100ms
11242 tcp-request content accept if ! too_fast
11243 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +010011244
Willy Tarreau0f30d262014-11-24 16:02:05 +010011245nbproc : integer
11246 Returns an integer value corresponding to the number of processes that were
11247 started (it equals the global "nbproc" setting). This is useful for logging
11248 and debugging purposes.
11249
Willy Tarreau74ca5042013-06-11 23:12:07 +020011250nbsrv([<backend>]) : integer
11251 Returns an integer value corresponding to the number of usable servers of
11252 either the current backend or the named backend. This is mostly used with
11253 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +010011254 switch to an alternate backend when the number of servers is too low to
11255 to handle some load. It is useful to report a failure when combined with
11256 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +010011257
Willy Tarreau0f30d262014-11-24 16:02:05 +010011258proc : integer
11259 Returns an integer value corresponding to the position of the process calling
11260 the function, between 1 and global.nbproc. This is useful for logging and
11261 debugging purposes.
11262
Willy Tarreau74ca5042013-06-11 23:12:07 +020011263queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010011264 Returns the total number of queued connections of the designated backend,
11265 including all the connections in server queues. If no backend name is
11266 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020011267 one. This is useful with ACLs or to pass statistics to backend servers. This
11268 can be used to take actions when queuing goes above a known level, generally
11269 indicating a surge of traffic or a massive slowdown on the servers. One
11270 possible action could be to reject new users but still accept old ones. See
11271 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
11272
Willy Tarreau84310e22014-02-14 11:59:04 +010011273rand([<range>]) : integer
11274 Returns a random integer value within a range of <range> possible values,
11275 starting at zero. If the range is not specified, it defaults to 2^32, which
11276 gives numbers between 0 and 4294967295. It can be useful to pass some values
11277 needed to take some routing decisions for example, or just for debugging
11278 purposes. This random must not be used for security purposes.
11279
Willy Tarreau74ca5042013-06-11 23:12:07 +020011280srv_conn([<backend>/]<server>) : integer
11281 Returns an integer value corresponding to the number of currently established
11282 connections on the designated server, possibly including the connection being
11283 evaluated. If <backend> is omitted, then the server is looked up in the
11284 current backend. It can be used to use a specific farm when one server is
11285 full, or to inform the server about our view of the number of active
11286 connections with it. See also the "fe_conn", "be_conn" and "queue" fetch
11287 methods.
11288
11289srv_is_up([<backend>/]<server>) : boolean
11290 Returns true when the designated server is UP, and false when it is either
11291 DOWN or in maintenance mode. If <backend> is omitted, then the server is
11292 looked up in the current backend. It is mainly used to take action based on
11293 an external status reported via a health check (eg: a geographical site's
11294 availability). Another possible use which is more of a hack consists in
11295 using dummy servers as boolean variables that can be enabled or disabled from
11296 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
11297
11298srv_sess_rate([<backend>/]<server>) : integer
11299 Returns an integer corresponding to the sessions creation rate on the
11300 designated server, in number of new sessions per second. If <backend> is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011301 omitted, then the server is looked up in the current backend. This is mostly
Willy Tarreau74ca5042013-06-11 23:12:07 +020011302 used with ACLs but can make sense with logs too. This is used to switch to an
11303 alternate backend when an expensive or fragile one reaches too high a session
11304 rate, or to limit abuse of service (eg. prevent latent requests from
11305 overloading servers).
11306
11307 Example :
11308 # Redirect to a separate back
11309 acl srv1_full srv_sess_rate(be1/srv1) gt 50
11310 acl srv2_full srv_sess_rate(be1/srv2) gt 50
11311 use_backend be2 if srv1_full or srv2_full
11312
Willy Tarreau0f30d262014-11-24 16:02:05 +010011313stopping : boolean
11314 Returns TRUE if the process calling the function is currently stopping. This
11315 can be useful for logging, or for relaxing certain checks or helping close
11316 certain connections upon graceful shutdown.
11317
Willy Tarreau74ca5042013-06-11 23:12:07 +020011318table_avl([<table>]) : integer
11319 Returns the total number of available entries in the current proxy's
11320 stick-table or in the designated stick-table. See also table_cnt.
11321
11322table_cnt([<table>]) : integer
11323 Returns the total number of entries currently in use in the current proxy's
11324 stick-table or in the designated stick-table. See also src_conn_cnt and
11325 table_avl for other entry counting methods.
11326
11327
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200113287.3.3. Fetching samples at Layer 4
Willy Tarreau74ca5042013-06-11 23:12:07 +020011329----------------------------------
11330
11331The layer 4 usually describes just the transport layer which in haproxy is
11332closest to the connection, where no content is yet made available. The fetch
11333methods described here are usable as low as the "tcp-request connection" rule
11334sets unless they require some future information. Those generally include
11335TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011336the incoming connection. For retrieving a value from a sticky counters, the
11337counter number can be explicitly set as 0, 1, or 2 using the pre-defined
11338"sc0_", "sc1_", or "sc2_" prefix, or it can be specified as the first integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011339argument when using the "sc_" prefix. An optional table may be specified with
11340the "sc*" form, in which case the currently tracked key will be looked up into
11341this alternate table instead of the table currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011342
11343be_id : integer
11344 Returns an integer containing the current backend's id. It can be used in
11345 frontends with responses to check which backend processed the request.
11346
11347dst : ip
11348 This is the destination IPv4 address of the connection on the client side,
11349 which is the address the client connected to. It can be useful when running
11350 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
11351 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
11352 RFC 4291.
11353
11354dst_conn : integer
11355 Returns an integer value corresponding to the number of currently established
11356 connections on the same socket including the one being evaluated. It is
11357 normally used with ACLs but can as well be used to pass the information to
11358 servers in an HTTP header or in logs. It can be used to either return a sorry
11359 page before hard-blocking, or to use a specific backend to drain new requests
11360 when the socket is considered saturated. This offers the ability to assign
11361 different limits to different listening ports or addresses. See also the
11362 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011363
Willy Tarreau74ca5042013-06-11 23:12:07 +020011364dst_port : integer
11365 Returns an integer value corresponding to the destination TCP port of the
11366 connection on the client side, which is the port the client connected to.
11367 This might be used when running in transparent mode, when assigning dynamic
11368 ports to some clients for a whole application session, to stick all users to
11369 a same server, or to pass the destination port information to a server using
11370 an HTTP header.
11371
11372fe_id : integer
11373 Returns an integer containing the current frontend's id. It can be used in
11374 backends to check from which backend it was called, or to stick all users
11375 coming via a same frontend to the same server.
11376
Cyril Bonté62ba8702014-04-22 23:52:25 +020011377sc_bytes_in_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011378sc0_bytes_in_rate([<table>]) : integer
11379sc1_bytes_in_rate([<table>]) : integer
11380sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011381 Returns the average client-to-server bytes rate from the currently tracked
11382 counters, measured in amount of bytes over the period configured in the
11383 table. See also src_bytes_in_rate.
11384
Cyril Bonté62ba8702014-04-22 23:52:25 +020011385sc_bytes_out_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011386sc0_bytes_out_rate([<table>]) : integer
11387sc1_bytes_out_rate([<table>]) : integer
11388sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011389 Returns the average server-to-client bytes rate from the currently tracked
11390 counters, measured in amount of bytes over the period configured in the
11391 table. See also src_bytes_out_rate.
11392
Cyril Bonté62ba8702014-04-22 23:52:25 +020011393sc_clr_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011394sc0_clr_gpc0([<table>]) : integer
11395sc1_clr_gpc0([<table>]) : integer
11396sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +020011397 Clears the first General Purpose Counter associated to the currently tracked
11398 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +010011399 stored value is zero, so first invocation will always return zero. This is
11400 typically used as a second ACL in an expression in order to mark a connection
11401 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020011402
11403 # block if 5 consecutive requests continue to come faster than 10 sess
11404 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020011405 acl abuse sc0_http_req_rate gt 10
11406 acl kill sc0_inc_gpc0 gt 5
11407 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020011408 tcp-request connection accept if !abuse save
11409 tcp-request connection reject if abuse kill
11410
Cyril Bonté62ba8702014-04-22 23:52:25 +020011411sc_conn_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011412sc0_conn_cnt([<table>]) : integer
11413sc1_conn_cnt([<table>]) : integer
11414sc2_conn_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011415 Returns the cumulated number of incoming connections from currently tracked
11416 counters. See also src_conn_cnt.
11417
Cyril Bonté62ba8702014-04-22 23:52:25 +020011418sc_conn_cur(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011419sc0_conn_cur([<table>]) : integer
11420sc1_conn_cur([<table>]) : integer
11421sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011422 Returns the current amount of concurrent connections tracking the same
11423 tracked counters. This number is automatically incremented when tracking
11424 begins and decremented when tracking stops. See also src_conn_cur.
11425
Cyril Bonté62ba8702014-04-22 23:52:25 +020011426sc_conn_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011427sc0_conn_rate([<table>]) : integer
11428sc1_conn_rate([<table>]) : integer
11429sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011430 Returns the average connection rate from the currently tracked counters,
11431 measured in amount of connections over the period configured in the table.
11432 See also src_conn_rate.
11433
Cyril Bonté62ba8702014-04-22 23:52:25 +020011434sc_get_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011435sc0_get_gpc0([<table>]) : integer
11436sc1_get_gpc0([<table>]) : integer
11437sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011438 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011439 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020011440
Cyril Bonté62ba8702014-04-22 23:52:25 +020011441sc_gpc0_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011442sc0_gpc0_rate([<table>]) : integer
11443sc1_gpc0_rate([<table>]) : integer
11444sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020011445 Returns the average increment rate of the first General Purpose Counter
11446 associated to the currently tracked counters. It reports the frequency
11447 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011448 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
11449 that the "gpc0_rate" counter must be stored in the stick-table for a value to
11450 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +020011451
Cyril Bonté62ba8702014-04-22 23:52:25 +020011452sc_http_err_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011453sc0_http_err_cnt([<table>]) : integer
11454sc1_http_err_cnt([<table>]) : integer
11455sc2_http_err_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011456 Returns the cumulated number of HTTP errors from the currently tracked
11457 counters. This includes the both request errors and 4xx error responses.
11458 See also src_http_err_cnt.
11459
Cyril Bonté62ba8702014-04-22 23:52:25 +020011460sc_http_err_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011461sc0_http_err_rate([<table>]) : integer
11462sc1_http_err_rate([<table>]) : integer
11463sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011464 Returns the average rate of HTTP errors from the currently tracked counters,
11465 measured in amount of errors over the period configured in the table. This
11466 includes the both request errors and 4xx error responses. See also
11467 src_http_err_rate.
11468
Cyril Bonté62ba8702014-04-22 23:52:25 +020011469sc_http_req_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011470sc0_http_req_cnt([<table>]) : integer
11471sc1_http_req_cnt([<table>]) : integer
11472sc2_http_req_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011473 Returns the cumulated number of HTTP requests from the currently tracked
11474 counters. This includes every started request, valid or not. See also
11475 src_http_req_cnt.
11476
Cyril Bonté62ba8702014-04-22 23:52:25 +020011477sc_http_req_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011478sc0_http_req_rate([<table>]) : integer
11479sc1_http_req_rate([<table>]) : integer
11480sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011481 Returns the average rate of HTTP requests from the currently tracked
11482 counters, measured in amount of requests over the period configured in
11483 the table. This includes every started request, valid or not. See also
11484 src_http_req_rate.
11485
Cyril Bonté62ba8702014-04-22 23:52:25 +020011486sc_inc_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011487sc0_inc_gpc0([<table>]) : integer
11488sc1_inc_gpc0([<table>]) : integer
11489sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011490 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +010011491 tracked counters, and returns its new value. Before the first invocation,
11492 the stored value is zero, so first invocation will increase it to 1 and will
11493 return 1. This is typically used as a second ACL in an expression in order
11494 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +020011495
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020011496 acl abuse sc0_http_req_rate gt 10
11497 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020011498 tcp-request connection reject if abuse kill
11499
Cyril Bonté62ba8702014-04-22 23:52:25 +020011500sc_kbytes_in(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011501sc0_kbytes_in([<table>]) : integer
11502sc1_kbytes_in([<table>]) : integer
11503sc2_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020011504 Returns the total amount of client-to-server data from the currently tracked
11505 counters, measured in kilobytes. The test is currently performed on 32-bit
11506 integers, which limits values to 4 terabytes. See also src_kbytes_in.
Willy Tarreaue9656522010-08-17 15:40:09 +020011507
Cyril Bonté62ba8702014-04-22 23:52:25 +020011508sc_kbytes_out(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011509sc0_kbytes_out([<table>]) : integer
11510sc1_kbytes_out([<table>]) : integer
11511sc2_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020011512 Returns the total amount of server-to-client data from the currently tracked
11513 counters, measured in kilobytes. The test is currently performed on 32-bit
11514 integers, which limits values to 4 terabytes. See also src_kbytes_out.
Willy Tarreaue9656522010-08-17 15:40:09 +020011515
Cyril Bonté62ba8702014-04-22 23:52:25 +020011516sc_sess_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011517sc0_sess_cnt([<table>]) : integer
11518sc1_sess_cnt([<table>]) : integer
11519sc2_sess_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011520 Returns the cumulated number of incoming connections that were transformed
11521 into sessions, which means that they were accepted by a "tcp-request
11522 connection" rule, from the currently tracked counters. A backend may count
11523 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -040011524 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +020011525 with the client. See also src_sess_cnt.
11526
Cyril Bonté62ba8702014-04-22 23:52:25 +020011527sc_sess_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011528sc0_sess_rate([<table>]) : integer
11529sc1_sess_rate([<table>]) : integer
11530sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011531 Returns the average session rate from the currently tracked counters,
11532 measured in amount of sessions over the period configured in the table. A
11533 session is a connection that got past the early "tcp-request connection"
11534 rules. A backend may count more sessions than connections because each
11535 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -040011536 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +020011537
Cyril Bonté62ba8702014-04-22 23:52:25 +020011538sc_tracked(<ctr>[,<table>]) : boolean
Willy Tarreau0f791d42013-07-23 19:56:43 +020011539sc0_tracked([<table>]) : boolean
11540sc1_tracked([<table>]) : boolean
11541sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +020011542 Returns true if the designated session counter is currently being tracked by
11543 the current session. This can be useful when deciding whether or not we want
11544 to set some values in a header passed to the server.
11545
Cyril Bonté62ba8702014-04-22 23:52:25 +020011546sc_trackers(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011547sc0_trackers([<table>]) : integer
11548sc1_trackers([<table>]) : integer
11549sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +010011550 Returns the current amount of concurrent connections tracking the same
11551 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020011552 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +010011553 that it does not rely on any stored information but on the table's reference
11554 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020011555 may sometimes be more suited for layer7 tracking. It can be used to tell a
11556 server how many concurrent connections there are from a given address for
11557 example.
Willy Tarreau2406db42012-12-09 12:16:43 +010011558
Willy Tarreau74ca5042013-06-11 23:12:07 +020011559so_id : integer
11560 Returns an integer containing the current listening socket's id. It is useful
11561 in frontends involving many "bind" lines, or to stick all users coming via a
11562 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011563
Willy Tarreau74ca5042013-06-11 23:12:07 +020011564src : ip
11565 This is the source IPv4 address of the client of the session. It is of type
11566 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
11567 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
11568 TCP-level source address which is used, and not the address of a client
11569 behind a proxy. However if the "accept-proxy" bind directive is used, it can
11570 be the address of a client behind another PROXY-protocol compatible component
11571 for all rule sets except "tcp-request connection" which sees the real address.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011572
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010011573 Example:
11574 # add an HTTP header in requests with the originating address' country
11575 http-request set-header X-Country %[src,map_ip(geoip.lst)]
11576
Willy Tarreau74ca5042013-06-11 23:12:07 +020011577src_bytes_in_rate([<table>]) : integer
11578 Returns the average bytes rate from the incoming connection's source address
11579 in the current proxy's stick-table or in the designated stick-table, measured
11580 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011581 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011582
Willy Tarreau74ca5042013-06-11 23:12:07 +020011583src_bytes_out_rate([<table>]) : integer
11584 Returns the average bytes rate to the incoming connection's source address in
11585 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +020011586 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011587 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011588
Willy Tarreau74ca5042013-06-11 23:12:07 +020011589src_clr_gpc0([<table>]) : integer
11590 Clears the first General Purpose Counter associated to the incoming
11591 connection's source address in the current proxy's stick-table or in the
11592 designated stick-table, and returns its previous value. If the address is not
11593 found, an entry is created and 0 is returned. This is typically used as a
11594 second ACL in an expression in order to mark a connection when a first ACL
11595 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020011596
11597 # block if 5 consecutive requests continue to come faster than 10 sess
11598 # per second, and reset the counter as soon as the traffic slows down.
11599 acl abuse src_http_req_rate gt 10
11600 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +010011601 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020011602 tcp-request connection accept if !abuse save
11603 tcp-request connection reject if abuse kill
11604
Willy Tarreau74ca5042013-06-11 23:12:07 +020011605src_conn_cnt([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020011606 Returns the cumulated number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +020011607 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020011608 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011609 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011610
Willy Tarreau74ca5042013-06-11 23:12:07 +020011611src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020011612 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +020011613 current incoming connection's source address in the current proxy's
11614 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011615 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011616
Willy Tarreau74ca5042013-06-11 23:12:07 +020011617src_conn_rate([<table>]) : integer
11618 Returns the average connection rate from the incoming connection's source
11619 address in the current proxy's stick-table or in the designated stick-table,
11620 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011621 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011622
Willy Tarreau74ca5042013-06-11 23:12:07 +020011623src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020011624 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +020011625 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020011626 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011627 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011628
Willy Tarreau74ca5042013-06-11 23:12:07 +020011629src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020011630 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +020011631 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +020011632 stick-table or in the designated stick-table. It reports the frequency
11633 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011634 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
11635 that the "gpc0_rate" counter must be stored in the stick-table for a value to
11636 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020011637
Willy Tarreau74ca5042013-06-11 23:12:07 +020011638src_http_err_cnt([<table>]) : integer
11639 Returns the cumulated number of HTTP errors from the incoming connection's
11640 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +020011641 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011642 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +020011643 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011644
Willy Tarreau74ca5042013-06-11 23:12:07 +020011645src_http_err_rate([<table>]) : integer
11646 Returns the average rate of HTTP errors from the incoming connection's source
11647 address in the current proxy's stick-table or in the designated stick-table,
11648 measured in amount of errors over the period configured in the table. This
11649 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011650 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011651
Willy Tarreau74ca5042013-06-11 23:12:07 +020011652src_http_req_cnt([<table>]) : integer
11653 Returns the cumulated number of HTTP requests from the incoming connection's
11654 source address in the current proxy's stick-table or in the designated stick-
11655 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011656 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011657
Willy Tarreau74ca5042013-06-11 23:12:07 +020011658src_http_req_rate([<table>]) : integer
11659 Returns the average rate of HTTP requests from the incoming connection's
11660 source address in the current proxy's stick-table or in the designated stick-
11661 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +020011662 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011663 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011664
Willy Tarreau74ca5042013-06-11 23:12:07 +020011665src_inc_gpc0([<table>]) : integer
11666 Increments the first General Purpose Counter associated to the incoming
11667 connection's source address in the current proxy's stick-table or in the
11668 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020011669 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011670 This is typically used as a second ACL in an expression in order to mark a
11671 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +020011672
11673 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +010011674 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020011675 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +020011676
Willy Tarreau74ca5042013-06-11 23:12:07 +020011677src_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020011678 Returns the total amount of data received from the incoming connection's
11679 source address in the current proxy's stick-table or in the designated
11680 stick-table, measured in kilobytes. If the address is not found, zero is
11681 returned. The test is currently performed on 32-bit integers, which limits
11682 values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011683
Willy Tarreau74ca5042013-06-11 23:12:07 +020011684src_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020011685 Returns the total amount of data sent to the incoming connection's source
11686 address in the current proxy's stick-table or in the designated stick-table,
11687 measured in kilobytes. If the address is not found, zero is returned. The
11688 test is currently performed on 32-bit integers, which limits values to 4
11689 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020011690
Willy Tarreau74ca5042013-06-11 23:12:07 +020011691src_port : integer
11692 Returns an integer value corresponding to the TCP source port of the
11693 connection on the client side, which is the port the client connected from.
11694 Usage of this function is very limited as modern protocols do not care much
11695 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010011696
Willy Tarreau74ca5042013-06-11 23:12:07 +020011697src_sess_cnt([<table>]) : integer
11698 Returns the cumulated number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +020011699 connection's source IPv4 address in the current proxy's stick-table or in the
11700 designated stick-table, that were transformed into sessions, which means that
11701 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011702 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011703
Willy Tarreau74ca5042013-06-11 23:12:07 +020011704src_sess_rate([<table>]) : integer
11705 Returns the average session rate from the incoming connection's source
11706 address in the current proxy's stick-table or in the designated stick-table,
11707 measured in amount of sessions over the period configured in the table. A
11708 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011709 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011710
Willy Tarreau74ca5042013-06-11 23:12:07 +020011711src_updt_conn_cnt([<table>]) : integer
11712 Creates or updates the entry associated to the incoming connection's source
11713 address in the current proxy's stick-table or in the designated stick-table.
11714 This table must be configured to store the "conn_cnt" data type, otherwise
11715 the match will be ignored. The current count is incremented by one, and the
11716 expiration timer refreshed. The updated count is returned, so this match
11717 can't return zero. This was used to reject service abusers based on their
11718 source address. Note: it is recommended to use the more complete "track-sc*"
11719 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020011720
11721 Example :
11722 # This frontend limits incoming SSH connections to 3 per 10 second for
11723 # each source address, and rejects excess connections until a 10 second
11724 # silence is observed. At most 20 addresses are tracked.
11725 listen ssh
11726 bind :22
11727 mode tcp
11728 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +020011729 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +020011730 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +020011731 server local 127.0.0.1:22
11732
Willy Tarreau74ca5042013-06-11 23:12:07 +020011733srv_id : integer
11734 Returns an integer containing the server's id when processing the response.
11735 While it's almost only used with ACLs, it may be used for logging or
11736 debugging.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +020011737
Hervé COMMOWICK35ed8012010-12-15 14:04:51 +010011738
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200117397.3.4. Fetching samples at Layer 5
Willy Tarreau74ca5042013-06-11 23:12:07 +020011740----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +020011741
Willy Tarreau74ca5042013-06-11 23:12:07 +020011742The layer 5 usually describes just the session layer which in haproxy is
11743closest to the session once all the connection handshakes are finished, but
11744when no content is yet made available. The fetch methods described here are
11745usable as low as the "tcp-request content" rule sets unless they require some
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011746future information. Those generally include the results of SSL negotiations.
