blob: b384b7b4aa9d9cbc57ad08b5d71ef687d682f136 [file] [log] [blame]
Willy Tarreau6a06a402007-07-15 20:15:28 +02001 ----------------------
Willy Tarreau8317b282014-04-23 01:49:41 +02002 HAProxy
Willy Tarreau6a06a402007-07-15 20:15:28 +02003 Configuration Manual
4 ----------------------
Willy Tarreau991b4782015-10-13 21:48:10 +02005 version 1.7
Willy Tarreau6a06a402007-07-15 20:15:28 +02006 willy tarreau
Willy Tarreaucb928252015-12-20 23:33:18 +01007 2015/12/20
Willy Tarreau6a06a402007-07-15 20:15:28 +02008
9
10This document covers the configuration language as implemented in the version
11specified above. It does not provide any hint, example or advice. For such
Willy Tarreau0ba27502007-12-24 16:55:16 +010012documentation, please refer to the Reference Manual or the Architecture Manual.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013The summary below is meant to help you search sections by name and navigate
14through the document.
Willy Tarreau6a06a402007-07-15 20:15:28 +020015
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016Note to documentation contributors :
Jamie Gloudonaaa21002012-08-25 00:18:33 -040017 This document is formatted with 80 columns per line, with even number of
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018 spaces for indentation and without tabs. Please follow these rules strictly
19 so that it remains easily printable everywhere. If a line needs to be
20 printed verbatim and does not fit, please end each line with a backslash
Willy Tarreau62a36c42010-08-17 15:53:10 +020021 ('\') and continue on next line, indented by two characters. It is also
22 sometimes useful to prefix all output lines (logs, console outs) with 3
23 closing angle brackets ('>>>') in order to help get the difference between
24 inputs and outputs when it can become ambiguous. If you add sections,
25 please update the summary below for easier searching.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020026
27
28Summary
29-------
30
311. Quick reminder about HTTP
321.1. The HTTP transaction model
331.2. HTTP request
341.2.1. The Request line
351.2.2. The request headers
361.3. HTTP response
371.3.1. The Response line
381.3.2. The response headers
39
402. Configuring HAProxy
412.1. Configuration file format
William Lallemandf9873ba2015-05-05 17:37:14 +0200422.2. Quoting and escaping
William Lallemandb2f07452015-05-12 14:27:13 +0200432.3. Environment variables
442.4. Time format
452.5. Examples
Willy Tarreauc57f0e22009-05-10 13:12:33 +020046
473. Global parameters
483.1. Process management and security
493.2. Performance tuning
503.3. Debugging
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100513.4. Userlists
Cyril Bontédc4d9032012-04-08 21:57:39 +0200523.5. Peers
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200533.6. Mailers
Willy Tarreauc57f0e22009-05-10 13:12:33 +020054
554. Proxies
564.1. Proxy keywords matrix
574.2. Alphabetically sorted keywords reference
58
Willy Tarreau086fbf52012-09-24 20:34:51 +0200595. Bind and Server options
605.1. Bind options
615.2. Server and default-server options
Baptiste Assmann1fa66662015-04-14 00:28:47 +0200625.3. Server DNS resolution
635.3.1. Global overview
645.3.2. The resolvers section
Willy Tarreauc57f0e22009-05-10 13:12:33 +020065
666. HTTP header manipulation
67
Willy Tarreau74ca5042013-06-11 23:12:07 +0200687. Using ACLs and fetching samples
697.1. ACL basics
707.1.1. Matching booleans
717.1.2. Matching integers
727.1.3. Matching strings
737.1.4. Matching regular expressions (regexes)
747.1.5. Matching arbitrary data blocks
757.1.6. Matching IPv4 and IPv6 addresses
767.2. Using ACLs to form conditions
777.3. Fetching samples
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200787.3.1. Converters
797.3.2. Fetching samples from internal states
807.3.3. Fetching samples at Layer 4
817.3.4. Fetching samples at Layer 5
827.3.5. Fetching samples from buffer contents (Layer 6)
837.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +0200847.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020085