Willy Tarreauc735a072011-03-29 00:57:02 +020011747
Emeric Brun645ae792014-04-30 14:21:06 +020011748ssl_bc : boolean
11749 Returns true when the back connection was made via an SSL/TLS transport
11750 layer and is locally deciphered. This means the outgoing connection was made
11751 other a server with the "ssl" option.
11752
11753ssl_bc_alg_keysize : integer
11754 Returns the symmetric cipher key size supported in bits when the outgoing
11755 connection was made over an SSL/TLS transport layer.
11756
11757ssl_bc_cipher : string
11758 Returns the name of the used cipher when the outgoing connection was made
11759 over an SSL/TLS transport layer.
11760
11761ssl_bc_protocol : string
11762 Returns the name of the used protocol when the outgoing connection was made
11763 over an SSL/TLS transport layer.
11764
Emeric Brunb73a9b02014-04-30 18:49:19 +020011765ssl_bc_unique_id : binary
Emeric Brun645ae792014-04-30 14:21:06 +020011766 When the outgoing connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020011767 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
11768 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
Emeric Brun645ae792014-04-30 14:21:06 +020011769
11770ssl_bc_session_id : binary
11771 Returns the SSL ID of the back connection when the outgoing connection was
11772 made over an SSL/TLS transport layer. It is useful to log if we want to know
11773 if session was reused or not.
11774
11775ssl_bc_use_keysize : integer
11776 Returns the symmetric cipher key size used in bits when the outgoing
11777 connection was made over an SSL/TLS transport layer.
11778
Willy Tarreau74ca5042013-06-11 23:12:07 +020011779ssl_c_ca_err : integer
11780 When the incoming connection was made over an SSL/TLS transport layer,
11781 returns the ID of the first error detected during verification of the client
11782 certificate at depth > 0, or 0 if no error was encountered during this
11783 verification process. Please refer to your SSL library's documentation to
11784 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +020011785
Willy Tarreau74ca5042013-06-11 23:12:07 +020011786ssl_c_ca_err_depth : integer
11787 When the incoming connection was made over an SSL/TLS transport layer,
11788 returns the depth in the CA chain of the first error detected during the
11789 verification of the client certificate. If no error is encountered, 0 is
11790 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +010011791
Emeric Brun43e79582014-10-29 19:03:26 +010011792ssl_c_der : binary
11793 Returns the DER formatted certificate presented by the client when the
11794 incoming connection was made over an SSL/TLS transport layer. When used for
11795 an ACL, the value(s) to match against can be passed in hexadecimal form.
11796
Willy Tarreau74ca5042013-06-11 23:12:07 +020011797ssl_c_err : integer
11798 When the incoming connection was made over an SSL/TLS transport layer,
11799 returns the ID of the first error detected during verification at depth 0, or
11800 0 if no error was encountered during this verification process. Please refer
11801 to your SSL library's documentation to find the exhaustive list of error
11802 codes.
Willy Tarreau62644772008-07-16 18:36:06 +020011803
Willy Tarreau74ca5042013-06-11 23:12:07 +020011804ssl_c_i_dn([<entry>[,<occ>]]) : string
11805 When the incoming connection was made over an SSL/TLS transport layer,
11806 returns the full distinguished name of the issuer of the certificate
11807 presented by the client when no <entry> is specified, or the value of the
11808 first given entry found from the beginning of the DN. If a positive/negative
11809 occurrence number is specified as the optional second argument, it returns
11810 the value of the nth given entry value from the beginning/end of the DN.
11811 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
11812 "ssl_c_i_dn(CN)" retrieves the common name.
Willy Tarreau62644772008-07-16 18:36:06 +020011813
Willy Tarreau74ca5042013-06-11 23:12:07 +020011814ssl_c_key_alg : string
11815 Returns the name of the algorithm used to generate the key of the certificate
11816 presented by the client when the incoming connection was made over an SSL/TLS
11817 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +020011818
Willy Tarreau74ca5042013-06-11 23:12:07 +020011819ssl_c_notafter : string
11820 Returns the end date presented by the client as a formatted string
11821 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
11822 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +020011823
Willy Tarreau74ca5042013-06-11 23:12:07 +020011824ssl_c_notbefore : string
11825 Returns the start date presented by the client as a formatted string
11826 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
11827 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +010011828
Willy Tarreau74ca5042013-06-11 23:12:07 +020011829ssl_c_s_dn([<entry>[,<occ>]]) : string
11830 When the incoming connection was made over an SSL/TLS transport layer,
11831 returns the full distinguished name of the subject of the certificate
11832 presented by the client when no <entry> is specified, or the value of the
11833 first given entry found from the beginning of the DN. If a positive/negative
11834 occurrence number is specified as the optional second argument, it returns
11835 the value of the nth given entry value from the beginning/end of the DN.
11836 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
11837 "ssl_c_s_dn(CN)" retrieves the common name.
Willy Tarreaub6672b52011-12-12 17:23:41 +010011838
Willy Tarreau74ca5042013-06-11 23:12:07 +020011839ssl_c_serial : binary
11840 Returns the serial of the certificate presented by the client when the
11841 incoming connection was made over an SSL/TLS transport layer. When used for
11842 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +020011843
Willy Tarreau74ca5042013-06-11 23:12:07 +020011844ssl_c_sha1 : binary
11845 Returns the SHA-1 fingerprint of the certificate presented by the client when
11846 the incoming connection was made over an SSL/TLS transport layer. This can be
11847 used to stick a client to a server, or to pass this information to a server.
Willy Tarreau2d0caa32014-07-02 19:01:22 +020011848 Note that the output is binary, so if you want to pass that signature to the
11849 server, you need to encode it in hex or base64, such as in the example below:
11850
11851 http-request set-header X-SSL-Client-SHA1 %[ssl_c_sha1,hex]
Emeric Brun2525b6b2012-10-18 15:59:43 +020011852
Willy Tarreau74ca5042013-06-11 23:12:07 +020011853ssl_c_sig_alg : string
11854 Returns the name of the algorithm used to sign the certificate presented by
11855 the client when the incoming connection was made over an SSL/TLS transport
11856 layer.
Emeric Brun87855892012-10-17 17:39:35 +020011857
Willy Tarreau74ca5042013-06-11 23:12:07 +020011858ssl_c_used : boolean
11859 Returns true if current SSL session uses a client certificate even if current
11860 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +020011861
Willy Tarreau74ca5042013-06-11 23:12:07 +020011862ssl_c_verify : integer
11863 Returns the verify result error ID when the incoming connection was made over
11864 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
11865 refer to your SSL library's documentation for an exhaustive list of error
11866 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +020011867
Willy Tarreau74ca5042013-06-11 23:12:07 +020011868ssl_c_version : integer
11869 Returns the version of the certificate presented by the client when the
11870 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +020011871
Emeric Brun43e79582014-10-29 19:03:26 +010011872ssl_f_der : binary
11873 Returns the DER formatted certificate presented by the frontend when the
11874 incoming connection was made over an SSL/TLS transport layer. When used for
11875 an ACL, the value(s) to match against can be passed in hexadecimal form.
11876
Willy Tarreau74ca5042013-06-11 23:12:07 +020011877ssl_f_i_dn([<entry>[,<occ>]]) : string
11878 When the incoming connection was made over an SSL/TLS transport layer,
11879 returns the full distinguished name of the issuer of the certificate
11880 presented by the frontend when no <entry> is specified, or the value of the
11881 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +020011882 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +020011883 the value of the nth given entry value from the beginning/end of the DN.
11884 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
11885 "ssl_f_i_dn(CN)" retrieves the common name.
Emeric Brun87855892012-10-17 17:39:35 +020011886
Willy Tarreau74ca5042013-06-11 23:12:07 +020011887ssl_f_key_alg : string
11888 Returns the name of the algorithm used to generate the key of the certificate
11889 presented by the frontend when the incoming connection was made over an
11890 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020011891
Willy Tarreau74ca5042013-06-11 23:12:07 +020011892ssl_f_notafter : string
11893 Returns the end date presented by the frontend as a formatted string
11894 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
11895 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +020011896
Willy Tarreau74ca5042013-06-11 23:12:07 +020011897ssl_f_notbefore : string
11898 Returns the start date presented by the frontend as a formatted string
11899 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
11900 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +020011901
Willy Tarreau74ca5042013-06-11 23:12:07 +020011902ssl_f_s_dn([<entry>[,<occ>]]) : string
11903 When the incoming connection was made over an SSL/TLS transport layer,
11904 returns the full distinguished name of the subject of the certificate
11905 presented by the frontend when no <entry> is specified, or the value of the
11906 first given entry found from the beginning of the DN. If a positive/negative
11907 occurrence number is specified as the optional second argument, it returns
11908 the value of the nth given entry value from the beginning/end of the DN.
11909 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
11910 "ssl_f_s_dn(CN)" retrieves the common name.
Emeric Brunce5ad802012-10-22 14:11:22 +020011911
Willy Tarreau74ca5042013-06-11 23:12:07 +020011912ssl_f_serial : binary
11913 Returns the serial of the certificate presented by the frontend when the
11914 incoming connection was made over an SSL/TLS transport layer. When used for
11915 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +020011916
Emeric Brun55f4fa82014-04-30 17:11:25 +020011917ssl_f_sha1 : binary
11918 Returns the SHA-1 fingerprint of the certificate presented by the frontend
11919 when the incoming connection was made over an SSL/TLS transport layer. This
11920 can be used to know which certificate was chosen using SNI.
11921
Willy Tarreau74ca5042013-06-11 23:12:07 +020011922ssl_f_sig_alg : string
11923 Returns the name of the algorithm used to sign the certificate presented by
11924 the frontend when the incoming connection was made over an SSL/TLS transport
11925 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020011926
Willy Tarreau74ca5042013-06-11 23:12:07 +020011927ssl_f_version : integer
11928 Returns the version of the certificate presented by the frontend when the
11929 incoming connection was made over an SSL/TLS transport layer.
11930
11931ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020011932 Returns true when the front connection was made via an SSL/TLS transport
11933 layer and is locally deciphered. This means it has matched a socket declared
11934 with a "bind" line having the "ssl" option.
11935
Willy Tarreau74ca5042013-06-11 23:12:07 +020011936 Example :
11937 # This passes "X-Proto: https" to servers when client connects over SSL
11938 listen http-https
11939 bind :80
11940 bind :443 ssl crt /etc/haproxy.pem
11941 http-request add-header X-Proto https if { ssl_fc }
11942
11943ssl_fc_alg_keysize : integer
11944 Returns the symmetric cipher key size supported in bits when the incoming
11945 connection was made over an SSL/TLS transport layer.
11946
11947ssl_fc_alpn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011948 This extracts the Application Layer Protocol Negotiation field from an
Willy Tarreau74ca5042013-06-11 23:12:07 +020011949 incoming connection made via a TLS transport layer and locally deciphered by
11950 haproxy. The result is a string containing the protocol name advertised by
11951 the client. The SSL library must have been built with support for TLS
11952 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
11953 not advertised unless the "alpn" keyword on the "bind" line specifies a
11954 protocol list. Also, nothing forces the client to pick a protocol from this
11955 list, any other one may be requested. The TLS ALPN extension is meant to
11956 replace the TLS NPN extension. See also "ssl_fc_npn".
11957
Willy Tarreau74ca5042013-06-11 23:12:07 +020011958ssl_fc_cipher : string
11959 Returns the name of the used cipher when the incoming connection was made
11960 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +020011961
Willy Tarreau74ca5042013-06-11 23:12:07 +020011962ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020011963 Returns true if a client certificate is present in an incoming connection over
11964 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +010011965 Note: on SSL session resumption with Session ID or TLS ticket, client
11966 certificate is not present in the current connection but may be retrieved
11967 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
11968 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +020011969
Willy Tarreau74ca5042013-06-11 23:12:07 +020011970ssl_fc_has_sni : boolean
11971 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020011972 in an incoming connection was made over an SSL/TLS transport layer. Returns
11973 true when the incoming connection presents a TLS SNI field. This requires
11974 that the SSL library is build with support for TLS extensions enabled (check
11975 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +020011976
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020011977ssl_fc_is_resumed: boolean
11978 Returns true if the SSL/TLS session has been resumed through the use of
11979 SSL session cache or TLS tickets.
11980
Willy Tarreau74ca5042013-06-11 23:12:07 +020011981ssl_fc_npn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011982 This extracts the Next Protocol Negotiation field from an incoming connection
Willy Tarreau74ca5042013-06-11 23:12:07 +020011983 made via a TLS transport layer and locally deciphered by haproxy. The result
11984 is a string containing the protocol name advertised by the client. The SSL
11985 library must have been built with support for TLS extensions enabled (check
11986 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
11987 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
11988 forces the client to pick a protocol from this list, any other one may be
11989 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +020011990
Willy Tarreau74ca5042013-06-11 23:12:07 +020011991ssl_fc_protocol : string
11992 Returns the name of the used protocol when the incoming connection was made
11993 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +020011994
Emeric Brunb73a9b02014-04-30 18:49:19 +020011995ssl_fc_unique_id : binary
David Sc1ad52e2014-04-08 18:48:47 -040011996 When the incoming connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020011997 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
11998 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
David Sc1ad52e2014-04-08 18:48:47 -040011999
Willy Tarreau74ca5042013-06-11 23:12:07 +020012000ssl_fc_session_id : binary
12001 Returns the SSL ID of the front connection when the incoming connection was
12002 made over an SSL/TLS transport layer. It is useful to stick a given client to
12003 a server. It is important to note that some browsers refresh their session ID
12004 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +020012005
Willy Tarreau74ca5042013-06-11 23:12:07 +020012006ssl_fc_sni : string
12007 This extracts the Server Name Indication TLS extension (SNI) field from an
12008 incoming connection made via an SSL/TLS transport layer and locally
12009 deciphered by haproxy. The result (when present) typically is a string
12010 matching the HTTPS host name (253 chars or less). The SSL library must have
12011 been built with support for TLS extensions enabled (check haproxy -vv).
12012
12013 This fetch is different from "req_ssl_sni" above in that it applies to the
12014 connection being deciphered by haproxy and not to SSL contents being blindly
12015 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +020012016 requires that the SSL library is build with support for TLS extensions
12017 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +020012018
Willy Tarreau74ca5042013-06-11 23:12:07 +020012019 ACL derivatives :
Willy Tarreau74ca5042013-06-11 23:12:07 +020012020 ssl_fc_sni_end : suffix match
12021 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +020012022
Willy Tarreau74ca5042013-06-11 23:12:07 +020012023ssl_fc_use_keysize : integer
12024 Returns the symmetric cipher key size used in bits when the incoming
12025 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020012026
Willy Tarreaub6fb4202008-07-20 11:18:28 +020012027
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200120287.3.5. Fetching samples from buffer contents (Layer 6)
Willy Tarreau74ca5042013-06-11 23:12:07 +020012029------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +020012030
Willy Tarreau74ca5042013-06-11 23:12:07 +020012031Fetching samples from buffer contents is a bit different from the previous
12032sample fetches above because the sampled data are ephemeral. These data can
12033only be used when they're available and will be lost when they're forwarded.
12034For this reason, samples fetched from buffer contents during a request cannot
12035be used in a response for example. Even while the data are being fetched, they
12036can change. Sometimes it is necessary to set some delays or combine multiple
12037sample fetch methods to ensure that the expected data are complete and usable,
12038for example through TCP request content inspection. Please see the "tcp-request
12039content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +020012040
Willy Tarreau74ca5042013-06-11 23:12:07 +020012041payload(<offset>,<length>) : binary (deprecated)
12042 This is an alias for "req.payload" when used in the context of a request (eg:
12043 "stick on", "stick match"), and for "res.payload" when used in the context of
12044 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010012045
Willy Tarreau74ca5042013-06-11 23:12:07 +020012046payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
12047 This is an alias for "req.payload_lv" when used in the context of a request
12048 (eg: "stick on", "stick match"), and for "res.payload_lv" when used in the
12049 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010012050
Willy Tarreau74ca5042013-06-11 23:12:07 +020012051req.len : integer
12052req_len : integer (deprecated)
12053 Returns an integer value corresponding to the number of bytes present in the
12054 request buffer. This is mostly used in ACL. It is important to understand
12055 that this test does not return false as long as the buffer is changing. This
12056 means that a check with equality to zero will almost always immediately match
12057 at the beginning of the session, while a test for more data will wait for
12058 that data to come in and return false only when haproxy is certain that no
12059 more data will come in. This test was designed to be used with TCP request
12060 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020012061
Willy Tarreau74ca5042013-06-11 23:12:07 +020012062req.payload(<offset>,<length>) : binary
12063 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020012064 in the request buffer. As a special case, if the <length> argument is zero,
12065 the the whole buffer from <offset> to the end is extracted. This can be used
12066 with ACLs in order to check for the presence of some content in a buffer at
12067 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020012068
Willy Tarreau74ca5042013-06-11 23:12:07 +020012069 ACL alternatives :
12070 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020012071
Willy Tarreau74ca5042013-06-11 23:12:07 +020012072req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
12073 This extracts a binary block whose size is specified at <offset1> for <length>
12074 bytes, and which starts at <offset2> if specified or just after the length in
12075 the request buffer. The <offset2> parameter also supports relative offsets if
12076 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020012077
Willy Tarreau74ca5042013-06-11 23:12:07 +020012078 ACL alternatives :
12079 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020012080
Willy Tarreau74ca5042013-06-11 23:12:07 +020012081 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020012082
Willy Tarreau74ca5042013-06-11 23:12:07 +020012083req.proto_http : boolean
12084req_proto_http : boolean (deprecated)
12085 Returns true when data in the request buffer look like HTTP and correctly
12086 parses as such. It is the same parser as the common HTTP request parser which
12087 is used so there should be no surprises. The test does not match until the
12088 request is complete, failed or timed out. This test may be used to report the
12089 protocol in TCP logs, but the biggest use is to block TCP request analysis
12090 until a complete HTTP request is present in the buffer, for example to track
12091 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020012092
Willy Tarreau74ca5042013-06-11 23:12:07 +020012093 Example:
12094 # track request counts per "base" (concatenation of Host+URL)
12095 tcp-request inspect-delay 10s
12096 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012097 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020012098
Willy Tarreau74ca5042013-06-11 23:12:07 +020012099req.rdp_cookie([<name>]) : string
12100rdp_cookie([<name>]) : string (deprecated)
12101 When the request buffer looks like the RDP protocol, extracts the RDP cookie
12102 <name>, or any cookie if unspecified. The parser only checks for the first
12103 cookie, as illustrated in the RDP protocol specification. The cookie name is
12104 case insensitive. Generally the "MSTS" cookie name will be used, as it can
12105 contain the user name of the client connecting to the server if properly
12106 configured on the client. The "MSTSHASH" cookie is often used as well for
12107 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020012108
Willy Tarreau74ca5042013-06-11 23:12:07 +020012109 This differs from "balance rdp-cookie" in that any balancing algorithm may be
12110 used and thus the distribution of clients to backend servers is not linked to
12111 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
12112 such as "balance roundrobin" or "balance leastconn" will lead to a more even
12113 distribution of clients to backend servers than the hash used by "balance
12114 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020012115
Willy Tarreau74ca5042013-06-11 23:12:07 +020012116 ACL derivatives :
12117 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020012118
Willy Tarreau74ca5042013-06-11 23:12:07 +020012119 Example :
12120 listen tse-farm
12121 bind 0.0.0.0:3389
12122 # wait up to 5s for an RDP cookie in the request
12123 tcp-request inspect-delay 5s
12124 tcp-request content accept if RDP_COOKIE
12125 # apply RDP cookie persistence
12126 persist rdp-cookie
12127 # Persist based on the mstshash cookie
12128 # This is only useful makes sense if
12129 # balance rdp-cookie is not used
12130 stick-table type string size 204800
12131 stick on req.rdp_cookie(mstshash)
12132 server srv1 1.1.1.1:3389
12133 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +020012134
Willy Tarreau74ca5042013-06-11 23:12:07 +020012135 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
12136 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020012137
Willy Tarreau74ca5042013-06-11 23:12:07 +020012138req.rdp_cookie_cnt([name]) : integer
12139rdp_cookie_cnt([name]) : integer (deprecated)
12140 Tries to parse the request buffer as RDP protocol, then returns an integer
12141 corresponding to the number of RDP cookies found. If an optional cookie name
12142 is passed, only cookies matching this name are considered. This is mostly
12143 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020012144
Willy Tarreau74ca5042013-06-11 23:12:07 +020012145 ACL derivatives :
12146 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020012147
Willy Tarreau74ca5042013-06-11 23:12:07 +020012148req.ssl_hello_type : integer
12149req_ssl_hello_type : integer (deprecated)
12150 Returns an integer value containing the type of the SSL hello message found
12151 in the request buffer if the buffer contains data that parse as a complete
12152 SSL (v3 or superior) client hello message. Note that this only applies to raw
12153 contents found in the request buffer and not to contents deciphered via an
12154 SSL data layer, so this will not work with "bind" lines having the "ssl"
12155 option. This is mostly used in ACL to detect presence of an SSL hello message
12156 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020012157
Willy Tarreau74ca5042013-06-11 23:12:07 +020012158req.ssl_sni : string
12159req_ssl_sni : string (deprecated)
12160 Returns a string containing the value of the Server Name TLS extension sent
12161 by a client in a TLS stream passing through the request buffer if the buffer
12162 contains data that parse as a complete SSL (v3 or superior) client hello
12163 message. Note that this only applies to raw contents found in the request
12164 buffer and not to contents deciphered via an SSL data layer, so this will not
12165 work with "bind" lines having the "ssl" option. SNI normally contains the
12166 name of the host the client tries to connect to (for recent browsers). SNI is
12167 useful for allowing or denying access to certain hosts when SSL/TLS is used
12168 by the client. This test was designed to be used with TCP request content
12169 inspection. If content switching is needed, it is recommended to first wait
12170 for a complete client hello (type 1), like in the example below. See also
12171 "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020012172
Willy Tarreau74ca5042013-06-11 23:12:07 +020012173 ACL derivatives :
12174 req_ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020012175
Willy Tarreau74ca5042013-06-11 23:12:07 +020012176 Examples :
12177 # Wait for a client hello for at most 5 seconds
12178 tcp-request inspect-delay 5s
12179 tcp-request content accept if { req_ssl_hello_type 1 }
12180 use_backend bk_allow if { req_ssl_sni -f allowed_sites }
12181 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +020012182
Willy Tarreau74ca5042013-06-11 23:12:07 +020012183res.ssl_hello_type : integer
12184rep_ssl_hello_type : integer (deprecated)
12185 Returns an integer value containing the type of the SSL hello message found
12186 in the response buffer if the buffer contains data that parses as a complete
12187 SSL (v3 or superior) hello message. Note that this only applies to raw
12188 contents found in the response buffer and not to contents deciphered via an
12189 SSL data layer, so this will not work with "server" lines having the "ssl"
12190 option. This is mostly used in ACL to detect presence of an SSL hello message
12191 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau51539362012-05-08 12:46:28 +020012192
Willy Tarreau74ca5042013-06-11 23:12:07 +020012193req.ssl_ver : integer
12194req_ssl_ver : integer (deprecated)
12195 Returns an integer value containing the version of the SSL/TLS protocol of a
12196 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
12197 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
12198 composed of the major version multiplied by 65536, added to the minor
12199 version. Note that this only applies to raw contents found in the request
12200 buffer and not to contents deciphered via an SSL data layer, so this will not
12201 work with "bind" lines having the "ssl" option. The ACL version of the test
12202 matches against a decimal notation in the form MAJOR.MINOR (eg: 3.1). This
12203 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012204
Willy Tarreau74ca5042013-06-11 23:12:07 +020012205 ACL derivatives :
12206 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +010012207
Willy Tarreau47e8eba2013-09-11 23:28:46 +020012208res.len : integer
12209 Returns an integer value corresponding to the number of bytes present in the
12210 response buffer. This is mostly used in ACL. It is important to understand
12211 that this test does not return false as long as the buffer is changing. This
12212 means that a check with equality to zero will almost always immediately match
12213 at the beginning of the session, while a test for more data will wait for
12214 that data to come in and return false only when haproxy is certain that no
12215 more data will come in. This test was designed to be used with TCP response
12216 content inspection.