868. Logging
878.1. Log levels
888.2. Log formats
898.2.1. Default log format
908.2.2. TCP log format
918.2.3. HTTP log format
William Lallemand48940402012-01-30 16:47:22 +0100928.2.4. Custom log format
Willy Tarreau5f51e1a2012-12-03 18:40:10 +0100938.2.5. Error log format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200948.3. Advanced logging options
958.3.1. Disabling logging of external tests
968.3.2. Logging before waiting for the session to terminate
978.3.3. Raising log level upon errors
988.3.4. Disabling logging of successful connections
998.4. Timing events
1008.5. Session state at disconnection
1018.6. Non-printable characters
1028.7. Capturing HTTP cookies
1038.8. Capturing HTTP headers
1048.9. Examples of logs
105
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200106
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
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200524 - cpu-map
Willy Tarreau6a06a402007-07-15 20:15:28 +0200525 - daemon
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200526 - description
527 - deviceatlas-json-file
528 - deviceatlas-log-level
529 - deviceatlas-separator
530 - deviceatlas-properties-cookie
Simon Horman98637e52014-06-20 12:30:16 +0900531 - external-check
Willy Tarreau6a06a402007-07-15 20:15:28 +0200532 - gid
533 - group
534 - log
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200535 - log-tag
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100536 - log-send-hostname
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200537 - lua-load
Willy Tarreau6a06a402007-07-15 20:15:28 +0200538 - nbproc
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200539 - node
Willy Tarreau6a06a402007-07-15 20:15:28 +0200540 - pidfile
541 - uid
542 - ulimit-n
543 - user
Willy Tarreaufbee7132007-10-18 13:53:22 +0200544 - stats
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200545 - ssl-default-bind-ciphers
546 - ssl-default-bind-options
547 - ssl-default-server-ciphers
548 - ssl-default-server-options
549 - ssl-dh-param-file
Emeric Brun850efd52014-01-29 12:24:34 +0100550 - ssl-server-verify
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100551 - unix-bind
Thomas Holmesdb04f192015-05-18 13:21:39 +0100552 - 51degrees-data-file
553 - 51degrees-property-name-list
Dragan Dosen93b38d92015-06-29 16:43:25 +0200554 - 51degrees-property-separator
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200555 - 51degrees-cache-size
Willy Tarreaud72758d2010-01-12 10:42:19 +0100556
Willy Tarreau6a06a402007-07-15 20:15:28 +0200557 * Performance tuning
Willy Tarreau1746eec2014-04-25 10:46:47 +0200558 - max-spread-checks
Willy Tarreau6a06a402007-07-15 20:15:28 +0200559 - maxconn
Willy Tarreau81c25d02011-09-07 15:17:21 +0200560 - maxconnrate
William Lallemandd85f9172012-11-09 17:05:39 +0100561 - maxcomprate
William Lallemand072a2bf2012-11-20 17:01:01 +0100562 - maxcompcpuusage
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100563 - maxpipes
Willy Tarreau93e7c002013-10-07 18:51:07 +0200564 - maxsessrate
Willy Tarreau403edff2012-09-06 11:58:37 +0200565 - maxsslconn
Willy Tarreaue43d5322013-10-07 20:01:52 +0200566 - maxsslrate
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200567 - maxzlibmem
Willy Tarreau6a06a402007-07-15 20:15:28 +0200568 - noepoll
569 - nokqueue
570 - nopoll
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100571 - nosplice
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300572 - nogetaddrinfo
Willy Tarreaufe255b72007-10-14 23:09:26 +0200573 - spread-checks
Baptiste Assmann5626f482015-08-23 10:00:10 +0200574 - server-state-base
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200575 - server-state-file
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200576 - tune.buffers.limit
577 - tune.buffers.reserve
Willy Tarreau27a674e2009-08-17 07:23:33 +0200578 - tune.bufsize
Willy Tarreau43961d52010-10-04 20:39:20 +0200579 - tune.chksize
William Lallemandf3747832012-11-09 12:33:10 +0100580 - tune.comp.maxlevel