12217
Willy Tarreau74ca5042013-06-11 23:12:07 +020012218res.payload(<offset>,<length>) : binary
12219 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020012220 in the response buffer. As a special case, if the <length> argument is zero,
12221 the the whole buffer from <offset> to the end is extracted. This can be used
12222 with ACLs in order to check for the presence of some content in a buffer at
12223 any location.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012224
Willy Tarreau74ca5042013-06-11 23:12:07 +020012225res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
12226 This extracts a binary block whose size is specified at <offset1> for <length>
12227 bytes, and which starts at <offset2> if specified or just after the length in
12228 the response buffer. The <offset2> parameter also supports relative offsets
12229 if prepended with a '+' or '-' sign.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012230
Willy Tarreau74ca5042013-06-11 23:12:07 +020012231 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012232
Willy Tarreau74ca5042013-06-11 23:12:07 +020012233wait_end : boolean
12234 This fetch either returns true when the inspection period is over, or does
12235 not fetch. It is only used in ACLs, in conjunction with content analysis to
12236 avoid returning a wrong verdict early. It may also be used to delay some
12237 actions, such as a delayed reject for some special addresses. Since it either
12238 stops the rules evaluation or immediately returns true, it is recommended to
12239 use this acl as the last one in a rule. Please note that the default ACL
12240 "WAIT_END" is always usable without prior declaration. This test was designed
12241 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012242
Willy Tarreau74ca5042013-06-11 23:12:07 +020012243 Examples :
12244 # delay every incoming request by 2 seconds
12245 tcp-request inspect-delay 2s
12246 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +010012247
Willy Tarreau74ca5042013-06-11 23:12:07 +020012248 # don't immediately tell bad guys they are rejected
12249 tcp-request inspect-delay 10s
12250 acl goodguys src 10.0.0.0/24
12251 acl badguys src 10.0.1.0/24
12252 tcp-request content accept if goodguys
12253 tcp-request content reject if badguys WAIT_END
12254 tcp-request content reject
12255
12256
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200122577.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +020012258--------------------------------------
12259
12260It is possible to fetch samples from HTTP contents, requests and responses.
12261This application layer is also called layer 7. It is only possible to fetch the
12262data in this section when a full HTTP request or response has been parsed from
12263its respective request or response buffer. This is always the case with all
12264HTTP specific rules and for sections running with "mode http". When using TCP
12265content inspection, it may be necessary to support an inspection delay in order
12266to let the request or response come in first. These fetches may require a bit
12267more CPU resources than the layer 4 ones, but not much since the request and
12268response are indexed.
12269
12270base : string
12271 This returns the concatenation of the first Host header and the path part of
12272 the request, which starts at the first slash and ends before the question
12273 mark. It can be useful in virtual hosted environments to detect URL abuses as
12274 well as to improve shared caches efficiency. Using this with a limited size
12275 stick table also allows one to collect statistics about most commonly
12276 requested objects by host/path. With ACLs it can allow simple content
12277 switching rules involving the host and the path at the same time, such as
12278 "www.example.com/favicon.ico". See also "path" and "uri".
12279
12280 ACL derivatives :
12281 base : exact string match
12282 base_beg : prefix match
12283 base_dir : subdir match
12284 base_dom : domain match
12285 base_end : suffix match
12286 base_len : length match
12287 base_reg : regex match
12288 base_sub : substring match
12289
12290base32 : integer
12291 This returns a 32-bit hash of the value returned by the "base" fetch method
12292 above. This is useful to track per-URL activity on high traffic sites without
12293 having to store all URLs. Instead a shorter hash is stored, saving a lot of
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012294 memory. The output type is an unsigned integer. The hash function used is
12295 SDBM with full avalanche on the output. Technically, base32 is exactly equal
12296 to "base,sdbm(1)".
Willy Tarreau74ca5042013-06-11 23:12:07 +020012297
12298base32+src : binary
12299 This returns the concatenation of the base32 fetch above and the src fetch
12300 below. The resulting type is of type binary, with a size of 8 or 20 bytes
12301 depending on the source address family. This can be used to track per-IP,
12302 per-URL counters.
12303
William Lallemand65ad6e12014-01-31 15:08:02 +010012304capture.req.hdr(<idx>) : string
12305 This extracts the content of the header captured by the "capture request
12306 header", idx is the position of the capture keyword in the configuration.
12307 The first entry is an index of 0. See also: "capture request header".
12308
12309capture.req.method : string
12310 This extracts the METHOD of an HTTP request. It can be used in both request
12311 and response. Unlike "method", it can be used in both request and response
12312 because it's allocated.
12313
12314capture.req.uri : string
12315 This extracts the request's URI, which starts at the first slash and ends
12316 before the first space in the request (without the host part). Unlike "path"
12317 and "url", it can be used in both request and response because it's
12318 allocated.
12319
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020012320capture.req.ver : string
12321 This extracts the request's HTTP version and returns either "HTTP/1.0" or
12322 "HTTP/1.1". Unlike "req.ver", it can be used in both request, response, and
12323 logs because it relies on a persistent flag.
12324
William Lallemand65ad6e12014-01-31 15:08:02 +010012325capture.res.hdr(<idx>) : string
12326 This extracts the content of the header captured by the "capture response
12327 header", idx is the position of the capture keyword in the configuration.
12328 The first entry is an index of 0.
12329 See also: "capture response header"
12330
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020012331capture.res.ver : string
12332 This extracts the response's HTTP version and returns either "HTTP/1.0" or
12333 "HTTP/1.1". Unlike "res.ver", it can be used in logs because it relies on a
12334 persistent flag.
12335
Willy Tarreaua5910cc2015-05-02 00:46:08 +020012336req.body : binary
12337 This returns the HTTP request's available body as a block of data. It
12338 requires that the request body has been buffered made available using
12339 "option http-buffer-request". In case of chunked-encoded body, currently only
12340 the first chunk is analyzed.
12341
Thierry FOURNIER9826c772015-05-20 15:50:54 +020012342req.body_param([<name>) : string
12343 This fetch assumes that the body of the POST request is url-encoded. The user
12344 can check if the "content-type" contains the value
12345 "application/x-www-form-urlencoded". This extracts the first occurrence of the
12346 parameter <name> in the body, which ends before '&'. The parameter name is
12347 case-sensitive. If no name is given, any parameter will match, and the first
12348 one will be returned. The result is a string corresponding to the value of the
12349 parameter <name> as presented in the request body (no URL decoding is
12350 performed). Note that the ACL version of this fetch iterates over multiple
12351 parameters and will iteratively report all parameters values if no name is
12352 given.
12353
Willy Tarreaua5910cc2015-05-02 00:46:08 +020012354req.body_len : integer
12355 This returns the length of the HTTP request's available body in bytes. It may
12356 be lower than the advertised length if the body is larger than the buffer. It
12357 requires that the request body has been buffered made available using
12358 "option http-buffer-request".
12359
12360req.body_size : integer
12361 This returns the advertised length of the HTTP request's body in bytes. It
12362 will represent the advertised Content-Length header, or the size of the first
12363 chunk in case of chunked encoding. In order to parse the chunks, it requires
12364 that the request body has been buffered made available using
12365 "option http-buffer-request".
12366
Willy Tarreau74ca5042013-06-11 23:12:07 +020012367req.cook([<name>]) : string
12368cook([<name>]) : string (deprecated)
12369 This extracts the last occurrence of the cookie name <name> on a "Cookie"
12370 header line from the request, and returns its value as string. If no name is
12371 specified, the first cookie value is returned. When used with ACLs, all
12372 matching cookies are evaluated. Spaces around the name and the value are
12373 ignored as requested by the Cookie header specification (RFC6265). The cookie
12374 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
12375 well return an empty value if it is present. Use the "found" match to detect
12376 presence. Use the res.cook() variant for response cookies sent by the server.
12377
12378 ACL derivatives :
12379 cook([<name>]) : exact string match
12380 cook_beg([<name>]) : prefix match
12381 cook_dir([<name>]) : subdir match
12382 cook_dom([<name>]) : domain match
12383 cook_end([<name>]) : suffix match
12384 cook_len([<name>]) : length match
12385 cook_reg([<name>]) : regex match
12386 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +010012387
Willy Tarreau74ca5042013-06-11 23:12:07 +020012388req.cook_cnt([<name>]) : integer
12389cook_cnt([<name>]) : integer (deprecated)
12390 Returns an integer value representing the number of occurrences of the cookie
12391 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012392
Willy Tarreau74ca5042013-06-11 23:12:07 +020012393req.cook_val([<name>]) : integer
12394cook_val([<name>]) : integer (deprecated)
12395 This extracts the last occurrence of the cookie name <name> on a "Cookie"
12396 header line from the request, and converts its value to an integer which is
12397 returned. If no name is specified, the first cookie value is returned. When
12398 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +020012399
Willy Tarreau74ca5042013-06-11 23:12:07 +020012400cookie([<name>]) : string (deprecated)
12401 This extracts the last occurrence of the cookie name <name> on a "Cookie"
12402 header line from the request, or a "Set-Cookie" header from the response, and
12403 returns its value as a string. A typical use is to get multiple clients
12404 sharing a same profile use the same server. This can be similar to what
12405 "appsession" does with the "request-learn" statement, but with support for
12406 multi-peer synchronization and state keeping across restarts. If no name is
12407 specified, the first cookie value is returned. This fetch should not be used
12408 anymore and should be replaced by req.cook() or res.cook() instead as it
12409 ambiguously uses the direction based on the context where it is used.
12410 See also : "appsession".
Willy Tarreaud63335a2010-02-26 12:56:52 +010012411
Willy Tarreau74ca5042013-06-11 23:12:07 +020012412hdr([<name>[,<occ>]]) : string
12413 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
12414 used on responses. Please refer to these respective fetches for more details.
12415 In case of doubt about the fetch direction, please use the explicit ones.
12416 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012417 unambiguously apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012418
Willy Tarreau74ca5042013-06-11 23:12:07 +020012419req.fhdr(<name>[,<occ>]) : string
12420 This extracts the last occurrence of header <name> in an HTTP request. When
12421 used from an ACL, all occurrences are iterated over until a match is found.
12422 Optionally, a specific occurrence might be specified as a position number.
12423 Positive values indicate a position from the first occurrence, with 1 being
12424 the first one. Negative values indicate positions relative to the last one,
12425 with -1 being the last one. It differs from req.hdr() in that any commas
12426 present in the value are returned and are not used as delimiters. This is
12427 sometimes useful with headers such as User-Agent.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012428
Willy Tarreau74ca5042013-06-11 23:12:07 +020012429req.fhdr_cnt([<name>]) : integer
12430 Returns an integer value representing the number of occurrences of request
12431 header field name <name>, or the total number of header fields if <name> is
12432 not specified. Contrary to its req.hdr_cnt() cousin, this function returns
12433 the number of full line headers and does not stop on commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012434
Willy Tarreau74ca5042013-06-11 23:12:07 +020012435req.hdr([<name>[,<occ>]]) : string
12436 This extracts the last occurrence of header <name> in an HTTP request. When
12437 used from an ACL, all occurrences are iterated over until a match is found.
12438 Optionally, a specific occurrence might be specified as a position number.
12439 Positive values indicate a position from the first occurrence, with 1 being
12440 the first one. Negative values indicate positions relative to the last one,
12441 with -1 being the last one. A typical use is with the X-Forwarded-For header
12442 once converted to IP, associated with an IP stick-table. The function
12443 considers any comma as a delimiter for distinct values. If full-line headers
12444 are desired instead, use req.fhdr(). Please carefully check RFC2616 to know
12445 how certain headers are supposed to be parsed. Also, some of them are case
12446 insensitive (eg: Connection).
Willy Tarreaud63335a2010-02-26 12:56:52 +010012447
Willy Tarreau74ca5042013-06-11 23:12:07 +020012448 ACL derivatives :
12449 hdr([<name>[,<occ>]]) : exact string match
12450 hdr_beg([<name>[,<occ>]]) : prefix match
12451 hdr_dir([<name>[,<occ>]]) : subdir match
12452 hdr_dom([<name>[,<occ>]]) : domain match
12453 hdr_end([<name>[,<occ>]]) : suffix match
12454 hdr_len([<name>[,<occ>]]) : length match
12455 hdr_reg([<name>[,<occ>]]) : regex match
12456 hdr_sub([<name>[,<occ>]]) : substring match
12457
12458req.hdr_cnt([<name>]) : integer
12459hdr_cnt([<header>]) : integer (deprecated)
12460 Returns an integer value representing the number of occurrences of request
12461 header field name <name>, or the total number of header field values if
12462 <name> is not specified. It is important to remember that one header line may
12463 count as several headers if it has several values. The function considers any
12464 comma as a delimiter for distinct values. If full-line headers are desired
12465 instead, req.fhdr_cnt() should be used instead. With ACLs, it can be used to
12466 detect presence, absence or abuse of a specific header, as well as to block
12467 request smuggling attacks by rejecting requests which contain more than one
12468 of certain headers. See "req.hdr" for more information on header matching.
12469
12470req.hdr_ip([<name>[,<occ>]]) : ip
12471hdr_ip([<name>[,<occ>]]) : ip (deprecated)
12472 This extracts the last occurrence of header <name> in an HTTP request,
12473 converts it to an IPv4 or IPv6 address and returns this address. When used
12474 with ACLs, all occurrences are checked, and if <name> is omitted, every value
12475 of every header is checked. Optionally, a specific occurrence might be
12476 specified as a position number. Positive values indicate a position from the
12477 first occurrence, with 1 being the first one. Negative values indicate
12478 positions relative to the last one, with -1 being the last one. A typical use
12479 is with the X-Forwarded-For and X-Client-IP headers.
12480
12481req.hdr_val([<name>[,<occ>]]) : integer
12482hdr_val([<name>[,<occ>]]) : integer (deprecated)
12483 This extracts the last occurrence of header <name> in an HTTP request, and
12484 converts it to an integer value. When used with ACLs, all occurrences are
12485 checked, and if <name> is omitted, every value of every header is checked.
12486 Optionally, a specific occurrence might be specified as a position number.
12487 Positive values indicate a position from the first occurrence, with 1 being
12488 the first one. Negative values indicate positions relative to the last one,
12489 with -1 being the last one. A typical use is with the X-Forwarded-For header.
12490
12491http_auth(<userlist>) : boolean
12492 Returns a boolean indicating whether the authentication data received from
12493 the client match a username & password stored in the specified userlist. This
12494 fetch function is not really useful outside of ACLs. Currently only http
12495 basic auth is supported.
12496
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010012497http_auth_group(<userlist>) : string
12498 Returns a string corresponding to the user name found in the authentication
12499 data received from the client if both the user name and password are valid
12500 according to the specified userlist. The main purpose is to use it in ACLs
12501 where it is then checked whether the user belongs to any group within a list.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012502 This fetch function is not really useful outside of ACLs. Currently only http
12503 basic auth is supported.
12504
12505 ACL derivatives :
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010012506 http_auth_group(<userlist>) : group ...
12507 Returns true when the user extracted from the request and whose password is
12508 valid according to the specified userlist belongs to at least one of the
12509 groups.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012510
12511http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020012512 Returns true when the request being processed is the first one of the
12513 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020012514 from some requests when a request is not the first one, or to help grouping
12515 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020012516
Willy Tarreau74ca5042013-06-11 23:12:07 +020012517method : integer + string
12518 Returns an integer value corresponding to the method in the HTTP request. For
12519 example, "GET" equals 1 (check sources to establish the matching). Value 9
12520 means "other method" and may be converted to a string extracted from the
12521 stream. This should not be used directly as a sample, this is only meant to
12522 be used from ACLs, which transparently convert methods from patterns to these
12523 integer + string values. Some predefined ACL already check for most common
12524 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012525
Willy Tarreau74ca5042013-06-11 23:12:07 +020012526 ACL derivatives :
12527 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020012528
Willy Tarreau74ca5042013-06-11 23:12:07 +020012529 Example :
12530 # only accept GET and HEAD requests
12531 acl valid_method method GET HEAD
12532 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020012533
Willy Tarreau74ca5042013-06-11 23:12:07 +020012534path : string
12535 This extracts the request's URL path, which starts at the first slash and
12536 ends before the question mark (without the host part). A typical use is with
12537 prefetch-capable caches, and with portals which need to aggregate multiple
12538 information from databases and keep them in caches. Note that with outgoing
12539 caches, it would be wiser to use "url" instead. With ACLs, it's typically
12540 used to match exact file names (eg: "/login.php"), or directory parts using
12541 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012542
Willy Tarreau74ca5042013-06-11 23:12:07 +020012543 ACL derivatives :
12544 path : exact string match
12545 path_beg : prefix match
12546 path_dir : subdir match
12547 path_dom : domain match
12548 path_end : suffix match
12549 path_len : length match
12550 path_reg : regex match
12551 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020012552
Willy Tarreau49ad95c2015-01-19 15:06:26 +010012553query : string
12554 This extracts the request's query string, which starts after the first
12555 question mark. If no question mark is present, this fetch returns nothing. If
12556 a question mark is present but nothing follows, it returns an empty string.
12557 This means it's possible to easily know whether a query string is present
12558 using the "found" matching method. This fetch is the completemnt of "path"
12559 which stops before the question mark.
12560
Willy Tarreaueb27ec72015-02-20 13:55:29 +010012561req.hdr_names([<delim>]) : string
12562 This builds a string made from the concatenation of all header names as they
12563 appear in the request when the rule is evaluated. The default delimiter is
12564 the comma (',') but it may be overridden as an optional argument <delim>. In
12565 this case, only the first character of <delim> is considered.
12566
Willy Tarreau74ca5042013-06-11 23:12:07 +020012567req.ver : string
12568req_ver : string (deprecated)
12569 Returns the version string from the HTTP request, for example "1.1". This can
12570 be useful for logs, but is mostly there for ACL. Some predefined ACL already
12571 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012572
Willy Tarreau74ca5042013-06-11 23:12:07 +020012573 ACL derivatives :
12574 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020012575
Willy Tarreau74ca5042013-06-11 23:12:07 +020012576res.comp : boolean
12577 Returns the boolean "true" value if the response has been compressed by
12578 HAProxy, otherwise returns boolean "false". This may be used to add
12579 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012580
Willy Tarreau74ca5042013-06-11 23:12:07 +020012581res.comp_algo : string
12582 Returns a string containing the name of the algorithm used if the response
12583 was compressed by HAProxy, for example : "deflate". This may be used to add
12584 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012585
Willy Tarreau74ca5042013-06-11 23:12:07 +020012586res.cook([<name>]) : string
12587scook([<name>]) : string (deprecated)
12588 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
12589 header line from the response, and returns its value as string. If no name is
12590 specified, the first cookie value is returned.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020012591
Willy Tarreau74ca5042013-06-11 23:12:07 +020012592 ACL derivatives :
12593 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020012594
Willy Tarreau74ca5042013-06-11 23:12:07 +020012595res.cook_cnt([<name>]) : integer
12596scook_cnt([<name>]) : integer (deprecated)
12597 Returns an integer value representing the number of occurrences of the cookie
12598 <name> in the response, or all cookies if <name> is not specified. This is
12599 mostly useful when combined with ACLs to detect suspicious responses.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012600
Willy Tarreau74ca5042013-06-11 23:12:07 +020012601res.cook_val([<name>]) : integer
12602scook_val([<name>]) : integer (deprecated)
12603 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
12604 header line from the response, and converts its value to an integer which is
12605 returned. If no name is specified, the first cookie value is returned.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012606
Willy Tarreau74ca5042013-06-11 23:12:07 +020012607res.fhdr([<name>[,<occ>]]) : string
12608 This extracts the last occurrence of header <name> in an HTTP response, or of
12609 the last header if no <name> is specified. Optionally, a specific occurrence
12610 might be specified as a position number. Positive values indicate a position
12611 from the first occurrence, with 1 being the first one. Negative values
12612 indicate positions relative to the last one, with -1 being the last one. It
12613 differs from res.hdr() in that any commas present in the value are returned
12614 and are not used as delimiters. If this is not desired, the res.hdr() fetch
12615 should be used instead. This is sometimes useful with headers such as Date or
12616 Expires.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012617
Willy Tarreau74ca5042013-06-11 23:12:07 +020012618res.fhdr_cnt([<name>]) : integer
12619 Returns an integer value representing the number of occurrences of response
12620 header field name <name>, or the total number of header fields if <name> is
12621 not specified. Contrary to its res.hdr_cnt() cousin, this function returns
12622 the number of full line headers and does not stop on commas. If this is not
12623 desired, the res.hdr_cnt() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012624
Willy Tarreau74ca5042013-06-11 23:12:07 +020012625res.hdr([<name>[,<occ>]]) : string
12626shdr([<name>[,<occ>]]) : string (deprecated)
12627 This extracts the last occurrence of header <name> in an HTTP response, or of
12628 the last header if no <name> is specified. Optionally, a specific occurrence
12629 might be specified as a position number. Positive values indicate a position
12630 from the first occurrence, with 1 being the first one. Negative values
12631 indicate positions relative to the last one, with -1 being the last one. This
12632 can be useful to learn some data into a stick-table. The function considers
12633 any comma as a delimiter for distinct values. If this is not desired, the
12634 res.fhdr() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012635
Willy Tarreau74ca5042013-06-11 23:12:07 +020012636 ACL derivatives :
12637 shdr([<name>[,<occ>]]) : exact string match
12638 shdr_beg([<name>[,<occ>]]) : prefix match
12639 shdr_dir([<name>[,<occ>]]) : subdir match
12640 shdr_dom([<name>[,<occ>]]) : domain match
12641 shdr_end([<name>[,<occ>]]) : suffix match
12642 shdr_len([<name>[,<occ>]]) : length match
12643 shdr_reg([<name>[,<occ>]]) : regex match
12644 shdr_sub([<name>[,<occ>]]) : substring match
12645
12646res.hdr_cnt([<name>]) : integer
12647shdr_cnt([<name>]) : integer (deprecated)
12648 Returns an integer value representing the number of occurrences of response
12649 header field name <name>, or the total number of header fields if <name> is
12650 not specified. The function considers any comma as a delimiter for distinct
12651 values. If this is not desired, the res.fhdr_cnt() fetch should be used
12652 instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012653
Willy Tarreau74ca5042013-06-11 23:12:07 +020012654res.hdr_ip([<name>[,<occ>]]) : ip
12655shdr_ip([<name>[,<occ>]]) : ip (deprecated)
12656 This extracts the last occurrence of header <name> in an HTTP response,
12657 convert it to an IPv4 or IPv6 address and returns this address. Optionally, a
12658 specific occurrence might be specified as a position number. Positive values
12659 indicate a position from the first occurrence, with 1 being the first one.
12660 Negative values indicate positions relative to the last one, with -1 being
12661 the last one. This can be useful to learn some data into a stick table.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012662
Willy Tarreaueb27ec72015-02-20 13:55:29 +010012663res.hdr_names([<delim>]) : string
12664 This builds a string made from the concatenation of all header names as they
12665 appear in the response when the rule is evaluated. The default delimiter is
12666 the comma (',') but it may be overridden as an optional argument <delim>. In
12667 this case, only the first character of <delim> is considered.
12668
Willy Tarreau74ca5042013-06-11 23:12:07 +020012669res.hdr_val([<name>[,<occ>]]) : integer
12670shdr_val([<name>[,<occ>]]) : integer (deprecated)
12671 This extracts the last occurrence of header <name> in an HTTP response, and
12672 converts it to an integer value. Optionally, a specific occurrence might be
12673 specified as a position number. Positive values indicate a position from the
12674 first occurrence, with 1 being the first one. Negative values indicate
12675 positions relative to the last one, with -1 being the last one. This can be
12676 useful to learn some data into a stick table.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010012677
Willy Tarreau74ca5042013-06-11 23:12:07 +020012678res.ver : string
12679resp_ver : string (deprecated)
12680 Returns the version string from the HTTP response, for example "1.1". This
12681 can be useful for logs, but is mostly there for ACL.
Willy Tarreau0e698542011-09-16 08:32:32 +020012682
Willy Tarreau74ca5042013-06-11 23:12:07 +020012683 ACL derivatives :
12684 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010012685
Willy Tarreau74ca5042013-06-11 23:12:07 +020012686set-cookie([<name>]) : string (deprecated)
12687 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
12688 header line from the response and uses the corresponding value to match. This
12689 can be comparable to what "appsession" does with default options, but with
12690 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010012691
Willy Tarreau74ca5042013-06-11 23:12:07 +020012692 This fetch function is deprecated and has been superseded by the "res.cook"
12693 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010012694
Willy Tarreau74ca5042013-06-11 23:12:07 +020012695 See also : "appsession"
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012696
Willy Tarreau74ca5042013-06-11 23:12:07 +020012697status : integer
12698 Returns an integer containing the HTTP status code in the HTTP response, for
12699 example, 302. It is mostly used within ACLs and integer ranges, for example,
12700 to remove any Location header if the response is not a 3xx.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012701
Willy Tarreau74ca5042013-06-11 23:12:07 +020012702url : string
12703 This extracts the request's URL as presented in the request. A typical use is
12704 with prefetch-capable caches, and with portals which need to aggregate
12705 multiple information from databases and keep them in caches. With ACLs, using
12706 "path" is preferred over using "url", because clients may send a full URL as
12707 is normally done with proxies. The only real use is to match "*" which does
12708 not match in "path", and for which there is already a predefined ACL. See
12709 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012710
Willy Tarreau74ca5042013-06-11 23:12:07 +020012711 ACL derivatives :
12712 url : exact string match
12713 url_beg : prefix match
12714 url_dir : subdir match
12715 url_dom : domain match
12716 url_end : suffix match
12717 url_len : length match
12718 url_reg : regex match
12719 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012720
Willy Tarreau74ca5042013-06-11 23:12:07 +020012721url_ip : ip
12722 This extracts the IP address from the request's URL when the host part is
12723 presented as an IP address. Its use is very limited. For instance, a
12724 monitoring system might use this field as an alternative for the source IP in
12725 order to test what path a given source address would follow, or to force an
12726 entry in a table for a given source address. With ACLs it can be used to
12727 restrict access to certain systems through a proxy, for example when combined
12728 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012729
Willy Tarreau74ca5042013-06-11 23:12:07 +020012730url_port : integer
12731 This extracts the port part from the request's URL. Note that if the port is
12732 not specified in the request, port 80 is assumed. With ACLs it can be used to
12733 restrict access to certain systems through a proxy, for example when combined
12734 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012735
Willy Tarreau1ede1da2015-05-07 16:06:18 +020012736urlp([<name>[,<delim>]]) : string
12737url_param([<name>[,<delim>]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020012738 This extracts the first occurrence of the parameter <name> in the query
12739 string, which begins after either '?' or <delim>, and which ends before '&',
Willy Tarreau1ede1da2015-05-07 16:06:18 +020012740 ';' or <delim>. The parameter name is case-sensitive. If no name is given,
12741 any parameter will match, and the first one will be returned. The result is
12742 a string corresponding to the value of the parameter <name> as presented in
12743 the request (no URL decoding is performed). This can be used for session
Willy Tarreau74ca5042013-06-11 23:12:07 +020012744 stickiness based on a client ID, to extract an application cookie passed as a
12745 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
Willy Tarreau1ede1da2015-05-07 16:06:18 +020012746 this fetch iterates over multiple parameters and will iteratively report all
12747 parameters values if no name is given
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012748
Willy Tarreau74ca5042013-06-11 23:12:07 +020012749 ACL derivatives :
12750 urlp(<name>[,<delim>]) : exact string match
12751 urlp_beg(<name>[,<delim>]) : prefix match
12752 urlp_dir(<name>[,<delim>]) : subdir match
12753 urlp_dom(<name>[,<delim>]) : domain match
12754 urlp_end(<name>[,<delim>]) : suffix match
12755 urlp_len(<name>[,<delim>]) : length match
12756 urlp_reg(<name>[,<delim>]) : regex match
12757 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012758
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012759
Willy Tarreau74ca5042013-06-11 23:12:07 +020012760 Example :
12761 # match http://example.com/foo?PHPSESSIONID=some_id
12762 stick on urlp(PHPSESSIONID)
12763 # match http://example.com/foo;JSESSIONID=some_id
12764 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012765
Willy Tarreau1ede1da2015-05-07 16:06:18 +020012766urlp_val([<name>[,<delim>])] : integer
Willy Tarreau74ca5042013-06-11 23:12:07 +020012767 See "urlp" above. This one extracts the URL parameter <name> in the request
12768 and converts it to an integer value. This can be used for session stickiness
12769 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020012770
Willy Tarreau198a7442008-01-17 12:05:32 +010012771
Willy Tarreau74ca5042013-06-11 23:12:07 +0200127727.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012773---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010012774
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012775Some predefined ACLs are hard-coded so that they do not have to be declared in
12776every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020012777order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010012778
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012779ACL name Equivalent to Usage
12780---------------+-----------------------------+---------------------------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012781FALSE always_false never match
Willy Tarreau2492d5b2009-07-11 00:06:00 +020012782HTTP req_proto_http match if protocol is valid HTTP
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012783HTTP_1.0 req_ver 1.0 match HTTP version 1.0
12784HTTP_1.1 req_ver 1.1 match HTTP version 1.1
Willy Tarreaud63335a2010-02-26 12:56:52 +010012785HTTP_CONTENT hdr_val(content-length) gt 0 match an existing content-length
12786HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
12787HTTP_URL_SLASH url_beg / match URL beginning with "/"
12788HTTP_URL_STAR url * match URL equal to "*"
12789LOCALHOST src 127.0.0.1/8 match connection from local host
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012790METH_CONNECT method CONNECT match HTTP CONNECT method
12791METH_GET method GET HEAD match HTTP GET or HEAD method
12792METH_HEAD method HEAD match HTTP HEAD method
12793METH_OPTIONS method OPTIONS match HTTP OPTIONS method
12794METH_POST method POST match HTTP POST method
12795METH_TRACE method TRACE match HTTP TRACE method
Emeric Brunbede3d02009-06-30 17:54:00 +020012796RDP_COOKIE req_rdp_cookie_cnt gt 0 match presence of an RDP cookie
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012797REQ_CONTENT req_len gt 0 match data in the request buffer
Willy Tarreaud63335a2010-02-26 12:56:52 +010012798TRUE always_true always match
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012799WAIT_END wait_end wait for end of content analysis
12800---------------+-----------------------------+---------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010012801
Willy Tarreaub937b7e2010-01-12 15:27:54 +010012802
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200128038. Logging
12804----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010012805
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012806One of HAProxy's strong points certainly lies is its precise logs. It probably
12807provides the finest level of information available for such a product, which is
12808very important for troubleshooting complex environments. Standard information
12809provided in logs include client ports, TCP/HTTP state timers, precise session
12810state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012811to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012812headers.
12813
12814In order to improve administrators reactivity, it offers a great transparency
12815about encountered problems, both internal and external, and it is possible to
12816send logs to different sources at the same time with different level filters :
12817
12818 - global process-level logs (system errors, start/stop, etc..)
12819 - per-instance system and internal errors (lack of resource, bugs, ...)
12820 - per-instance external troubles (servers up/down, max connections)
12821 - per-instance activity (client connections), either at the establishment or
12822 at the termination.
12823
12824The ability to distribute different levels of logs to different log servers
12825allow several production teams to interact and to fix their problems as soon
12826as possible. For example, the system team might monitor system-wide errors,
12827while the application team might be monitoring the up/down for their servers in
12828real time, and the security team might analyze the activity logs with one hour
12829delay.
12830
12831
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200128328.1. Log levels
12833---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012834
Simon Hormandf791f52011-05-29 15:01:10 +090012835TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012836source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090012837HTTP request, HTTP return code, number of bytes transmitted, conditions
12838in which the session ended, and even exchanged cookies values. For example
12839track a particular user's problems. All messages may be sent to up to two
12840syslog servers. Check the "log" keyword in section 4.2 for more information
12841about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012842
12843
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200128448.2. Log formats
12845----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012846
William Lallemand48940402012-01-30 16:47:22 +010012847HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090012848and will be detailed in the following sections. A few of them may vary
12849slightly with the configuration, due to indicators specific to certain
12850options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012851
12852 - the default format, which is very basic and very rarely used. It only
12853 provides very basic information about the incoming connection at the moment
12854 it is accepted : source IP:port, destination IP:port, and frontend-name.
12855 This mode will eventually disappear so it will not be described to great
12856 extents.
12857
12858 - the TCP format, which is more advanced. This format is enabled when "option
12859 tcplog" is set on the frontend. HAProxy will then usually wait for the
12860 connection to terminate before logging. This format provides much richer
12861 information, such as timers, connection counts, queue size, etc... This
12862 format is recommended for pure TCP proxies.
12863
12864 - the HTTP format, which is the most advanced for HTTP proxying. This format
12865 is enabled when "option httplog" is set on the frontend. It provides the
12866 same information as the TCP format with some HTTP-specific fields such as
12867 the request, the status code, and captures of headers and cookies. This
12868 format is recommended for HTTP proxies.
12869
Emeric Brun3a058f32009-06-30 18:26:00 +020012870 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
12871 fields arranged in the same order as the CLF format. In this mode, all
12872 timers, captures, flags, etc... appear one per field after the end of the
12873 common fields, in the same order they appear in the standard HTTP format.
12874
William Lallemand48940402012-01-30 16:47:22 +010012875 - the custom log format, allows you to make your own log line.
12876
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012877Next sections will go deeper into details for each of these formats. Format
12878specification will be performed on a "field" basis. Unless stated otherwise, a
12879field is a portion of text delimited by any number of spaces. Since syslog
12880servers are susceptible of inserting fields at the beginning of a line, it is
12881always assumed that the first field is the one containing the process name and
12882identifier.
12883
12884Note : Since log lines may be quite long, the log examples in sections below
12885 might be broken into multiple lines. The example log lines will be
12886 prefixed with 3 closing angle brackets ('>>>') and each time a log is
12887 broken into multiple lines, each non-final line will end with a
12888 backslash ('\') and the next line will start indented by two characters.
12889
12890
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200128918.2.1. Default log format
12892-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012893
12894This format is used when no specific option is set. The log is emitted as soon
12895as the connection is accepted. One should note that this currently is the only
12896format which logs the request's destination IP and ports.
12897
12898 Example :
12899 listen www
12900 mode http
12901 log global
12902 server srv1 127.0.0.1:8000
12903
12904 >>> Feb 6 12:12:09 localhost \
12905 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
12906 (www/HTTP)
12907
12908 Field Format Extract from the example above
12909 1 process_name '[' pid ']:' haproxy[14385]:
12910 2 'Connect from' Connect from
12911 3 source_ip ':' source_port 10.0.1.2:33312
12912 4 'to' to
12913 5 destination_ip ':' destination_port 10.0.3.31:8012
12914 6 '(' frontend_name '/' mode ')' (www/HTTP)
12915
12916Detailed fields description :
12917 - "source_ip" is the IP address of the client which initiated the connection.
12918 - "source_port" is the TCP port of the client which initiated the connection.
12919 - "destination_ip" is the IP address the client connected to.
12920 - "destination_port" is the TCP port the client connected to.
12921 - "frontend_name" is the name of the frontend (or listener) which received
12922 and processed the connection.
12923 - "mode is the mode the frontend is operating (TCP or HTTP).
12924
Willy Tarreauceb24bc2010-11-09 12:46:41 +010012925In case of a UNIX socket, the source and destination addresses are marked as
12926"unix:" and the ports reflect the internal ID of the socket which accepted the
12927connection (the same ID as reported in the stats).
12928
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012929It is advised not to use this deprecated format for newer installations as it
12930will eventually disappear.
12931
12932
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200129338.2.2. TCP log format
12934---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012935
12936The TCP format is used when "option tcplog" is specified in the frontend, and
12937is the recommended format for pure TCP proxies. It provides a lot of precious
12938information for troubleshooting. Since this format includes timers and byte
12939counts, the log is normally emitted at the end of the session. It can be
12940emitted earlier if "option logasap" is specified, which makes sense in most
12941environments with long sessions such as remote terminals. Sessions which match
12942the "monitor" rules are never logged. It is also possible not to emit logs for
12943sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020012944specifying "option dontlognull" in the frontend. Successful connections will
12945not be logged if "option dontlog-normal" is specified in the frontend. A few
12946fields may slightly vary depending on some configuration options, those are
12947marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012948
12949 Example :
12950 frontend fnt
12951 mode tcp
12952 option tcplog
12953 log global
12954 default_backend bck
12955
12956 backend bck
12957 server srv1 127.0.0.1:8000
12958
12959 >>> Feb 6 12:12:56 localhost \
12960 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
12961 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
12962
12963 Field Format Extract from the example above
12964 1 process_name '[' pid ']:' haproxy[14387]:
12965 2 client_ip ':' client_port 10.0.1.2:33313
12966 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
12967 4 frontend_name fnt
12968 5 backend_name '/' server_name bck/srv1
12969 6 Tw '/' Tc '/' Tt* 0/0/5007
12970 7 bytes_read* 212
12971 8 termination_state --
12972 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
12973 10 srv_queue '/' backend_queue 0/0
12974
12975Detailed fields description :
12976 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010012977 connection to haproxy. If the connection was accepted on a UNIX socket
12978 instead, the IP address would be replaced with the word "unix". Note that
12979 when the connection is accepted on a socket configured with "accept-proxy"
12980 and the PROXY protocol is correctly used, then the logs will reflect the
12981 forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012982
12983 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010012984 If the connection was accepted on a UNIX socket instead, the port would be
12985 replaced with the ID of the accepting socket, which is also reported in the
12986 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012987
12988 - "accept_date" is the exact date when the connection was received by haproxy
12989 (which might be very slightly different from the date observed on the
12990 network if there was some queuing in the system's backlog). This is usually
12991 the same date which may appear in any upstream firewall's log.
12992
12993 - "frontend_name" is the name of the frontend (or listener) which received
12994 and processed the connection.
12995
12996 - "backend_name" is the name of the backend (or listener) which was selected
12997 to manage the connection to the server. This will be the same as the
12998 frontend if no switching rule has been applied, which is common for TCP
12999 applications.
13000
13001 - "server_name" is the name of the last server to which the connection was
13002 sent, which might differ from the first one if there were connection errors
13003 and a redispatch occurred. Note that this server belongs to the backend
13004 which processed the request. If the connection was aborted before reaching
13005 a server, "<NOSRV>" is indicated instead of a server name.
13006
13007 - "Tw" is the total time in milliseconds spent waiting in the various queues.
13008 It can be "-1" if the connection was aborted before reaching the queue.
13009 See "Timers" below for more details.
13010
13011 - "Tc" is the total time in milliseconds spent waiting for the connection to
13012 establish to the final server, including retries. It can be "-1" if the
13013 connection was aborted before a connection could be established. See
13014 "Timers" below for more details.
13015
13016 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013017 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013018 "option logasap" was specified, then the time counting stops at the moment
13019 the log is emitted. In this case, a '+' sign is prepended before the value,
13020 indicating that the final one will be larger. See "Timers" below for more
13021 details.
13022
13023 - "bytes_read" is the total number of bytes transmitted from the server to
13024 the client when the log is emitted. If "option logasap" is specified, the
13025 this value will be prefixed with a '+' sign indicating that the final one
13026 may be larger. Please note that this value is a 64-bit counter, so log
13027 analysis tools must be able to handle it without overflowing.
13028
13029 - "termination_state" is the condition the session was in when the session
13030 ended. This indicates the session state, which side caused the end of
13031 session to happen, and for what reason (timeout, error, ...). The normal
13032 flags should be "--", indicating the session was closed by either end with
13033 no data remaining in buffers. See below "Session state at disconnection"
13034 for more details.
13035
13036 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040013037 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013038 limits have been reached. For instance, if actconn is close to 512 when
13039 multiple connection errors occur, chances are high that the system limits
13040 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013041 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013042
13043 - "feconn" is the total number of concurrent connections on the frontend when
13044 the session was logged. It is useful to estimate the amount of resource
13045 required to sustain high loads, and to detect when the frontend's "maxconn"
13046 has been reached. Most often when this value increases by huge jumps, it is
13047 because there is congestion on the backend servers, but sometimes it can be
13048 caused by a denial of service attack.
13049
13050 - "beconn" is the total number of concurrent connections handled by the
13051 backend when the session was logged. It includes the total number of
13052 concurrent connections active on servers as well as the number of
13053 connections pending in queues. It is useful to estimate the amount of
13054 additional servers needed to support high loads for a given application.
13055 Most often when this value increases by huge jumps, it is because there is
13056 congestion on the backend servers, but sometimes it can be caused by a
13057 denial of service attack.
13058
13059 - "srv_conn" is the total number of concurrent connections still active on
13060 the server when the session was logged. It can never exceed the server's
13061 configured "maxconn" parameter. If this value is very often close or equal
13062 to the server's "maxconn", it means that traffic regulation is involved a
13063 lot, meaning that either the server's maxconn value is too low, or that
13064 there aren't enough servers to process the load with an optimal response
13065 time. When only one of the server's "srv_conn" is high, it usually means
13066 that this server has some trouble causing the connections to take longer to
13067 be processed than on other servers.
13068
13069 - "retries" is the number of connection retries experienced by this session
13070 when trying to connect to the server. It must normally be zero, unless a
13071 server is being stopped at the same moment the connection was attempted.
13072 Frequent retries generally indicate either a network problem between
13073 haproxy and the server, or a misconfigured system backlog on the server
13074 preventing new connections from being queued. This field may optionally be
13075 prefixed with a '+' sign, indicating that the session has experienced a
13076 redispatch after the maximal retry count has been reached on the initial
13077 server. In this case, the server name appearing in the log is the one the
13078 connection was redispatched to, and not the first one, though both may
13079 sometimes be the same in case of hashing for instance. So as a general rule
13080 of thumb, when a '+' is present in front of the retry count, this count
13081 should not be attributed to the logged server.
13082
13083 - "srv_queue" is the total number of requests which were processed before
13084 this one in the server queue. It is zero when the request has not gone
13085 through the server queue. It makes it possible to estimate the approximate
13086 server's response time by dividing the time spent in queue by the number of
13087 requests in the queue. It is worth noting that if a session experiences a
13088 redispatch and passes through two server queues, their positions will be
13089 cumulated. A request should not pass through both the server queue and the
13090 backend queue unless a redispatch occurs.
13091
13092 - "backend_queue" is the total number of requests which were processed before
13093 this one in the backend's global queue. It is zero when the request has not
13094 gone through the global queue. It makes it possible to estimate the average
13095 queue length, which easily translates into a number of missing servers when
13096 divided by a server's "maxconn" parameter. It is worth noting that if a
13097 session experiences a redispatch, it may pass twice in the backend's queue,
13098 and then both positions will be cumulated. A request should not pass
13099 through both the server queue and the backend queue unless a redispatch
13100 occurs.
13101
13102
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200131038.2.3. HTTP log format
13104----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013105
13106The HTTP format is the most complete and the best suited for HTTP proxies. It
13107is enabled by when "option httplog" is specified in the frontend. It provides
13108the same level of information as the TCP format with additional features which
13109are specific to the HTTP protocol. Just like the TCP format, the log is usually
13110emitted at the end of the session, unless "option logasap" is specified, which
13111generally only makes sense for download sites. A session which matches the
13112"monitor" rules will never logged. It is also possible not to log sessions for
13113which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020013114frontend. Successful connections will not be logged if "option dontlog-normal"
13115is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013116
13117Most fields are shared with the TCP log, some being different. A few fields may
13118slightly vary depending on some configuration options. Those ones are marked
13119with a star ('*') after the field name below.