Willy Tarreau193b8c62012-11-22 00:17:38 +0100581 - tune.http.cookielen
Willy Tarreauac1932d2011-10-24 19:14:41 +0200582 - tune.http.maxhdr
Willy Tarreau7e312732014-02-12 16:35:14 +0100583 - tune.idletimer
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100584 - tune.lua.forced-yield
Willy Tarreau32f61e22015-03-18 17:54:59 +0100585 - tune.lua.maxmem
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100586 - tune.lua.session-timeout
587 - tune.lua.task-timeout
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +0200588 - tune.lua.service-timeout
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100589 - tune.maxaccept
590 - tune.maxpollevents
Willy Tarreau27a674e2009-08-17 07:23:33 +0200591 - tune.maxrewrite
Willy Tarreauf3045d22015-04-29 16:24:50 +0200592 - tune.pattern.cache-size
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200593 - tune.pipesize
Willy Tarreaue803de22010-01-21 17:43:04 +0100594 - tune.rcvbuf.client
595 - tune.rcvbuf.server
Willy Tarreaub22fc302015-12-14 12:04:35 +0100596 - tune.recv_enough
Willy Tarreaue803de22010-01-21 17:43:04 +0100597 - tune.sndbuf.client
598 - tune.sndbuf.server
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100599 - tune.ssl.cachesize
Willy Tarreaubfd59462013-02-21 07:46:09 +0100600 - tune.ssl.lifetime
Emeric Brun8dc60392014-05-09 13:52:00 +0200601 - tune.ssl.force-private-cache
Willy Tarreaubfd59462013-02-21 07:46:09 +0100602 - tune.ssl.maxrecord
Remi Gacognef46cd6e2014-06-12 14:58:40 +0200603 - tune.ssl.default-dh-param
Christopher Faulet31af49d2015-06-09 17:29:50 +0200604 - tune.ssl.ssl-ctx-cache-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200605 - tune.vars.global-max-size
606 - tune.vars.reqres-max-size
607 - tune.vars.sess-max-size
608 - tune.vars.txn-max-size
William Lallemanda509e4c2012-11-07 16:54:34 +0100609 - tune.zlib.memlevel
610 - tune.zlib.windowsize
Willy Tarreaud72758d2010-01-12 10:42:19 +0100611
Willy Tarreau6a06a402007-07-15 20:15:28 +0200612 * Debugging
613 - debug
614 - quiet
Willy Tarreau6a06a402007-07-15 20:15:28 +0200615
616
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006173.1. Process management and security
Willy Tarreau6a06a402007-07-15 20:15:28 +0200618------------------------------------
619
Emeric Brunc8e8d122012-10-02 18:42:10 +0200620ca-base <dir>
621 Assigns a default directory to fetch SSL CA certificates and CRLs from when a
Emeric Brunfd33a262012-10-11 16:28:27 +0200622 relative path is used with "ca-file" or "crl-file" directives. Absolute
623 locations specified in "ca-file" and "crl-file" prevail and ignore "ca-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +0200624
Willy Tarreau6a06a402007-07-15 20:15:28 +0200625chroot <jail dir>
626 Changes current directory to <jail dir> and performs a chroot() there before
627 dropping privileges. This increases the security level in case an unknown
628 vulnerability would be exploited, since it would make it very hard for the
629 attacker to exploit the system. This only works when the process is started
630 with superuser privileges. It is important to ensure that <jail_dir> is both
631 empty and unwritable to anyone.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100632
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100633cpu-map <"all"|"odd"|"even"|process_num> <cpu-set>...
634 On Linux 2.6 and above, it is possible to bind a process to a specific CPU
635 set. This means that the process will never run on other CPUs. The "cpu-map"
636 directive specifies CPU sets for process sets. The first argument is the
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100637 process number to bind. This process must have a number between 1 and 32 or
638 64, depending on the machine's word size, and any process IDs above nbproc
639 are ignored. It is possible to specify all processes at once using "all",
640 only odd numbers using "odd" or even numbers using "even", just like with the
641 "bind-process" directive. The second and forthcoming arguments are CPU sets.