13120
13121 Example :
13122 frontend http-in
13123 mode http
13124 option httplog
13125 log global
13126 default_backend bck
13127
13128 backend static
13129 server srv1 127.0.0.1:8000
13130
13131 >>> Feb 6 12:14:14 localhost \
13132 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
13133 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 +010013134 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013135
13136 Field Format Extract from the example above
13137 1 process_name '[' pid ']:' haproxy[14389]:
13138 2 client_ip ':' client_port 10.0.1.2:33317
13139 3 '[' accept_date ']' [06/Feb/2009:12:14:14.655]
13140 4 frontend_name http-in
13141 5 backend_name '/' server_name static/srv1
13142 6 Tq '/' Tw '/' Tc '/' Tr '/' Tt* 10/0/30/69/109
13143 7 status_code 200
13144 8 bytes_read* 2750
13145 9 captured_request_cookie -
13146 10 captured_response_cookie -
13147 11 termination_state ----
13148 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
13149 13 srv_queue '/' backend_queue 0/0
13150 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
13151 15 '{' captured_response_headers* '}' {}
13152 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010013153
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013154
13155Detailed fields description :
13156 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010013157 connection to haproxy. If the connection was accepted on a UNIX socket
13158 instead, the IP address would be replaced with the word "unix". Note that
13159 when the connection is accepted on a socket configured with "accept-proxy"
13160 and the PROXY protocol is correctly used, then the logs will reflect the
13161 forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013162
13163 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010013164 If the connection was accepted on a UNIX socket instead, the port would be
13165 replaced with the ID of the accepting socket, which is also reported in the
13166 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013167
13168 - "accept_date" is the exact date when the TCP connection was received by
13169 haproxy (which might be very slightly different from the date observed on
13170 the network if there was some queuing in the system's backlog). This is
13171 usually the same date which may appear in any upstream firewall's log. This
13172 does not depend on the fact that the client has sent the request or not.
13173
13174 - "frontend_name" is the name of the frontend (or listener) which received
13175 and processed the connection.
13176
13177 - "backend_name" is the name of the backend (or listener) which was selected
13178 to manage the connection to the server. This will be the same as the
13179 frontend if no switching rule has been applied.
13180
13181 - "server_name" is the name of the last server to which the connection was
13182 sent, which might differ from the first one if there were connection errors
13183 and a redispatch occurred. Note that this server belongs to the backend
13184 which processed the request. If the request was aborted before reaching a
13185 server, "<NOSRV>" is indicated instead of a server name. If the request was
13186 intercepted by the stats subsystem, "<STATS>" is indicated instead.
13187
13188 - "Tq" is the total time in milliseconds spent waiting for the client to send
13189 a full HTTP request, not counting data. It can be "-1" if the connection
13190 was aborted before a complete request could be received. It should always
13191 be very small because a request generally fits in one single packet. Large
13192 times here generally indicate network trouble between the client and
13193 haproxy. See "Timers" below for more details.
13194
13195 - "Tw" is the total time in milliseconds spent waiting in the various queues.
13196 It can be "-1" if the connection was aborted before reaching the queue.
13197 See "Timers" below for more details.
13198
13199 - "Tc" is the total time in milliseconds spent waiting for the connection to
13200 establish to the final server, including retries. It can be "-1" if the
13201 request was aborted before a connection could be established. See "Timers"
13202 below for more details.
13203
13204 - "Tr" is the total time in milliseconds spent waiting for the server to send
13205 a full HTTP response, not counting data. It can be "-1" if the request was
13206 aborted before a complete response could be received. It generally matches
13207 the server's processing time for the request, though it may be altered by
13208 the amount of data sent by the client to the server. Large times here on
13209 "GET" requests generally indicate an overloaded server. See "Timers" below
13210 for more details.
13211
13212 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013213 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013214 "option logasap" was specified, then the time counting stops at the moment
13215 the log is emitted. In this case, a '+' sign is prepended before the value,
13216 indicating that the final one will be larger. See "Timers" below for more
13217 details.
13218
13219 - "status_code" is the HTTP status code returned to the client. This status
13220 is generally set by the server, but it might also be set by haproxy when
13221 the server cannot be reached or when its response is blocked by haproxy.
13222
13223 - "bytes_read" is the total number of bytes transmitted to the client when
13224 the log is emitted. This does include HTTP headers. If "option logasap" is
13225 specified, the this value will be prefixed with a '+' sign indicating that
13226 the final one may be larger. Please note that this value is a 64-bit
13227 counter, so log analysis tools must be able to handle it without
13228 overflowing.
13229
13230 - "captured_request_cookie" is an optional "name=value" entry indicating that
13231 the client had this cookie in the request. The cookie name and its maximum
13232 length are defined by the "capture cookie" statement in the frontend
13233 configuration. The field is a single dash ('-') when the option is not
13234 set. Only one cookie may be captured, it is generally used to track session
13235 ID exchanges between a client and a server to detect session crossing
13236 between clients due to application bugs. For more details, please consult
13237 the section "Capturing HTTP headers and cookies" below.
13238
13239 - "captured_response_cookie" is an optional "name=value" entry indicating
13240 that the server has returned a cookie with its response. The cookie name
13241 and its maximum length are defined by the "capture cookie" statement in the
13242 frontend configuration. The field is a single dash ('-') when the option is
13243 not set. Only one cookie may be captured, it is generally used to track
13244 session ID exchanges between a client and a server to detect session
13245 crossing between clients due to application bugs. For more details, please
13246 consult the section "Capturing HTTP headers and cookies" below.
13247
13248 - "termination_state" is the condition the session was in when the session
13249 ended. This indicates the session state, which side caused the end of
13250 session to happen, for what reason (timeout, error, ...), just like in TCP
13251 logs, and information about persistence operations on cookies in the last
13252 two characters. The normal flags should begin with "--", indicating the
13253 session was closed by either end with no data remaining in buffers. See
13254 below "Session state at disconnection" for more details.
13255
13256 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040013257 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013258 limits have been reached. For instance, if actconn is close to 512 or 1024
13259 when multiple connection errors occur, chances are high that the system
13260 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013261 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013262 system.
13263
13264 - "feconn" is the total number of concurrent connections on the frontend when
13265 the session was logged. It is useful to estimate the amount of resource
13266 required to sustain high loads, and to detect when the frontend's "maxconn"
13267 has been reached. Most often when this value increases by huge jumps, it is
13268 because there is congestion on the backend servers, but sometimes it can be
13269 caused by a denial of service attack.
13270
13271 - "beconn" is the total number of concurrent connections handled by the
13272 backend when the session was logged. It includes the total number of
13273 concurrent connections active on servers as well as the number of
13274 connections pending in queues. It is useful to estimate the amount of
13275 additional servers needed to support high loads for a given application.
13276 Most often when this value increases by huge jumps, it is because there is
13277 congestion on the backend servers, but sometimes it can be caused by a
13278 denial of service attack.
13279
13280 - "srv_conn" is the total number of concurrent connections still active on
13281 the server when the session was logged. It can never exceed the server's
13282 configured "maxconn" parameter. If this value is very often close or equal
13283 to the server's "maxconn", it means that traffic regulation is involved a
13284 lot, meaning that either the server's maxconn value is too low, or that
13285 there aren't enough servers to process the load with an optimal response
13286 time. When only one of the server's "srv_conn" is high, it usually means
13287 that this server has some trouble causing the requests to take longer to be
13288 processed than on other servers.
13289
13290 - "retries" is the number of connection retries experienced by this session
13291 when trying to connect to the server. It must normally be zero, unless a
13292 server is being stopped at the same moment the connection was attempted.
13293 Frequent retries generally indicate either a network problem between
13294 haproxy and the server, or a misconfigured system backlog on the server
13295 preventing new connections from being queued. This field may optionally be
13296 prefixed with a '+' sign, indicating that the session has experienced a
13297 redispatch after the maximal retry count has been reached on the initial
13298 server. In this case, the server name appearing in the log is the one the
13299 connection was redispatched to, and not the first one, though both may
13300 sometimes be the same in case of hashing for instance. So as a general rule
13301 of thumb, when a '+' is present in front of the retry count, this count
13302 should not be attributed to the logged server.
13303
13304 - "srv_queue" is the total number of requests which were processed before
13305 this one in the server queue. It is zero when the request has not gone
13306 through the server queue. It makes it possible to estimate the approximate
13307 server's response time by dividing the time spent in queue by the number of
13308 requests in the queue. It is worth noting that if a session experiences a
13309 redispatch and passes through two server queues, their positions will be
13310 cumulated. A request should not pass through both the server queue and the
13311 backend queue unless a redispatch occurs.
13312
13313 - "backend_queue" is the total number of requests which were processed before
13314 this one in the backend's global queue. It is zero when the request has not
13315 gone through the global queue. It makes it possible to estimate the average
13316 queue length, which easily translates into a number of missing servers when
13317 divided by a server's "maxconn" parameter. It is worth noting that if a
13318 session experiences a redispatch, it may pass twice in the backend's queue,
13319 and then both positions will be cumulated. A request should not pass
13320 through both the server queue and the backend queue unless a redispatch
13321 occurs.
13322
13323 - "captured_request_headers" is a list of headers captured in the request due
13324 to the presence of the "capture request header" statement in the frontend.
13325 Multiple headers can be captured, they will be delimited by a vertical bar
13326 ('|'). When no capture is enabled, the braces do not appear, causing a
13327 shift of remaining fields. It is important to note that this field may
13328 contain spaces, and that using it requires a smarter log parser than when
13329 it's not used. Please consult the section "Capturing HTTP headers and
13330 cookies" below for more details.
13331
13332 - "captured_response_headers" is a list of headers captured in the response
13333 due to the presence of the "capture response header" statement in the
13334 frontend. Multiple headers can be captured, they will be delimited by a
13335 vertical bar ('|'). When no capture is enabled, the braces do not appear,
13336 causing a shift of remaining fields. It is important to note that this
13337 field may contain spaces, and that using it requires a smarter log parser
13338 than when it's not used. Please consult the section "Capturing HTTP headers
13339 and cookies" below for more details.
13340
13341 - "http_request" is the complete HTTP request line, including the method,
13342 request and HTTP version string. Non-printable characters are encoded (see
13343 below the section "Non-printable characters"). This is always the last
13344 field, and it is always delimited by quotes and is the only one which can
13345 contain quotes. If new fields are added to the log format, they will be
13346 added before this field. This field might be truncated if the request is
13347 huge and does not fit in the standard syslog buffer (1024 characters). This
13348 is the reason why this field must always remain the last one.
13349
13350
Cyril Bontédc4d9032012-04-08 21:57:39 +0200133518.2.4. Custom log format
13352------------------------
William Lallemand48940402012-01-30 16:47:22 +010013353
Willy Tarreau2beef582012-12-20 17:22:52 +010013354The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010013355mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010013356
13357HAproxy understands some log format variables. % precedes log format variables.
13358Variables can take arguments using braces ('{}'), and multiple arguments are
13359separated by commas within the braces. Flags may be added or removed by
13360prefixing them with a '+' or '-' sign.
13361
13362Special variable "%o" may be used to propagate its flags to all other
13363variables on the same format string. This is particularly handy with quoted
13364string formats ("Q").
13365
Willy Tarreauc8368452012-12-21 00:09:23 +010013366If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreaube722a22014-06-13 16:31:59 +020013367as a sample expression rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010013368less common information such as the client's SSL certificate's DN, or to log
13369the key that would be used to store an entry into a stick table.
13370
William Lallemand48940402012-01-30 16:47:22 +010013371Note: spaces must be escaped. A space character is considered as a separator.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013372In order to emit a verbatim '%', it must be preceded by another '%' resulting
Willy Tarreau06d97f92013-12-02 17:45:48 +010013373in '%%'. HAProxy will automatically merge consecutive separators.
William Lallemand48940402012-01-30 16:47:22 +010013374
13375Flags are :
13376 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040013377 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
William Lallemand48940402012-01-30 16:47:22 +010013378
13379 Example:
13380
13381 log-format %T\ %t\ Some\ Text
13382 log-format %{+Q}o\ %t\ %s\ %{-Q}r
13383
13384At the moment, the default HTTP format is defined this way :
13385
Willy Tarreau2beef582012-12-20 17:22:52 +010013386 log-format %ci:%cp\ [%t]\ %ft\ %b/%s\ %Tq/%Tw/%Tc/%Tr/%Tt\ %ST\ %B\ %CC\ \
13387 %CS\ %tsc\ %ac/%fc/%bc/%sc/%rc\ %sq/%bq\ %hr\ %hs\ %{+Q}r
William Lallemand48940402012-01-30 16:47:22 +010013388
William Lallemandbddd4fd2012-02-27 11:23:10 +010013389the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010013390
Willy Tarreau2beef582012-12-20 17:22:52 +010013391 log-format %{+Q}o\ %{-Q}ci\ -\ -\ [%T]\ %r\ %ST\ %B\ \"\"\ \"\"\ %cp\ \
Willy Tarreau773d65f2012-10-12 14:56:11 +020013392 %ms\ %ft\ %b\ %s\ \%Tq\ %Tw\ %Tc\ %Tr\ %Tt\ %tsc\ %ac\ %fc\ \
Willy Tarreau2beef582012-12-20 17:22:52 +010013393 %bc\ %sc\ %rc\ %sq\ %bq\ %CC\ %CS\ \%hrl\ %hsl
William Lallemand48940402012-01-30 16:47:22 +010013394
William Lallemandbddd4fd2012-02-27 11:23:10 +010013395and the default TCP format is defined this way :
13396
Willy Tarreau2beef582012-12-20 17:22:52 +010013397 log-format %ci:%cp\ [%t]\ %ft\ %b/%s\ %Tw/%Tc/%Tt\ %B\ %ts\ \
William Lallemandbddd4fd2012-02-27 11:23:10 +010013398 %ac/%fc/%bc/%sc/%rc\ %sq/%bq
13399
William Lallemand48940402012-01-30 16:47:22 +010013400Please refer to the table below for currently defined variables :
13401
William Lallemandbddd4fd2012-02-27 11:23:10 +010013402 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020013403 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010013404 +---+------+-----------------------------------------------+-------------+
13405 | | %o | special variable, apply flags on all next var | |
13406 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010013407 | | %B | bytes_read (from server to client) | numeric |
13408 | H | %CC | captured_request_cookie | string |
13409 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020013410 | | %H | hostname | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000013411 | H | %HM | HTTP method (ex: POST) | string |
13412 | H | %HP | HTTP request URI without query string (path) | string |
13413 | H | %HU | HTTP request URI (ex: /foo?bar=baz) | string |
13414 | H | %HV | HTTP version (ex: HTTP/1.0) | string |
William Lallemanda73203e2012-03-12 12:48:57 +010013415 | | %ID | unique-id | string |
Willy Tarreau4bf99632014-06-13 12:21:40 +020013416 | | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020013417 | | %T | gmt_date_time | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010013418 | | %Tc | Tc | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080013419 | | %Tl | local_date_time | date |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020013420 | H | %Tq | Tq | numeric |
13421 | H | %Tr | Tr | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020013422 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010013423 | | %Tt | Tt | numeric |
13424 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010013425 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010013426 | | %ac | actconn | numeric |
13427 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010013428 | | %bc | beconn (backend concurrent connections) | numeric |
13429 | | %bi | backend_source_ip (connecting address) | IP |
13430 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010013431 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010013432 | | %ci | client_ip (accepted address) | IP |
13433 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010013434 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010013435 | | %fc | feconn (frontend concurrent connections) | numeric |
13436 | | %fi | frontend_ip (accepting address) | IP |
13437 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020013438 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreau7346acb2014-08-28 15:03:15 +020013439 | | %lc | frontend_log_counter | numeric |
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020013440 | | %hr | captured_request_headers default style | string |
13441 | | %hrl | captured_request_headers CLF style | string list |
13442 | | %hs | captured_response_headers default style | string |
13443 | | %hsl | captured_response_headers CLF style | string list |
William Lallemandbddd4fd2012-02-27 11:23:10 +010013444 | | %ms | accept date milliseconds | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020013445 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020013446 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010013447 | | %rc | retries | numeric |
Willy Tarreau1f0da242014-01-25 11:01:50 +010013448 | | %rt | request_counter (HTTP req or TCP session) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010013449 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010013450 | | %sc | srv_conn (server concurrent connections) | numeric |
13451 | | %si | server_IP (target address) | IP |
13452 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010013453 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020013454 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
13455 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010013456 | | %t | date_time (with millisecond resolution) | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010013457 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020013458 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010013459 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010013460
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020013461 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010013462
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010013463
134648.2.5. Error log format
13465-----------------------
13466
13467When an incoming connection fails due to an SSL handshake or an invalid PROXY
13468protocol header, haproxy will log the event using a shorter, fixed line format.
13469By default, logs are emitted at the LOG_INFO level, unless the option
13470"log-separate-errors" is set in the backend, in which case the LOG_ERR level
13471will be used. Connections on which no data are exchanged (eg: probes) are not
13472logged if the "dontlognull" option is set.
13473
13474The format looks like this :
13475
13476 >>> Dec 3 18:27:14 localhost \
13477 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
13478 Connection error during SSL handshake
13479
13480 Field Format Extract from the example above
13481 1 process_name '[' pid ']:' haproxy[6103]:
13482 2 client_ip ':' client_port 127.0.0.1:56059
13483 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
13484 4 frontend_name "/" bind_name ":" frt/f1:
13485 5 message Connection error during SSL handshake
13486
13487These fields just provide minimal information to help debugging connection
13488failures.
13489
13490
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200134918.3. Advanced logging options
13492-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013493
13494Some advanced logging options are often looked for but are not easy to find out
13495just by looking at the various options. Here is an entry point for the few
13496options which can enable better logging. Please refer to the keywords reference
13497for more information about their usage.
13498
13499
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200135008.3.1. Disabling logging of external tests
13501------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013502
13503It is quite common to have some monitoring tools perform health checks on
13504haproxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
13505commercial load-balancer, and sometimes it will simply be a more complete
13506monitoring system such as Nagios. When the tests are very frequent, users often
13507ask how to disable logging for those checks. There are three possibilities :
13508
13509 - if connections come from everywhere and are just TCP probes, it is often
13510 desired to simply disable logging of connections without data exchange, by
13511 setting "option dontlognull" in the frontend. It also disables logging of
13512 port scans, which may or may not be desired.
13513
13514 - if the connection come from a known source network, use "monitor-net" to
13515 declare this network as monitoring only. Any host in this network will then
13516 only be able to perform health checks, and their requests will not be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013517 logged. This is generally appropriate to designate a list of equipment
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013518 such as other load-balancers.
13519
13520 - if the tests are performed on a known URI, use "monitor-uri" to declare
13521 this URI as dedicated to monitoring. Any host sending this request will
13522 only get the result of a health-check, and the request will not be logged.
13523
13524
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200135258.3.2. Logging before waiting for the session to terminate
13526----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013527
13528The problem with logging at end of connection is that you have no clue about
13529what is happening during very long sessions, such as remote terminal sessions
13530or large file downloads. This problem can be worked around by specifying
13531"option logasap" in the frontend. Haproxy will then log as soon as possible,
13532just before data transfer begins. This means that in case of TCP, it will still
13533log the connection status to the server, and in case of HTTP, it will log just
13534after processing the server headers. In this case, the number of bytes reported
13535is the number of header bytes sent to the client. In order to avoid confusion
13536with normal logs, the total time field and the number of bytes are prefixed
13537with a '+' sign which means that real numbers are certainly larger.
13538
13539
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200135408.3.3. Raising log level upon errors
13541------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020013542
13543Sometimes it is more convenient to separate normal traffic from errors logs,
13544for instance in order to ease error monitoring from log files. When the option
13545"log-separate-errors" is used, connections which experience errors, timeouts,
13546retries, redispatches or HTTP status codes 5xx will see their syslog level
13547raised from "info" to "err". This will help a syslog daemon store the log in
13548a separate file. It is very important to keep the errors in the normal traffic
13549file too, so that log ordering is not altered. You should also be careful if
13550you already have configured your syslog daemon to store all logs higher than
13551"notice" in an "admin" file, because the "err" level is higher than "notice".
13552
13553
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200135548.3.4. Disabling logging of successful connections
13555--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020013556
13557Although this may sound strange at first, some large sites have to deal with
13558multiple thousands of logs per second and are experiencing difficulties keeping
13559them intact for a long time or detecting errors within them. If the option
13560"dontlog-normal" is set on the frontend, all normal connections will not be
13561logged. In this regard, a normal connection is defined as one without any
13562error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
13563and a response with a status 5xx is not considered normal and will be logged
13564too. Of course, doing is is really discouraged as it will remove most of the
13565useful information from the logs. Do this only if you have no other
13566alternative.
13567
13568
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200135698.4. Timing events
13570------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013571
13572Timers provide a great help in troubleshooting network problems. All values are
13573reported in milliseconds (ms). These timers should be used in conjunction with
13574the session termination flags. In TCP mode with "option tcplog" set on the
13575frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
13576mode, 5 control points are reported under the form "Tq/Tw/Tc/Tr/Tt" :
13577
13578 - Tq: total time to get the client request (HTTP mode only). It's the time
13579 elapsed between the moment the client connection was accepted and the
13580 moment the proxy received the last HTTP header. The value "-1" indicates
13581 that the end of headers (empty line) has never been seen. This happens when
13582 the client closes prematurely or times out.
13583
13584 - Tw: total time spent in the queues waiting for a connection slot. It
13585 accounts for backend queue as well as the server queues, and depends on the
13586 queue size, and the time needed for the server to complete previous
13587 requests. The value "-1" means that the request was killed before reaching
13588 the queue, which is generally what happens with invalid or denied requests.
13589
13590 - Tc: total time to establish the TCP connection to the server. It's the time
13591 elapsed between the moment the proxy sent the connection request, and the
13592 moment it was acknowledged by the server, or between the TCP SYN packet and
13593 the matching SYN/ACK packet in return. The value "-1" means that the
13594 connection never established.
13595
13596 - Tr: server response time (HTTP mode only). It's the time elapsed between
13597 the moment the TCP connection was established to the server and the moment
13598 the server sent its complete response headers. It purely shows its request
13599 processing time, without the network overhead due to the data transmission.
13600 It is worth noting that when the client has data to send to the server, for
13601 instance during a POST request, the time already runs, and this can distort
13602 apparent response time. For this reason, it's generally wise not to trust
13603 too much this field for POST requests initiated from clients behind an
13604 untrusted network. A value of "-1" here means that the last the response
13605 header (empty line) was never seen, most likely because the server timeout
13606 stroke before the server managed to process the request.
13607
13608 - Tt: total session duration time, between the moment the proxy accepted it
13609 and the moment both ends were closed. The exception is when the "logasap"
13610 option is specified. In this case, it only equals (Tq+Tw+Tc+Tr), and is
13611 prefixed with a '+' sign. From this field, we can deduce "Td", the data
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013612 transmission time, by subtracting other timers when valid :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013613
13614 Td = Tt - (Tq + Tw + Tc + Tr)
13615
13616 Timers with "-1" values have to be excluded from this equation. In TCP
13617 mode, "Tq" and "Tr" have to be excluded too. Note that "Tt" can never be
13618 negative.
13619
13620These timers provide precious indications on trouble causes. Since the TCP
13621protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
13622that timers close to multiples of 3s are nearly always related to lost packets
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010013623due to network problems (wires, negotiation, congestion). Moreover, if "Tt" is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013624close to a timeout value specified in the configuration, it often means that a
13625session has been aborted on timeout.
13626
13627Most common cases :
13628
13629 - If "Tq" is close to 3000, a packet has probably been lost between the
13630 client and the proxy. This is very rare on local networks but might happen
13631 when clients are on far remote networks and send large requests. It may
13632 happen that values larger than usual appear here without any network cause.
13633 Sometimes, during an attack or just after a resource starvation has ended,
13634 haproxy may accept thousands of connections in a few milliseconds. The time
13635 spent accepting these connections will inevitably slightly delay processing
13636 of other connections, and it can happen that request times in the order of
13637 a few tens of milliseconds are measured after a few thousands of new
Patrick Mezard105faca2010-06-12 17:02:46 +020013638 connections have been accepted at once. Setting "option http-server-close"
13639 may display larger request times since "Tq" also measures the time spent
13640 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013641
13642 - If "Tc" is close to 3000, a packet has probably been lost between the
13643 server and the proxy during the server connection phase. This value should
13644 always be very low, such as 1 ms on local networks and less than a few tens
13645 of ms on remote networks.
13646
Willy Tarreau55165fe2009-05-10 12:02:55 +020013647 - If "Tr" is nearly always lower than 3000 except some rare values which seem
13648 to be the average majored by 3000, there are probably some packets lost
13649 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013650
13651 - If "Tt" is large even for small byte counts, it generally is because
13652 neither the client nor the server decides to close the connection, for
13653 instance because both have agreed on a keep-alive connection mode. In order
13654 to solve this issue, it will be needed to specify "option httpclose" on
13655 either the frontend or the backend. If the problem persists, it means that
13656 the server ignores the "close" connection mode and expects the client to
13657 close. Then it will be required to use "option forceclose". Having the
13658 smallest possible 'Tt' is important when connection regulation is used with
13659 the "maxconn" option on the servers, since no new connection will be sent
13660 to the server until another one is released.
13661
13662Other noticeable HTTP log cases ('xx' means any value to be ignored) :
13663
13664 Tq/Tw/Tc/Tr/+Tt The "option logasap" is present on the frontend and the log
13665 was emitted before the data phase. All the timers are valid
13666 except "Tt" which is shorter than reality.
13667
13668 -1/xx/xx/xx/Tt The client was not able to send a complete request in time
13669 or it aborted too early. Check the session termination flags
13670 then "timeout http-request" and "timeout client" settings.
13671
13672 Tq/-1/xx/xx/Tt It was not possible to process the request, maybe because
13673 servers were out of order, because the request was invalid
13674 or forbidden by ACL rules. Check the session termination
13675 flags.
13676
13677 Tq/Tw/-1/xx/Tt The connection could not establish on the server. Either it
13678 actively refused it or it timed out after Tt-(Tq+Tw) ms.
13679 Check the session termination flags, then check the
13680 "timeout connect" setting. Note that the tarpit action might
13681 return similar-looking patterns, with "Tw" equal to the time
13682 the client connection was maintained open.
13683
13684 Tq/Tw/Tc/-1/Tt The server has accepted the connection but did not return
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013685 a complete response in time, or it closed its connection
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013686 unexpectedly after Tt-(Tq+Tw+Tc) ms. Check the session
13687 termination flags, then check the "timeout server" setting.
13688
13689
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200136908.5. Session state at disconnection
13691-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013692
13693TCP and HTTP logs provide a session termination indicator in the
13694"termination_state" field, just before the number of active connections. It is
136952-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
13696each of which has a special meaning :
13697
13698 - On the first character, a code reporting the first event which caused the
13699 session to terminate :
13700
13701 C : the TCP session was unexpectedly aborted by the client.
13702
13703 S : the TCP session was unexpectedly aborted by the server, or the
13704 server explicitly refused it.
13705
13706 P : the session was prematurely aborted by the proxy, because of a
13707 connection limit enforcement, because a DENY filter was matched,
13708 because of a security check which detected and blocked a dangerous
13709 error in server response which might have caused information leak
Willy Tarreau570f2212013-06-10 16:42:09 +020013710 (eg: cacheable cookie).
13711
13712 L : the session was locally processed by haproxy and was not passed to
13713 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013714
13715 R : a resource on the proxy has been exhausted (memory, sockets, source
13716 ports, ...). Usually, this appears during the connection phase, and
13717 system logs should contain a copy of the precise error. If this
13718 happens, it must be considered as a very serious anomaly which
13719 should be fixed as soon as possible by any means.
13720
13721 I : an internal error was identified by the proxy during a self-check.
13722 This should NEVER happen, and you are encouraged to report any log
13723 containing this, because this would almost certainly be a bug. It
13724 would be wise to preventively restart the process after such an
13725 event too, in case it would be caused by memory corruption.
13726
Simon Horman752dc4a2011-06-21 14:34:59 +090013727 D : the session was killed by haproxy because the server was detected
13728 as down and was configured to kill all connections when going down.
13729
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070013730 U : the session was killed by haproxy on this backup server because an
13731 active server was detected as up and was configured to kill all
13732 backup connections when going up.
13733
Willy Tarreaua2a64e92011-09-07 23:01:56 +020013734 K : the session was actively killed by an admin operating on haproxy.
13735
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013736 c : the client-side timeout expired while waiting for the client to
13737 send or receive data.
13738
13739 s : the server-side timeout expired while waiting for the server to
13740 send or receive data.
13741
13742 - : normal session completion, both the client and the server closed
13743 with nothing left in the buffers.
13744
13745 - on the second character, the TCP or HTTP session state when it was closed :
13746
Willy Tarreauf7b30a92010-12-06 22:59:17 +010013747 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013748 (HTTP mode only). Nothing was sent to any server.
13749
13750 Q : the proxy was waiting in the QUEUE for a connection slot. This can
13751 only happen when servers have a 'maxconn' parameter set. It can
13752 also happen in the global queue after a redispatch consecutive to
13753 a failed attempt to connect to a dying server. If no redispatch is
13754 reported, then no connection attempt was made to any server.
13755
13756 C : the proxy was waiting for the CONNECTION to establish on the
13757 server. The server might at most have noticed a connection attempt.
13758
13759 H : the proxy was waiting for complete, valid response HEADERS from the
13760 server (HTTP only).
13761
13762 D : the session was in the DATA phase.
13763
13764 L : the proxy was still transmitting LAST data to the client while the
13765 server had already finished. This one is very rare as it can only
13766 happen when the client dies while receiving the last packets.
13767
13768 T : the request was tarpitted. It has been held open with the client
13769 during the whole "timeout tarpit" duration or until the client
13770 closed, both of which will be reported in the "Tw" timer.
13771
13772 - : normal session completion after end of data transfer.
13773
13774 - the third character tells whether the persistence cookie was provided by
13775 the client (only in HTTP mode) :
13776
13777 N : the client provided NO cookie. This is usually the case for new
13778 visitors, so counting the number of occurrences of this flag in the
13779 logs generally indicate a valid trend for the site frequentation.
13780
13781 I : the client provided an INVALID cookie matching no known server.
13782 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020013783 cookies between HTTP/HTTPS sites, persistence conditionally
13784 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013785
13786 D : the client provided a cookie designating a server which was DOWN,
13787 so either "option persist" was used and the client was sent to
13788 this server, or it was not set and the client was redispatched to
13789 another server.
13790
Willy Tarreau996a92c2010-10-13 19:30:47 +020013791 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013792 server.
13793
Willy Tarreau996a92c2010-10-13 19:30:47 +020013794 E : the client provided a valid cookie, but with a last date which was
13795 older than what is allowed by the "maxidle" cookie parameter, so
13796 the cookie is consider EXPIRED and is ignored. The request will be
13797 redispatched just as if there was no cookie.
13798
13799 O : the client provided a valid cookie, but with a first date which was
13800 older than what is allowed by the "maxlife" cookie parameter, so
13801 the cookie is consider too OLD and is ignored. The request will be
13802 redispatched just as if there was no cookie.
13803
Willy Tarreauc89ccb62012-04-05 21:18:22 +020013804 U : a cookie was present but was not used to select the server because
13805 some other server selection mechanism was used instead (typically a
13806 "use-server" rule).
13807
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013808 - : does not apply (no cookie set in configuration).
13809
13810 - the last character reports what operations were performed on the persistence
13811 cookie returned by the server (only in HTTP mode) :
13812
13813 N : NO cookie was provided by the server, and none was inserted either.
13814
13815 I : no cookie was provided by the server, and the proxy INSERTED one.
13816 Note that in "cookie insert" mode, if the server provides a cookie,
13817 it will still be overwritten and reported as "I" here.
13818
Willy Tarreau996a92c2010-10-13 19:30:47 +020013819 U : the proxy UPDATED the last date in the cookie that was presented by
13820 the client. This can only happen in insert mode with "maxidle". It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013821 happens every time there is activity at a different date than the
Willy Tarreau996a92c2010-10-13 19:30:47 +020013822 date indicated in the cookie. If any other change happens, such as
13823 a redispatch, then the cookie will be marked as inserted instead.
13824
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013825 P : a cookie was PROVIDED by the server and transmitted as-is.
13826
13827 R : the cookie provided by the server was REWRITTEN by the proxy, which
13828 happens in "cookie rewrite" or "cookie prefix" modes.
13829
13830 D : the cookie provided by the server was DELETED by the proxy.
13831
13832 - : does not apply (no cookie set in configuration).
13833
Willy Tarreau996a92c2010-10-13 19:30:47 +020013834The combination of the two first flags gives a lot of information about what
13835was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013836helpful to detect server saturation, network troubles, local system resource
13837starvation, attacks, etc...
13838
13839The most common termination flags combinations are indicated below. They are
13840alphabetically sorted, with the lowercase set just after the upper case for
13841easier finding and understanding.
13842
13843 Flags Reason
13844
13845 -- Normal termination.
13846
13847 CC The client aborted before the connection could be established to the
13848 server. This can happen when haproxy tries to connect to a recently
13849 dead (or unchecked) server, and the client aborts while haproxy is
13850 waiting for the server to respond or for "timeout connect" to expire.
13851
13852 CD The client unexpectedly aborted during data transfer. This can be
13853 caused by a browser crash, by an intermediate equipment between the
13854 client and haproxy which decided to actively break the connection,
13855 by network routing issues between the client and haproxy, or by a
13856 keep-alive session between the server and the client terminated first
13857 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010013858
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013859 cD The client did not send nor acknowledge any data for as long as the
13860 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020013861 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013862
13863 CH The client aborted while waiting for the server to start responding.
13864 It might be the server taking too long to respond or the client
13865 clicking the 'Stop' button too fast.
13866
13867 cH The "timeout client" stroke while waiting for client data during a
13868 POST request. This is sometimes caused by too large TCP MSS values
13869 for PPPoE networks which cannot transport full-sized packets. It can
13870 also happen when client timeout is smaller than server timeout and
13871 the server takes too long to respond.
13872
13873 CQ The client aborted while its session was queued, waiting for a server
13874 with enough empty slots to accept it. It might be that either all the
13875 servers were saturated or that the assigned server was taking too
13876 long a time to respond.
13877
13878 CR The client aborted before sending a full HTTP request. Most likely
13879 the request was typed by hand using a telnet client, and aborted
13880 too early. The HTTP status code is likely a 400 here. Sometimes this
13881 might also be caused by an IDS killing the connection between haproxy
Willy Tarreau0f228a02015-05-01 15:37:53 +020013882 and the client. "option http-ignore-probes" can be used to ignore
13883 connections without any data transfer.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013884
13885 cR The "timeout http-request" stroke before the client sent a full HTTP
13886 request. This is sometimes caused by too large TCP MSS values on the
13887 client side for PPPoE networks which cannot transport full-sized
13888 packets, or by clients sending requests by hand and not typing fast
13889 enough, or forgetting to enter the empty line at the end of the
Willy Tarreau2705a612014-05-23 17:38:34 +020013890 request. The HTTP status code is likely a 408 here. Note: recently,
Willy Tarreau0f228a02015-05-01 15:37:53 +020013891 some browsers started to implement a "pre-connect" feature consisting
13892 in speculatively connecting to some recently visited web sites just
13893 in case the user would like to visit them. This results in many
13894 connections being established to web sites, which end up in 408
13895 Request Timeout if the timeout strikes first, or 400 Bad Request when
13896 the browser decides to close them first. These ones pollute the log
13897 and feed the error counters. Some versions of some browsers have even
13898 been reported to display the error code. It is possible to work
13899 around the undesirable effects of this behaviour by adding "option
13900 http-ignore-probes" in the frontend, resulting in connections with
13901 zero data transfer to be totally ignored. This will definitely hide
13902 the errors of people experiencing connectivity issues though.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013903
13904 CT The client aborted while its session was tarpitted. It is important to
13905 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020013906 wrong tarpit rules have been written. If a lot of them happen, it
13907 might make sense to lower the "timeout tarpit" value to something
13908 closer to the average reported "Tw" timer, in order not to consume
13909 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013910
Willy Tarreau570f2212013-06-10 16:42:09 +020013911 LR The request was intercepted and locally handled by haproxy. Generally
13912 it means that this was a redirect or a stats request.
13913
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010013914 SC The server or an equipment between it and haproxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013915 the TCP connection (the proxy received a TCP RST or an ICMP message
13916 in return). Under some circumstances, it can also be the network
13917 stack telling the proxy that the server is unreachable (eg: no route,
13918 or no ARP response on local network). When this happens in HTTP mode,
13919 the status code is likely a 502 or 503 here.
13920
13921 sC The "timeout connect" stroke before a connection to the server could
13922 complete. When this happens in HTTP mode, the status code is likely a
13923 503 or 504 here.
13924
13925 SD The connection to the server died with an error during the data
13926 transfer. This usually means that haproxy has received an RST from
13927 the server or an ICMP message from an intermediate equipment while
13928 exchanging data with the server. This can be caused by a server crash
13929 or by a network issue on an intermediate equipment.
13930
13931 sD The server did not send nor acknowledge any data for as long as the
13932 "timeout server" setting during the data phase. This is often caused
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010013933 by too short timeouts on L4 equipments before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013934 load-balancers, ...), as well as keep-alive sessions maintained
13935 between the client and the server expiring first on haproxy.
13936
13937 SH The server aborted before sending its full HTTP response headers, or
13938 it crashed while processing the request. Since a server aborting at
13939 this moment is very rare, it would be wise to inspect its logs to
13940 control whether it crashed and why. The logged request may indicate a
13941 small set of faulty requests, demonstrating bugs in the application.
13942 Sometimes this might also be caused by an IDS killing the connection
13943 between haproxy and the server.
13944
13945 sH The "timeout server" stroke before the server could return its
13946 response headers. This is the most common anomaly, indicating too
13947 long transactions, probably caused by server or database saturation.
13948 The immediate workaround consists in increasing the "timeout server"
13949 setting, but it is important to keep in mind that the user experience
13950 will suffer from these long response times. The only long term
13951 solution is to fix the application.
13952
13953 sQ The session spent too much time in queue and has been expired. See
13954 the "timeout queue" and "timeout connect" settings to find out how to
13955 fix this if it happens too often. If it often happens massively in
13956 short periods, it may indicate general problems on the affected
13957 servers due to I/O or database congestion, or saturation caused by
13958 external attacks.
13959
13960 PC The proxy refused to establish a connection to the server because the
13961 process' socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020013962 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013963 so that it does not happen anymore. This status is very rare and
13964 might happen when the global "ulimit-n" parameter is forced by hand.
13965
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010013966 PD The proxy blocked an incorrectly formatted chunked encoded message in
13967 a request or a response, after the server has emitted its headers. In
13968 most cases, this will indicate an invalid message from the server to
Willy Tarreauf3a3e132013-08-31 08:16:26 +020013969 the client. Haproxy supports chunk sizes of up to 2GB - 1 (2147483647
13970 bytes). Any larger size will be considered as an error.
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010013971
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013972 PH The proxy blocked the server's response, because it was invalid,
13973 incomplete, dangerous (cache control), or matched a security filter.
13974 In any case, an HTTP 502 error is sent to the client. One possible
13975 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010013976 containing unauthorized characters. It is also possible but quite
13977 rare, that the proxy blocked a chunked-encoding request from the
13978 client due to an invalid syntax, before the server responded. In this
13979 case, an HTTP 400 error is sent to the client and reported in the
13980 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013981
13982 PR The proxy blocked the client's HTTP request, either because of an
13983 invalid HTTP syntax, in which case it returned an HTTP 400 error to
13984 the client, or because a deny filter matched, in which case it
13985 returned an HTTP 403 error.
13986
13987 PT The proxy blocked the client's request and has tarpitted its
13988 connection before returning it a 500 server error. Nothing was sent
13989 to the server. The connection was maintained open for as long as
13990 reported by the "Tw" timer field.
13991
13992 RC A local resource has been exhausted (memory, sockets, source ports)
13993 preventing the connection to the server from establishing. The error
13994 logs will tell precisely what was missing. This is very rare and can
13995 only be solved by proper system tuning.
13996
Willy Tarreau996a92c2010-10-13 19:30:47 +020013997The combination of the two last flags gives a lot of information about how
13998persistence was handled by the client, the server and by haproxy. This is very
13999important to troubleshoot disconnections, when users complain they have to
14000re-authenticate. The commonly encountered flags are :
14001
14002 -- Persistence cookie is not enabled.
14003
14004 NN No cookie was provided by the client, none was inserted in the
14005 response. For instance, this can be in insert mode with "postonly"
14006 set on a GET request.
14007
14008 II A cookie designating an invalid server was provided by the client,
14009 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014010 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020014011 value can be presented by a client when no other server knows it.
14012
14013 NI No cookie was provided by the client, one was inserted in the
14014 response. This typically happens for first requests from every user
14015 in "insert" mode, which makes it an easy way to count real users.
14016
14017 VN A cookie was provided by the client, none was inserted in the
14018 response. This happens for most responses for which the client has
14019 already got a cookie.
14020
14021 VU A cookie was provided by the client, with a last visit date which is
14022 not completely up-to-date, so an updated cookie was provided in
14023 response. This can also happen if there was no date at all, or if
14024 there was a date but the "maxidle" parameter was not set, so that the
14025 cookie can be switched to unlimited time.
14026
14027 EI A cookie was provided by the client, with a last visit date which is
14028 too old for the "maxidle" parameter, so the cookie was ignored and a
14029 new cookie was inserted in the response.
14030
14031 OI A cookie was provided by the client, with a first visit date which is
14032 too old for the "maxlife" parameter, so the cookie was ignored and a
14033 new cookie was inserted in the response.
14034
14035 DI The server designated by the cookie was down, a new server was
14036 selected and a new cookie was emitted in the response.
14037
14038 VI The server designated by the cookie was not marked dead but could not
14039 be reached. A redispatch happened and selected another one, which was
14040 then advertised in the response.
14041
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014042
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200140438.6. Non-printable characters
14044-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014045
14046In order not to cause trouble to log analysis tools or terminals during log
14047consulting, non-printable characters are not sent as-is into log files, but are
14048converted to the two-digits hexadecimal representation of their ASCII code,
14049prefixed by the character '#'. The only characters that can be logged without
14050being escaped are comprised between 32 and 126 (inclusive). Obviously, the
14051escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
14052is the same for the character '"' which becomes "#22", as well as '{', '|' and
14053'}' when logging headers.
14054
14055Note that the space character (' ') is not encoded in headers, which can cause
14056issues for tools relying on space count to locate fields. A typical header
14057containing spaces is "User-Agent".
14058
14059Last, it has been observed that some syslog daemons such as syslog-ng escape
14060the quote ('"') with a backslash ('\'). The reverse operation can safely be
14061performed since no quote may appear anywhere else in the logs.
14062
14063
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200140648.7. Capturing HTTP cookies
14065---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014066
14067Cookie capture simplifies the tracking a complete user session. This can be
14068achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014069section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014070cookie will simultaneously be checked in the request ("Cookie:" header) and in
14071the response ("Set-Cookie:" header). The respective values will be reported in
14072the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014073locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014074not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
14075user switches to a new session for example, because the server will reassign it
14076a new cookie. It is also possible to detect if a server unexpectedly sets a
14077wrong cookie to a client, leading to session crossing.
14078
14079 Examples :
14080 # capture the first cookie whose name starts with "ASPSESSION"
14081 capture cookie ASPSESSION len 32
14082
14083 # capture the first cookie whose name is exactly "vgnvisitor"
14084 capture cookie vgnvisitor= len 32
14085
14086
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200140878.8. Capturing HTTP headers
14088---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014089
14090Header captures are useful to track unique request identifiers set by an upper
14091proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
14092the response, one can search for information about the response length, how the
14093server asked the cache to behave, or an object location during a redirection.
14094
14095Header captures are performed using the "capture request header" and "capture
14096response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014097section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014098
14099It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010014100time. Non-existent headers are logged as empty strings, and if one header
14101appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014102are grouped within braces '{' and '}' in the same order as they were declared,
14103and delimited with a vertical bar '|' without any space. Response headers
14104follow the same representation, but are displayed after a space following the
14105request headers block. These blocks are displayed just before the HTTP request
14106in the logs.
14107
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020014108As a special case, it is possible to specify an HTTP header capture in a TCP
14109frontend. The purpose is to enable logging of headers which will be parsed in
14110an HTTP backend if the request is then switched to this HTTP backend.
14111
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014112 Example :
14113 # This instance chains to the outgoing proxy
14114 listen proxy-out
14115 mode http
14116 option httplog
14117 option logasap
14118 log global
14119 server cache1 192.168.1.1:3128
14120
14121 # log the name of the virtual server
14122 capture request header Host len 20
14123
14124 # log the amount of data uploaded during a POST
14125 capture request header Content-Length len 10
14126
14127 # log the beginning of the referrer
14128 capture request header Referer len 20
14129
14130 # server name (useful for outgoing proxies only)
14131 capture response header Server len 20
14132
14133 # logging the content-length is useful with "option logasap"
14134 capture response header Content-Length len 10
14135
14136 # log the expected cache behaviour on the response
14137 capture response header Cache-Control len 8
14138
14139 # the Via header will report the next proxy's name
14140 capture response header Via len 20
14141
14142 # log the URL location during a redirection
14143 capture response header Location len 20
14144
14145 >>> Aug 9 20:26:09 localhost \
14146 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
14147 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
14148 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
14149 "GET http://fr.adserver.yahoo.com/"
14150
14151 >>> Aug 9 20:30:46 localhost \
14152 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
14153 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
14154 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010014155 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014156
14157 >>> Aug 9 20:30:46 localhost \
14158 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
14159 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
14160 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
14161 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010014162 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014163
14164
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200141658.9. Examples of logs
14166---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014167
14168These are real-world examples of logs accompanied with an explanation. Some of
14169them have been made up by hand. The syslog part has been removed for better
14170reading. Their sole purpose is to explain how to decipher them.
14171
14172 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
14173 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
14174 "HEAD / HTTP/1.0"
14175
14176 => long request (6.5s) entered by hand through 'telnet'. The server replied
14177 in 147 ms, and the session ended normally ('----')
14178
14179 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
14180 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
14181 0/9 "HEAD / HTTP/1.0"
14182
14183 => Idem, but the request was queued in the global queue behind 9 other
14184 requests, and waited there for 1230 ms.