642 Each CPU set is either a unique number between 0 and 31 or 63 or a range with
643 two such numbers delimited by a dash ('-'). Multiple CPU numbers or ranges
644 may be specified, and the processes will be allowed to bind to all of them.
645 Obviously, multiple "cpu-map" directives may be specified. Each "cpu-map"
646 directive will replace the previous ones when they overlap.
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100647
Emeric Brunc8e8d122012-10-02 18:42:10 +0200648crt-base <dir>
649 Assigns a default directory to fetch SSL certificates from when a relative
650 path is used with "crtfile" directives. Absolute locations specified after
651 "crtfile" prevail and ignore "crt-base".
652
Willy Tarreau6a06a402007-07-15 20:15:28 +0200653daemon
654 Makes the process fork into background. This is the recommended mode of
655 operation. It is equivalent to the command line "-D" argument. It can be
656 disabled by the command line "-db" argument.
657
David Carlier8167f302015-06-01 13:50:06 +0200658deviceatlas-json-file <path>
659 Sets the path of the DeviceAtlas JSON data file to be loaded by the API.
660 The path must be a valid JSON data file and accessible by Haproxy process.
661
662deviceatlas-log-level <value>
663 Sets the level of informations returned by the API. This directive is
664 optional and set to 0 by default if not set.
665
666deviceatlas-separator <char>
667 Sets the character separator for the API properties results. This directive
668 is optional and set to | by default if not set.
669
Cyril Bonté0306c4a2015-10-26 22:37:38 +0100670deviceatlas-properties-cookie <name>
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200671 Sets the client cookie's name used for the detection if the DeviceAtlas
672 Client-side component was used during the request. This directive is optional
673 and set to DAPROPS by default if not set.
David Carlier29b3ca32015-09-25 14:09:21 +0100674
Simon Horman98637e52014-06-20 12:30:16 +0900675external-check
676 Allows the use of an external agent to perform health checks.
677 This is disabled by default as a security precaution.
678 See "option external-check".
679
Willy Tarreau6a06a402007-07-15 20:15:28 +0200680gid <number>
681 Changes the process' group ID to <number>. It is recommended that the group
682 ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
683 be started with a user belonging to this group, or with superuser privileges.
Michael Schererab012dd2013-01-12 18:35:19 +0100684 Note that if haproxy is started from a user having supplementary groups, it
685 will only be able to drop these groups if started with superuser privileges.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200686 See also "group" and "uid".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100687
Willy Tarreau6a06a402007-07-15 20:15:28 +0200688group <group name>
689 Similar to "gid" but uses the GID of group name <group name> from /etc/group.
690 See also "gid" and "user".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100691
Dragan Dosen7ad31542015-09-28 17:16:47 +0200692log <address> [len <length>] [format <format>] <facility> [max level [min level]]
Willy Tarreau6a06a402007-07-15 20:15:28 +0200693 Adds a global syslog server. Up to two global servers can be defined. They
694 will receive logs for startups and exits, as well as all logs from proxies
Robert Tsai81ae1952007-12-05 10:47:29 +0100695 configured with "log global".
696
697 <address> can be one of:
698
Willy Tarreau2769aa02007-12-27 18:26:09 +0100699 - An IPv4 address optionally followed by a colon and a UDP port. If
Robert Tsai81ae1952007-12-05 10:47:29 +0100700 no port is specified, 514 is used by default (the standard syslog
701 port).
702
David du Colombier24bb5f52011-03-17 10:40:23 +0100703 - An IPv6 address followed by a colon and optionally a UDP port. If
704 no port is specified, 514 is used by default (the standard syslog
705 port).
706
Robert Tsai81ae1952007-12-05 10:47:29 +0100707 - A filesystem path to a UNIX domain socket, keeping in mind
708 considerations for chroot (be sure the path is accessible inside
709 the chroot) and uid/gid (be sure the path is appropriately
710 writeable).