14185
14186 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
14187 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
14188 "GET /image.iso HTTP/1.0"
14189
14190 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010014191 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014192 14 ms, 243 bytes of headers were sent to the client, and total time from
14193 accept to first data byte is 30 ms.
14194
14195 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
14196 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
14197 "GET /cgi-bin/bug.cgi? HTTP/1.0"
14198
14199 => the proxy blocked a server response either because of an "rspdeny" or
14200 "rspideny" filter, or because the response was improperly formatted and
Willy Tarreau3c92c5f2011-08-28 09:45:47 +020014201 not HTTP-compliant, or because it blocked sensitive information which
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014202 risked being cached. In this case, the response is replaced with a "502
14203 bad gateway". The flags ("PH--") tell us that it was haproxy who decided
14204 to return the 502 and not the server.
14205
14206 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010014207 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 +010014208
14209 => the client never completed its request and aborted itself ("C---") after
14210 8.5s, while the proxy was waiting for the request headers ("-R--").
14211 Nothing was sent to any server.
14212
14213 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
14214 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
14215
14216 => The client never completed its request, which was aborted by the
14217 time-out ("c---") after 50s, while the proxy was waiting for the request
14218 headers ("-R--"). Nothing was sent to any server, but the proxy could
14219 send a 408 return code to the client.
14220
14221 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
14222 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
14223
14224 => This log was produced with "option tcplog". The client timed out after
14225 5 seconds ("c----").
14226
14227 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
14228 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010014229 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014230
14231 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014232 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014233 (config says 'retries 3'), and no redispatch (otherwise we would have
14234 seen "/+3"). Status code 503 was returned to the client. There were 115
14235 connections on this server, 202 connections on this proxy, and 205 on
14236 the global process. It is possible that the server refused the
14237 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010014238
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010014239
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200142409. Statistics and monitoring
14241----------------------------
14242
14243It is possible to query HAProxy about its status. The most commonly used
14244mechanism is the HTTP statistics page. This page also exposes an alternative
14245CSV output format for monitoring tools. The same format is provided on the
14246Unix socket.
14247
14248
142499.1. CSV format
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010014250---------------
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +010014251
Willy Tarreau7f062c42009-03-05 18:43:00 +010014252The statistics may be consulted either from the unix socket or from the HTTP
Willy Tarreaua3310dc2014-06-16 15:43:21 +020014253page. Both means provide a CSV format whose fields follow. The first line
14254begins with a sharp ('#') and has one word per comma-delimited field which
14255represents the title of the column. All other lines starting at the second one
14256use a classical CSV format using a comma as the delimiter, and the double quote
14257('"') as an optional text delimiter, but only if the enclosed text is ambiguous
14258(if it contains a quote or a comma). The double-quote character ('"') in the
14259text is doubled ('""'), which is the format that most tools recognize. Please
14260do not insert any column before these ones in order not to break tools which
14261use hard-coded column positions.
Willy Tarreau7f062c42009-03-05 18:43:00 +010014262
James Westbyebe62d62014-07-08 10:14:57 -040014263In brackets after each field name are the types which may have a value for
14264that field. The types are L (Listeners), F (Frontends), B (Backends), and
14265S (Servers).
14266
14267 0. pxname [LFBS]: proxy name
14268 1. svname [LFBS]: service name (FRONTEND for frontend, BACKEND for backend,
14269 any name for server/listener)
14270 2. qcur [..BS]: current queued requests. For the backend this reports the
14271 number queued without a server assigned.
14272 3. qmax [..BS]: max value of qcur
14273 4. scur [LFBS]: current sessions
14274 5. smax [LFBS]: max sessions
14275 6. slim [LFBS]: configured session limit
14276 7. stot [LFBS]: cumulative number of connections
14277 8. bin [LFBS]: bytes in
14278 9. bout [LFBS]: bytes out
14279 10. dreq [LFB.]: requests denied because of security concerns.
14280 - For tcp this is because of a matched tcp-request content rule.
14281 - For http this is because of a matched http-request or tarpit rule.
14282 11. dresp [LFBS]: responses denied because of security concerns.
14283 - For http this is because of a matched http-request rule, or
14284 "option checkcache".
14285 12. ereq [LF..]: request errors. Some of the possible causes are:
14286 - early termination from the client, before the request has been sent.
14287 - read error from the client
14288 - client timeout
14289 - client closed connection
14290 - various bad requests from the client.
14291 - request was tarpitted.
14292 13. econ [..BS]: number of requests that encountered an error trying to
14293 connect to a backend server. The backend stat is the sum of the stat
14294 for all servers of that backend, plus any connection errors not
14295 associated with a particular server (such as the backend having no
14296 active servers).
14297 14. eresp [..BS]: response errors. srv_abrt will be counted here also.
14298 Some other errors are:
14299 - write error on the client socket (won't be counted for the server stat)
14300 - failure applying filters to the response.
14301 15. wretr [..BS]: number of times a connection to a server was retried.
14302 16. wredis [..BS]: number of times a request was redispatched to another
14303 server. The server value counts the number of times that server was
14304 switched away from.
14305 17. status [LFBS]: status (UP/DOWN/NOLB/MAINT/MAINT(via)...)
Pavlos Parissis1f673c72015-05-02 20:30:44 +020014306 18. weight [..BS]: total weight (backend), server weight (server)
14307 19. act [..BS]: number of active servers (backend), server is active (server)
14308 20. bck [..BS]: number of backup servers (backend), server is backup (server)
James Westbyebe62d62014-07-08 10:14:57 -040014309 21. chkfail [...S]: number of failed checks. (Only counts checks failed when
14310 the server is up.)
14311 22. chkdown [..BS]: number of UP->DOWN transitions. The backend counter counts
14312 transitions to the whole backend being down, rather than the sum of the
14313 counters for each server.
14314 23. lastchg [..BS]: number of seconds since the last UP<->DOWN transition
14315 24. downtime [..BS]: total downtime (in seconds). The value for the backend
14316 is the downtime for the whole backend, not the sum of the server downtime.
14317 25. qlimit [...S]: configured maxqueue for the server, or nothing in the
14318 value is 0 (default, meaning no limit)
14319 26. pid [LFBS]: process id (0 for first instance, 1 for second, ...)
14320 27. iid [LFBS]: unique proxy id
14321 28. sid [L..S]: server id (unique inside a proxy)
14322 29. throttle [...S]: current throttle percentage for the server, when
14323 slowstart is active, or no value if not in slowstart.
14324 30. lbtot [..BS]: total number of times a server was selected, either for new
14325 sessions, or when re-dispatching. The server counter is the number
14326 of times that server was selected.
14327 31. tracked [...S]: id of proxy/server if tracking is enabled.
14328 32. type [LFBS]: (0=frontend, 1=backend, 2=server, 3=socket/listener)
14329 33. rate [.FBS]: number of sessions per second over last elapsed second
14330 34. rate_lim [.F..]: configured limit on new sessions per second
14331 35. rate_max [.FBS]: max number of new sessions per second
14332 36. check_status [...S]: status of last health check, one of:
Cyril Bontéf0c60612010-02-06 14:44:47 +010014333 UNK -> unknown
14334 INI -> initializing
14335 SOCKERR -> socket error
14336 L4OK -> check passed on layer 4, no upper layers testing enabled
Jason Harvey83104802015-04-16 11:13:21 -080014337 L4TOUT -> layer 1-4 timeout
Cyril Bontéf0c60612010-02-06 14:44:47 +010014338 L4CON -> layer 1-4 connection problem, for example
14339 "Connection refused" (tcp rst) or "No route to host" (icmp)
14340 L6OK -> check passed on layer 6
14341 L6TOUT -> layer 6 (SSL) timeout
14342 L6RSP -> layer 6 invalid response - protocol error
14343 L7OK -> check passed on layer 7
14344 L7OKC -> check conditionally passed on layer 7, for example 404 with
14345 disable-on-404
14346 L7TOUT -> layer 7 (HTTP/SMTP) timeout
14347 L7RSP -> layer 7 invalid response - protocol error
14348 L7STS -> layer 7 response error, for example HTTP 5xx
James Westbyebe62d62014-07-08 10:14:57 -040014349 37. check_code [...S]: layer5-7 code, if available
14350 38. check_duration [...S]: time in ms took to finish last health check
14351 39. hrsp_1xx [.FBS]: http responses with 1xx code
14352 40. hrsp_2xx [.FBS]: http responses with 2xx code
14353 41. hrsp_3xx [.FBS]: http responses with 3xx code
14354 42. hrsp_4xx [.FBS]: http responses with 4xx code
14355 43. hrsp_5xx [.FBS]: http responses with 5xx code
14356 44. hrsp_other [.FBS]: http responses with other codes (protocol error)
14357 45. hanafail [...S]: failed health checks details
14358 46. req_rate [.F..]: HTTP requests per second over last elapsed second
14359 47. req_rate_max [.F..]: max number of HTTP requests per second observed
14360 48. req_tot [.F..]: total number of HTTP requests received
14361 49. cli_abrt [..BS]: number of data transfers aborted by the client
14362 50. srv_abrt [..BS]: number of data transfers aborted by the server
14363 (inc. in eresp)
14364 51. comp_in [.FB.]: number of HTTP response bytes fed to the compressor
14365 52. comp_out [.FB.]: number of HTTP response bytes emitted by the compressor
14366 53. comp_byp [.FB.]: number of bytes that bypassed the HTTP compressor
14367 (CPU/BW limit)
14368 54. comp_rsp [.FB.]: number of HTTP responses that were compressed
14369 55. lastsess [..BS]: number of seconds since last session assigned to
14370 server/backend
14371 56. last_chk [...S]: last health check contents or textual error
14372 57. last_agt [...S]: last agent check contents or textual error
14373 58. qtime [..BS]: the average queue time in ms over the 1024 last requests
14374 59. ctime [..BS]: the average connect time in ms over the 1024 last requests
14375 60. rtime [..BS]: the average response time in ms over the 1024 last requests
14376 (0 for TCP)
14377 61. ttime [..BS]: the average total session time in ms over the 1024 last
14378 requests
Willy Tarreau844e3c52008-01-11 16:28:18 +010014379
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010014380
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200143819.2. Unix Socket commands
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010014382-------------------------
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010014383
Willy Tarreau468f4932013-08-01 16:50:16 +020014384The stats socket is not enabled by default. In order to enable it, it is
14385necessary to add one line in the global section of the haproxy configuration.
14386A second line is recommended to set a larger timeout, always appreciated when
14387issuing commands by hand :
14388
14389 global
14390 stats socket /var/run/haproxy.sock mode 600 level admin
14391 stats timeout 2m
14392
14393It is also possible to add multiple instances of the stats socket by repeating
14394the line, and make them listen to a TCP port instead of a UNIX socket. This is
14395never done by default because this is dangerous, but can be handy in some
14396situations :
14397
14398 global
14399 stats socket /var/run/haproxy.sock mode 600 level admin
14400 stats socket ipv4@192.168.0.1:9999 level admin
14401 stats timeout 2m
14402
14403To access the socket, an external utility such as "socat" is required. Socat is a
14404swiss-army knife to connect anything to anything. We use it to connect terminals
14405to the socket, or a couple of stdin/stdout pipes to it for scripts. The two main
14406syntaxes we'll use are the following :
14407
14408 # socat /var/run/haproxy.sock stdio
14409 # socat /var/run/haproxy.sock readline
14410
14411The first one is used with scripts. It is possible to send the output of a
14412script to haproxy, and pass haproxy's output to another script. That's useful
14413for retrieving counters or attack traces for example.
14414
14415The second one is only useful for issuing commands by hand. It has the benefit
14416that the terminal is handled by the readline library which supports line
14417editing and history, which is very convenient when issuing repeated commands
14418(eg: watch a counter).
14419
14420The socket supports two operation modes :
14421 - interactive
14422 - non-interactive
14423
14424The non-interactive mode is the default when socat connects to the socket. In
14425this mode, a single line may be sent. It is processed as a whole, responses are
14426sent back, and the connection closes after the end of the response. This is the
14427mode that scripts and monitoring tools use. It is possible to send multiple
14428commands in this mode, they need to be delimited by a semi-colon (';'). For
14429example :
14430
14431 # echo "show info;show stat;show table" | socat /var/run/haproxy stdio
14432
14433The interactive mode displays a prompt ('>') and waits for commands to be
14434entered on the line, then processes them, and displays the prompt again to wait
14435for a new command. This mode is entered via the "prompt" command which must be
14436sent on the first line in non-interactive mode. The mode is a flip switch, if
14437"prompt" is sent in interactive mode, it is disabled and the connection closes
14438after processing the last command of the same line.
14439
14440For this reason, when debugging by hand, it's quite common to start with the
14441"prompt" command :
14442
14443 # socat /var/run/haproxy readline
14444 prompt
14445 > show info
14446 ...
14447 >
14448
14449Since multiple commands may be issued at once, haproxy uses the empty line as a
14450delimiter to mark an end of output for each command, and takes care of ensuring
14451that no command can emit an empty line on output. A script can thus easily
14452parse the output even when multiple commands were pipelined on a single line.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010014453
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014454It is important to understand that when multiple haproxy processes are started
14455on the same sockets, any process may pick up the request and will output its
14456own stats.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010014457
Willy Tarreau468f4932013-08-01 16:50:16 +020014458The list of commands currently supported on the stats socket is provided below.
14459If an unknown command is sent, haproxy displays the usage message which reminds
14460all supported commands. Some commands support a more complex syntax, generally
14461it will explain what part of the command is invalid when this happens.
14462
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010014463add acl <acl> <pattern>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014464 Add an entry into the acl <acl>. <acl> is the #<id> or the <file> returned by
14465 "show acl". This command does not verify if the entry already exists. This
14466 command cannot be used if the reference <acl> is a file also used with a map.
14467 In this case, you must use the command "add map" in place of "add acl".
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010014468
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010014469add map <map> <key> <value>
14470 Add an entry into the map <map> to associate the value <value> to the key
14471 <key>. This command does not verify if the entry already exists. It is
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010014472 mainly used to fill a map after a clear operation. Note that if the reference
14473 <map> is a file and is shared with a map, this map will contain also a new
14474 pattern entry.
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010014475
Willy Tarreaud63335a2010-02-26 12:56:52 +010014476clear counters
14477 Clear the max values of the statistics counters in each proxy (frontend &
14478 backend) and in each server. The cumulated counters are not affected. This
14479 can be used to get clean counters after an incident, without having to
14480 restart nor to clear traffic counters. This command is restricted and can
14481 only be issued on sockets configured for levels "operator" or "admin".
14482
14483clear counters all
14484 Clear all statistics counters in each proxy (frontend & backend) and in each
14485 server. This has the same effect as restarting. This command is restricted
14486 and can only be issued on sockets configured for level "admin".
14487
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010014488clear acl <acl>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014489 Remove all entries from the acl <acl>. <acl> is the #<id> or the <file>
14490 returned by "show acl". Note that if the reference <acl> is a file and is
14491 shared with a map, this map will be also cleared.
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010014492
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010014493clear map <map>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014494 Remove all entries from the map <map>. <map> is the #<id> or the <file>
14495 returned by "show map". Note that if the reference <map> is a file and is
14496 shared with a acl, this acl will be also cleared.
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010014497
Simon Hormanc88b8872011-06-15 15:18:49 +090014498clear table <table> [ data.<type> <operator> <value> ] | [ key <key> ]
14499 Remove entries from the stick-table <table>.
14500
14501 This is typically used to unblock some users complaining they have been
14502 abusively denied access to a service, but this can also be used to clear some
14503 stickiness entries matching a server that is going to be replaced (see "show
14504 table" below for details). Note that sometimes, removal of an entry will be
14505 refused because it is currently tracked by a session. Retrying a few seconds
14506 later after the session ends is usual enough.
14507
14508 In the case where no options arguments are given all entries will be removed.
14509
14510 When the "data." form is used entries matching a filter applied using the
14511 stored data (see "stick-table" in section 4.2) are removed. A stored data
14512 type must be specified in <type>, and this data type must be stored in the
14513 table otherwise an error is reported. The data is compared according to
14514 <operator> with the 64-bit integer <value>. Operators are the same as with
14515 the ACLs :
14516
14517 - eq : match entries whose data is equal to this value
14518 - ne : match entries whose data is not equal to this value
14519 - le : match entries whose data is less than or equal to this value
14520 - ge : match entries whose data is greater than or equal to this value
14521 - lt : match entries whose data is less than this value
14522 - gt : match entries whose data is greater than this value
14523
14524 When the key form is used the entry <key> is removed. The key must be of the
Simon Horman619e3cc2011-06-15 15:18:52 +090014525 same type as the table, which currently is limited to IPv4, IPv6, integer and
14526 string.
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014527
14528 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020014529 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Emeric Brun7c6b82e2010-09-24 16:34:28 +020014530 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020014531 >>> 0x80e6a4c: key=127.0.0.1 use=0 exp=3594729 gpc0=0 conn_rate(30000)=1 \
14532 bytes_out_rate(60000)=187
14533 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
14534 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014535
14536 $ echo "clear table http_proxy key 127.0.0.1" | socat stdio /tmp/sock1
14537
14538 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Emeric Brun7c6b82e2010-09-24 16:34:28 +020014539 >>> # table: http_proxy, type: ip, size:204800, used:1
Willy Tarreau62a36c42010-08-17 15:53:10 +020014540 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
14541 bytes_out_rate(60000)=191
Simon Hormanc88b8872011-06-15 15:18:49 +090014542 $ echo "clear table http_proxy data.gpc0 eq 1" | socat stdio /tmp/sock1
14543 $ echo "show table http_proxy" | socat stdio /tmp/sock1
14544 >>> # table: http_proxy, type: ip, size:204800, used:1
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014545
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010014546del acl <acl> [<key>|#<ref>]
14547 Delete all the acl entries from the acl <acl> corresponding to the key <key>.
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014548 <acl> is the #<id> or the <file> returned by "show acl". If the <ref> is used,
14549 this command delete only the listed reference. The reference can be found with
14550 listing the content of the acl. Note that if the reference <acl> is a file and
14551 is shared with a map, the entry will be also deleted in the map.
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010014552
14553del map <map> [<key>|#<ref>]
Thierry FOURNIER0b90f312014-01-29 20:40:18 +010014554 Delete all the map entries from the map <map> corresponding to the key <key>.
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014555 <map> is the #<id> or the <file> returned by "show map". If the <ref> is used,
14556 this command delete only the listed reference. The reference can be found with
14557 listing the content of the map. Note that if the reference <map> is a file and
14558 is shared with a acl, the entry will be also deleted in the map.
Thierry FOURNIER0b90f312014-01-29 20:40:18 +010014559
14560disable agent <backend>/<server>
Simon Horman671b6f02013-11-25 10:46:39 +090014561 Mark the auxiliary agent check as temporarily stopped.
14562
14563 In the case where an agent check is being run as a auxiliary check, due
14564 to the agent-check parameter of a server directive, new checks are only
14565 initialised when the agent is in the enabled. Thus, disable agent will
14566 prevent any new agent checks from begin initiated until the agent
14567 re-enabled using enable agent.
14568
14569 When an agent is disabled the processing of an auxiliary agent check that
14570 was initiated while the agent was set as enabled is as follows: All
14571 results that would alter the weight, specifically "drain" or a weight
14572 returned by the agent, are ignored. The processing of agent check is
14573 otherwise unchanged.
14574
14575 The motivation for this feature is to allow the weight changing effects
14576 of the agent checks to be paused to allow the weight of a server to be
14577 configured using set weight without being overridden by the agent.
14578
14579 This command is restricted and can only be issued on sockets configured for
14580 level "admin".
14581
Willy Tarreau532a4502011-09-07 22:37:44 +020014582disable frontend <frontend>
14583 Mark the frontend as temporarily stopped. This corresponds to the mode which
14584 is used during a soft restart : the frontend releases the port but can be
14585 enabled again if needed. This should be used with care as some non-Linux OSes
14586 are unable to enable it back. This is intended to be used in environments
14587 where stopping a proxy is not even imaginable but a misconfigured proxy must
14588 be fixed. That way it's possible to release the port and bind it into another
14589 process to restore operations. The frontend will appear with status "STOP"
14590 on the stats page.
14591
14592 The frontend may be specified either by its name or by its numeric ID,
14593 prefixed with a sharp ('#').
14594
14595 This command is restricted and can only be issued on sockets configured for
14596 level "admin".
14597
Willy Tarreau9b5aecd2014-05-23 11:53:10 +020014598disable health <backend>/<server>
14599 Mark the primary health check as temporarily stopped. This will disable
14600 sending of health checks, and the last health check result will be ignored.
14601 The server will be in unchecked state and considered UP unless an auxiliary
14602 agent check forces it down.
14603
14604 This command is restricted and can only be issued on sockets configured for
14605 level "admin".
14606
Willy Tarreaud63335a2010-02-26 12:56:52 +010014607disable server <backend>/<server>
14608 Mark the server DOWN for maintenance. In this mode, no more checks will be
14609 performed on the server until it leaves maintenance.
14610 If the server is tracked by other servers, those servers will be set to DOWN
14611 during the maintenance.
14612
14613 In the statistics page, a server DOWN for maintenance will appear with a
14614 "MAINT" status, its tracking servers with the "MAINT(via)" one.
14615
14616 Both the backend and the server may be specified either by their name or by
Willy Tarreauf5f31922011-08-02 11:32:07 +020014617 their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010014618
14619 This command is restricted and can only be issued on sockets configured for
14620 level "admin".
14621
Simon Horman671b6f02013-11-25 10:46:39 +090014622enable agent <backend>/<server>
14623 Resume auxiliary agent check that was temporarily stopped.
14624
14625 See "disable agent" for details of the effect of temporarily starting
14626 and stopping an auxiliary agent.
14627
14628 This command is restricted and can only be issued on sockets configured for
14629 level "admin".
14630
Willy Tarreau532a4502011-09-07 22:37:44 +020014631enable frontend <frontend>
14632 Resume a frontend which was temporarily stopped. It is possible that some of
14633 the listening ports won't be able to bind anymore (eg: if another process
14634 took them since the 'disable frontend' operation). If this happens, an error
14635 is displayed. Some operating systems might not be able to resume a frontend
14636 which was disabled.
14637
14638 The frontend may be specified either by its name or by its numeric ID,
14639 prefixed with a sharp ('#').
14640
14641 This command is restricted and can only be issued on sockets configured for
14642 level "admin".
14643
Willy Tarreau9b5aecd2014-05-23 11:53:10 +020014644enable health <backend>/<server>
14645 Resume a primary health check that was temporarily stopped. This will enable
14646 sending of health checks again. Please see "disable health" for details.
14647
14648 This command is restricted and can only be issued on sockets configured for
14649 level "admin".
14650
Willy Tarreaud63335a2010-02-26 12:56:52 +010014651enable server <backend>/<server>
14652 If the server was previously marked as DOWN for maintenance, this marks the
14653 server UP and checks are re-enabled.
14654
14655 Both the backend and the server may be specified either by their name or by
Willy Tarreauf5f31922011-08-02 11:32:07 +020014656 their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010014657
14658 This command is restricted and can only be issued on sockets configured for
14659 level "admin".
14660
Thierry FOURNIER5b16df72014-02-26 18:07:38 +010014661get map <map> <value>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014662get acl <acl> <value>
14663 Lookup the value <value> in the map <map> or in the ACL <acl>. <map> or <acl>
14664 are the #<id> or the <file> returned by "show map" or "show acl". This command
14665 returns all the matching patterns associated with this map. This is useful for
14666 debugging maps and ACLs. The output format is composed by one line par
14667 matching type. Each line is composed by space-delimited series of words.
Thierry FOURNIER5b16df72014-02-26 18:07:38 +010014668
14669 The first two words are:
14670
14671 <match method>: The match method applied. It can be "found", "bool",
14672 "int", "ip", "bin", "len", "str", "beg", "sub", "dir",
14673 "dom", "end" or "reg".
14674
14675 <match result>: The result. Can be "match" or "no-match".
14676
14677 The following words are returned only if the pattern matches an entry.
14678
14679 <index type>: "tree" or "list". The internal lookup algorithm.
14680
14681 <case>: "case-insensitive" or "case-sensitive". The
14682 interpretation of the case.
14683
14684 <entry matched>: match="<entry>". Return the matched pattern. It is
14685 useful with regular expressions.
14686
14687 The two last word are used to show the returned value and its type. With the
14688 "acl" case, the pattern doesn't exist.
14689
14690 return=nothing: No return because there are no "map".
14691 return="<value>": The value returned in the string format.
14692 return=cannot-display: The value cannot be converted as string.
14693
14694 type="<type>": The type of the returned sample.
14695
Willy Tarreaud63335a2010-02-26 12:56:52 +010014696get weight <backend>/<server>
14697 Report the current weight and the initial weight of server <server> in
14698 backend <backend> or an error if either doesn't exist. The initial weight is
14699 the one that appears in the configuration file. Both are normally equal
14700 unless the current weight has been changed. Both the backend and the server
14701 may be specified either by their name or by their numeric ID, prefixed with a
Willy Tarreauf5f31922011-08-02 11:32:07 +020014702 sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010014703
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014704help
14705 Print the list of known keywords and their basic usage. The same help screen
14706 is also displayed for unknown commands.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010014707
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014708prompt
14709 Toggle the prompt at the beginning of the line and enter or leave interactive
14710 mode. In interactive mode, the connection is not closed after a command
14711 completes. Instead, the prompt will appear again, indicating the user that
14712 the interpreter is waiting for a new command. The prompt consists in a right
14713 angle bracket followed by a space "> ". This mode is particularly convenient
14714 when one wants to periodically check information such as stats or errors.
14715 It is also a good idea to enter interactive mode before issuing a "help"
14716 command.
14717
14718quit
14719 Close the connection when in interactive mode.
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010014720
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010014721set map <map> [<key>|#<ref>] <value>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014722 Modify the value corresponding to each key <key> in a map <map>. <map> is the
14723 #<id> or <file> returned by "show map". If the <ref> is used in place of
14724 <key>, only the entry pointed by <ref> is changed. The new value is <value>.
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010014725
Willy Tarreau2a0f4d22011-08-02 11:49:05 +020014726set maxconn frontend <frontend> <value>
Willy Tarreau3c7a79d2012-09-26 21:07:15 +020014727 Dynamically change the specified frontend's maxconn setting. Any positive
14728 value is allowed including zero, but setting values larger than the global
14729 maxconn does not make much sense. If the limit is increased and connections
14730 were pending, they will immediately be accepted. If it is lowered to a value
14731 below the current number of connections, new connections acceptation will be
Willy Tarreau2a0f4d22011-08-02 11:49:05 +020014732 delayed until the threshold is reached. The frontend might be specified by
14733 either its name or its numeric ID prefixed with a sharp ('#').
14734
Willy Tarreau91886b62011-09-07 14:38:31 +020014735set maxconn global <maxconn>
14736 Dynamically change the global maxconn setting within the range defined by the
14737 initial global maxconn setting. If it is increased and connections were
14738 pending, they will immediately be accepted. If it is lowered to a value below
14739 the current number of connections, new connections acceptation will be
14740 delayed until the threshold is reached. A value of zero restores the initial
14741 setting.
14742
Willy Tarreauf5b22872011-09-07 16:13:44 +020014743set rate-limit connections global <value>
14744 Change the process-wide connection rate limit, which is set by the global
14745 'maxconnrate' setting. A value of zero disables the limitation. This limit
14746 applies to all frontends and the change has an immediate effect. The value
14747 is passed in number of connections per second.
14748
William Lallemandd85f9172012-11-09 17:05:39 +010014749set rate-limit http-compression global <value>
14750 Change the maximum input compression rate, which is set by the global
14751 'maxcomprate' setting. A value of zero disables the limitation. The value is
William Lallemand096f5542012-11-19 17:26:05 +010014752 passed in number of kilobytes per second. The value is available in the "show
14753 info" on the line "CompressBpsRateLim" in bytes.
William Lallemandd85f9172012-11-09 17:05:39 +010014754
Willy Tarreau93e7c002013-10-07 18:51:07 +020014755set rate-limit sessions global <value>
14756 Change the process-wide session rate limit, which is set by the global
14757 'maxsessrate' setting. A value of zero disables the limitation. This limit
14758 applies to all frontends and the change has an immediate effect. The value
14759 is passed in number of sessions per second.
14760
Willy Tarreaue43d5322013-10-07 20:01:52 +020014761set rate-limit ssl-sessions global <value>
14762 Change the process-wide SSL session rate limit, which is set by the global
14763 'maxsslrate' setting. A value of zero disables the limitation. This limit
14764 applies to all frontends and the change has an immediate effect. The value
14765 is passed in number of sessions per second sent to the SSL stack. It applies
14766 before the handshake in order to protect the stack against handshake abuses.
14767
Willy Tarreau2a4b70f2014-05-22 18:42:35 +020014768set server <backend>/<server> agent [ up | down ]
14769 Force a server's agent to a new state. This can be useful to immediately
14770 switch a server's state regardless of some slow agent checks for example.
14771 Note that the change is propagated to tracking servers if any.
14772
14773set server <backend>/<server> health [ up | stopping | down ]
14774 Force a server's health to a new state. This can be useful to immediately
14775 switch a server's state regardless of some slow health checks for example.
14776 Note that the change is propagated to tracking servers if any.
14777
14778set server <backend>/<server> state [ ready | drain | maint ]
14779 Force a server's administrative state to a new state. This can be useful to
14780 disable load balancing and/or any traffic to a server. Setting the state to
14781 "ready" puts the server in normal mode, and the command is the equivalent of
14782 the "enable server" command. Setting the state to "maint" disables any traffic
14783 to the server as well as any health checks. This is the equivalent of the
14784 "disable server" command. Setting the mode to "drain" only removes the server
14785 from load balancing but still allows it to be checked and to accept new
14786 persistent connections. Changes are propagated to tracking servers if any.
14787
14788set server <backend>/<server> weight <weight>[%]
14789 Change a server's weight to the value passed in argument. This is the exact
14790 equivalent of the "set weight" command below.
14791
Emeric Brun4147b2e2014-06-16 18:36:30 +020014792set ssl ocsp-response <response>
14793 This command is used to update an OCSP Response for a certificate (see "crt"
14794 on "bind" lines). Same controls are performed as during the initial loading of
14795 the response. The <response> must be passed as a base64 encoded string of the
14796 DER encoded response from the OCSP server.
14797
14798 Example:
14799 openssl ocsp -issuer issuer.pem -cert server.pem \
14800 -host ocsp.issuer.com:80 -respout resp.der
14801 echo "set ssl ocsp-response $(base64 -w 10000 resp.der)" | \
14802 socat stdio /var/run/haproxy.stat
14803
Nenad Merdanovicc6985f02015-05-09 08:46:02 +020014804set ssl tls-key <id> <tlskey>
14805 Set the next TLS key for the <id> listener to <tlskey>. This key becomes the
14806 ultimate key, while the penultimate one is used for encryption (others just
14807 decrypt). The oldest TLS key present is overwritten. <id> is either a numeric
14808 #<id> or <file> returned by "show tls-keys". <tlskey> is a base64 encoded 48
14809 bit TLS ticket key (ex. openssl rand -base64 48).
14810
Willy Tarreau47060b62013-08-01 21:11:42 +020014811set table <table> key <key> [data.<data_type> <value>]*
Willy Tarreau654694e2012-06-07 01:03:16 +020014812 Create or update a stick-table entry in the table. If the key is not present,
14813 an entry is inserted. See stick-table in section 4.2 to find all possible
14814 values for <data_type>. The most likely use consists in dynamically entering
14815 entries for source IP addresses, with a flag in gpc0 to dynamically block an
Willy Tarreau47060b62013-08-01 21:11:42 +020014816 IP address or affect its quality of service. It is possible to pass multiple
14817 data_types in a single call.
Willy Tarreau654694e2012-06-07 01:03:16 +020014818
Willy Tarreaud63335a2010-02-26 12:56:52 +010014819set timeout cli <delay>
14820 Change the CLI interface timeout for current connection. This can be useful
14821 during long debugging sessions where the user needs to constantly inspect
14822 some indicators without being disconnected. The delay is passed in seconds.
14823
14824set weight <backend>/<server> <weight>[%]
14825 Change a server's weight to the value passed in argument. If the value ends
14826 with the '%' sign, then the new weight will be relative to the initially
Simon Horman58b5d292013-02-12 10:45:52 +090014827 configured weight. Absolute weights are permitted between 0 and 256.
14828 Relative weights must be positive with the resulting absolute weight is
14829 capped at 256. Servers which are part of a farm running a static
14830 load-balancing algorithm have stricter limitations because the weight
14831 cannot change once set. Thus for these servers, the only accepted values
14832 are 0 and 100% (or 0 and the initial weight). Changes take effect
14833 immediately, though certain LB algorithms require a certain amount of
14834 requests to consider changes. A typical usage of this command is to
14835 disable a server during an update by setting its weight to zero, then to
14836 enable it again after the update by setting it back to 100%. This command
14837 is restricted and can only be issued on sockets configured for level
14838 "admin". Both the backend and the server may be specified either by their
14839 name or by their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010014840
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010014841show errors [<iid>]
14842 Dump last known request and response errors collected by frontends and
14843 backends. If <iid> is specified, the limit the dump to errors concerning
Willy Tarreau6162db22009-10-10 17:13:00 +020014844 either frontend or backend whose ID is <iid>. This command is restricted
14845 and can only be issued on sockets configured for levels "operator" or
14846 "admin".
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010014847
14848 The errors which may be collected are the last request and response errors
14849 caused by protocol violations, often due to invalid characters in header
14850 names. The report precisely indicates what exact character violated the
14851 protocol. Other important information such as the exact date the error was
14852 detected, frontend and backend names, the server name (when known), the
14853 internal session ID and the source address which has initiated the session
14854 are reported too.
14855
14856 All characters are returned, and non-printable characters are encoded. The
14857 most common ones (\t = 9, \n = 10, \r = 13 and \e = 27) are encoded as one
14858 letter following a backslash. The backslash itself is encoded as '\\' to
14859 avoid confusion. Other non-printable characters are encoded '\xNN' where
14860 NN is the two-digits hexadecimal representation of the character's ASCII
14861 code.
14862
14863 Lines are prefixed with the position of their first character, starting at 0
14864 for the beginning of the buffer. At most one input line is printed per line,
14865 and large lines will be broken into multiple consecutive output lines so that
14866 the output never goes beyond 79 characters wide. It is easy to detect if a
14867 line was broken, because it will not end with '\n' and the next line's offset
14868 will be followed by a '+' sign, indicating it is a continuation of previous
14869 line.
14870
14871 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020014872 $ echo "show errors" | socat stdio /tmp/sock1
14873 >>> [04/Mar/2009:15:46:56.081] backend http-in (#2) : invalid response
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010014874 src 127.0.0.1, session #54, frontend fe-eth0 (#1), server s2 (#1)
14875 response length 213 bytes, error at position 23:
14876
14877 00000 HTTP/1.0 200 OK\r\n
14878 00017 header/bizarre:blah\r\n
14879 00038 Location: blah\r\n
14880 00054 Long-line: this is a very long line which should b
14881 00104+ e broken into multiple lines on the output buffer,
14882 00154+ otherwise it would be too large to print in a ter
14883 00204+ minal\r\n
14884 00211 \r\n
14885
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014886 In the example above, we see that the backend "http-in" which has internal
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010014887 ID 2 has blocked an invalid response from its server s2 which has internal
14888 ID 1. The request was on session 54 initiated by source 127.0.0.1 and
14889 received by frontend fe-eth0 whose ID is 1. The total response length was
14890 213 bytes when the error was detected, and the error was at byte 23. This
14891 is the slash ('/') in header name "header/bizarre", which is not a valid
14892 HTTP character for a header name.
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010014893
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014894show info
14895 Dump info about haproxy status on current process.
14896
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010014897show map [<map>]
14898 Dump info about map converters. Without argument, the list of all available
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014899 maps is returned. If a <map> is specified, its contents are dumped. <map> is
14900 the #<id> or <file>. The first column is a unique identifier. It can be used
14901 as reference for the operation "del map" and "set map". The second column is
14902 the pattern and the third column is the sample if available. The data returned
14903 are not directly a list of available maps, but are the list of all patterns
14904 composing any map. Many of these patterns can be shared with ACL.
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010014905
14906show acl [<acl>]
14907 Dump info about acl converters. Without argument, the list of all available
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014908 acls is returned. If a <acl> is specified, its contents are dumped. <acl> if
14909 the #<id> or <file>. The dump format is the same than the map even for the
14910 sample value. The data returned are not a list of available ACL, but are the
14911 list of all patterns composing any ACL. Many of these patterns can be shared
14912 with maps.
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010014913
Willy Tarreau12833bb2014-01-28 16:49:56 +010014914show pools
14915 Dump the status of internal memory pools. This is useful to track memory
14916 usage when suspecting a memory leak for example. It does exactly the same
14917 as the SIGQUIT when running in foreground except that it does not flush
14918 the pools.
14919
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014920show sess
14921 Dump all known sessions. Avoid doing this on slow connections as this can
Willy Tarreau6162db22009-10-10 17:13:00 +020014922 be huge. This command is restricted and can only be issued on sockets
14923 configured for levels "operator" or "admin".
14924
Willy Tarreau66dc20a2010-03-05 17:53:32 +010014925show sess <id>
14926 Display a lot of internal information about the specified session identifier.
14927 This identifier is the first field at the beginning of the lines in the dumps
14928 of "show sess" (it corresponds to the session pointer). Those information are
14929 useless to most users but may be used by haproxy developers to troubleshoot a
14930 complex bug. The output format is intentionally not documented so that it can
Olivierce31e6e2014-09-05 18:49:10 +020014931 freely evolve depending on demands. You may find a description of all fields
14932 returned in src/dumpstats.c
14933
14934 The special id "all" dumps the states of all sessions, which must be avoided
14935 as much as possible as it is highly CPU intensive and can take a lot of time.
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014936
14937show stat [<iid> <type> <sid>]
14938 Dump statistics in the CSV format. By passing <id>, <type> and <sid>, it is
14939 possible to dump only selected items :
14940 - <iid> is a proxy ID, -1 to dump everything
14941 - <type> selects the type of dumpable objects : 1 for frontends, 2 for
14942 backends, 4 for servers, -1 for everything. These values can be ORed,
14943 for example:
14944 1 + 2 = 3 -> frontend + backend.
14945 1 + 2 + 4 = 7 -> frontend + backend + server.
14946 - <sid> is a server ID, -1 to dump everything from the selected proxy.
14947
14948 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020014949 $ echo "show info;show stat" | socat stdio unix-connect:/tmp/sock1
14950 >>> Name: HAProxy
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014951 Version: 1.4-dev2-49
14952 Release_date: 2009/09/23
14953 Nbproc: 1
14954 Process_num: 1
14955 (...)
14956
14957 # pxname,svname,qcur,qmax,scur,smax,slim,stot,bin,bout,dreq, (...)
14958 stats,FRONTEND,,,0,0,1000,0,0,0,0,0,0,,,,,OPEN,,,,,,,,,1,1,0, (...)
14959 stats,BACKEND,0,0,0,0,1000,0,0,0,0,0,,0,0,0,0,UP,0,0,0,,0,250,(...)
14960 (...)
14961 www1,BACKEND,0,0,0,0,1000,0,0,0,0,0,,0,0,0,0,UP,1,1,0,,0,250, (...)
14962
14963 $
14964
14965 Here, two commands have been issued at once. That way it's easy to find
14966 which process the stats apply to in multi-process mode. Notice the empty
14967 line after the information output which marks the end of the first block.
14968 A similar empty line appears at the end of the second block (stats) so that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010014969 the reader knows the output has not been truncated.
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014970
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014971show table
14972 Dump general information on all known stick-tables. Their name is returned
14973 (the name of the proxy which holds them), their type (currently zero, always
14974 IP), their size in maximum possible number of entries, and the number of
14975 entries currently in use.
14976
14977 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020014978 $ echo "show table" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090014979 >>> # table: front_pub, type: ip, size:204800, used:171454
14980 >>> # table: back_rdp, type: ip, size:204800, used:0
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014981
Simon Horman17bce342011-06-15 15:18:47 +090014982show table <name> [ data.<type> <operator> <value> ] | [ key <key> ]
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014983 Dump contents of stick-table <name>. In this mode, a first line of generic
14984 information about the table is reported as with "show table", then all
14985 entries are dumped. Since this can be quite heavy, it is possible to specify
Simon Horman17bce342011-06-15 15:18:47 +090014986 a filter in order to specify what entries to display.
14987
14988 When the "data." form is used the filter applies to the stored data (see
14989 "stick-table" in section 4.2). A stored data type must be specified
14990 in <type>, and this data type must be stored in the table otherwise an
14991 error is reported. The data is compared according to <operator> with the
14992 64-bit integer <value>. Operators are the same as with the ACLs :
14993
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014994 - eq : match entries whose data is equal to this value
14995 - ne : match entries whose data is not equal to this value
14996 - le : match entries whose data is less than or equal to this value
14997 - ge : match entries whose data is greater than or equal to this value
14998 - lt : match entries whose data is less than this value
14999 - gt : match entries whose data is greater than this value
15000
Simon Hormanc88b8872011-06-15 15:18:49 +090015001
15002 When the key form is used the entry <key> is shown. The key must be of the
Simon Horman619e3cc2011-06-15 15:18:52 +090015003 same type as the table, which currently is limited to IPv4, IPv6, integer,
15004 and string.
Simon Horman17bce342011-06-15 15:18:47 +090015005
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020015006 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020015007 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090015008 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020015009 >>> 0x80e6a4c: key=127.0.0.1 use=0 exp=3594729 gpc0=0 conn_rate(30000)=1 \
15010 bytes_out_rate(60000)=187
15011 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
15012 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020015013
Willy Tarreau62a36c42010-08-17 15:53:10 +020015014 $ echo "show table http_proxy data.gpc0 gt 0" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090015015 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020015016 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
15017 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020015018
Willy Tarreau62a36c42010-08-17 15:53:10 +020015019 $ echo "show table http_proxy data.conn_rate gt 5" | \
15020 socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090015021 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020015022 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
15023 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020015024
Simon Horman17bce342011-06-15 15:18:47 +090015025 $ echo "show table http_proxy key 127.0.0.2" | \
15026 socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090015027 >>> # table: http_proxy, type: ip, size:204800, used:2
Simon Horman17bce342011-06-15 15:18:47 +090015028 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
15029 bytes_out_rate(60000)=191
15030
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020015031 When the data criterion applies to a dynamic value dependent on time such as
15032 a bytes rate, the value is dynamically computed during the evaluation of the
15033 entry in order to decide whether it has to be dumped or not. This means that
15034 such a filter could match for some time then not match anymore because as
15035 time goes, the average event rate drops.
15036
15037 It is possible to use this to extract lists of IP addresses abusing the
15038 service, in order to monitor them or even blacklist them in a firewall.
15039 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020015040 $ echo "show table http_proxy data.gpc0 gt 0" \
15041 | socat stdio /tmp/sock1 \
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020015042 | fgrep 'key=' | cut -d' ' -f2 | cut -d= -f2 > abusers-ip.txt
15043 ( or | awk '/key/{ print a[split($2,a,"=")]; }' )
Krzysztof Piotr Oledzki719e7262009-10-04 15:02:46 +020015044
Nenad Merdanovicc6985f02015-05-09 08:46:02 +020015045show tls-keys
15046 Dump all loaded TLS ticket keys. The TLS ticket key reference ID and the
15047 file from which the keys have been loaded is shown. Both of those can be
15048 used to update the TLS keys using "set ssl tls-key".
15049
Willy Tarreau532a4502011-09-07 22:37:44 +020015050shutdown frontend <frontend>
15051 Completely delete the specified frontend. All the ports it was bound to will
15052 be released. It will not be possible to enable the frontend anymore after
15053 this operation. This is intended to be used in environments where stopping a
15054 proxy is not even imaginable but a misconfigured proxy must be fixed. That
15055 way it's possible to release the port and bind it into another process to
15056 restore operations. The frontend will not appear at all on the stats page
15057 once it is terminated.
15058
15059 The frontend may be specified either by its name or by its numeric ID,
15060 prefixed with a sharp ('#').
15061
15062 This command is restricted and can only be issued on sockets configured for
15063 level "admin".
15064
Willy Tarreaua295edc2011-09-07 23:21:03 +020015065shutdown session <id>
15066 Immediately terminate the session matching the specified session identifier.
15067 This identifier is the first field at the beginning of the lines in the dumps
15068 of "show sess" (it corresponds to the session pointer). This can be used to
15069 terminate a long-running session without waiting for a timeout or when an
15070 endless transfer is ongoing. Such terminated sessions are reported with a 'K'
15071 flag in the logs.
15072
Cyril Bontée63a1eb2014-07-12 18:22:42 +020015073shutdown sessions server <backend>/<server>
Willy Tarreau52b2d222011-09-07 23:48:48 +020015074 Immediately terminate all the sessions attached to the specified server. This
15075 can be used to terminate long-running sessions after a server is put into
15076 maintenance mode, for instance. Such terminated sessions are reported with a
15077 'K' flag in the logs.
15078
Willy Tarreau0ba27502007-12-24 16:55:16 +010015079/*
15080 * Local variables:
15081 * fill-column: 79
15082 * End:
15083 */