711
William Lallemandb2f07452015-05-12 14:27:13 +0200712 You may want to reference some environment variables in the address
713 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +0100714
Willy Tarreau18324f52014-06-27 18:10:07 +0200715 <length> is an optional maximum line length. Log lines larger than this value
716 will be truncated before being sent. The reason is that syslog
717 servers act differently on log line length. All servers support the
718 default value of 1024, but some servers simply drop larger lines
719 while others do log them. If a server supports long lines, it may
720 make sense to set this value here in order to avoid truncating long
721 lines. Similarly, if a server drops long lines, it is preferable to
722 truncate them before sending them. Accepted values are 80 to 65535
723 inclusive. The default value of 1024 is generally fine for all
724 standard usages. Some specific cases of long captures or
725 JSON-formated logs may require larger values.
726
Dragan Dosen7ad31542015-09-28 17:16:47 +0200727 <format> is the log format used when generating syslog messages. It may be
728 one of the following :
729
730 rfc3164 The RFC3164 syslog message format. This is the default.
731 (https://tools.ietf.org/html/rfc3164)
732
733 rfc5424 The RFC5424 syslog message format.
734 (https://tools.ietf.org/html/rfc5424)
735
Robert Tsai81ae1952007-12-05 10:47:29 +0100736 <facility> must be one of the 24 standard syslog facilities :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200737
738 kern user mail daemon auth syslog lpr news
739 uucp cron auth2 ftp ntp audit alert cron2
740 local0 local1 local2 local3 local4 local5 local6 local7
741
742 An optional level can be specified to filter outgoing messages. By default,
Willy Tarreauf7edefa2009-05-10 17:20:05 +0200743 all messages are sent. If a maximum level is specified, only messages with a
744 severity at least as important as this level will be sent. An optional minimum
745 level can be specified. If it is set, logs emitted with a more severe level
746 than this one will be capped to this level. This is used to avoid sending
747 "emerg" messages on all terminals on some default syslog configurations.
748 Eight levels are known :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200749
Cyril Bontédc4d9032012-04-08 21:57:39 +0200750 emerg alert crit err warning notice info debug
Willy Tarreau6a06a402007-07-15 20:15:28 +0200751
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100752log-send-hostname [<string>]
753 Sets the hostname field in the syslog header. If optional "string" parameter
754 is set the header is set to the string contents, otherwise uses the hostname
755 of the system. Generally used if one is not relaying logs through an
756 intermediate syslog server or for simply customizing the hostname printed in
757 the logs.
758
Kevinm48936af2010-12-22 16:08:21 +0000759log-tag <string>
760 Sets the tag field in the syslog header to this string. It defaults to the
761 program name as launched from the command line, which usually is "haproxy".
762 Sometimes it can be useful to differentiate between multiple processes
Willy Tarreau094af4e2015-01-07 15:03:42 +0100763 running on the same host. See also the per-proxy "log-tag" directive.
Kevinm48936af2010-12-22 16:08:21 +0000764
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100765lua-load <file>
766 This global directive loads and executes a Lua file. This directive can be
767 used multiple times.
768
Willy Tarreau6a06a402007-07-15 20:15:28 +0200769nbproc <number>
770 Creates <number> processes when going daemon. This requires the "daemon"
771 mode. By default, only one process is created, which is the recommended mode
772 of operation. For systems limited to small sets of file descriptors per
773 process, it may be needed to fork multiple daemons. USING MULTIPLE PROCESSES
774 IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. See also "daemon".
775
776pidfile <pidfile>
777 Writes pids of all daemons into file <pidfile>. This option is equivalent to
778 the "-p" command line argument. The file must be accessible to the user
779 starting the process. See also "daemon".
780
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100781stats bind-process [ all | odd | even | <number 1-64>[-<number 1-64>] ] ...
Willy Tarreau35b7b162012-10-22 23:17:18 +0200782 Limits the stats socket to a certain set of processes numbers. By default the
783 stats socket is bound to all processes, causing a warning to be emitted when
784 nbproc is greater than 1 because there is no way to select the target process
785 when connecting. However, by using this setting, it becomes possible to pin
786 the stats socket to a specific set of processes, typically the first one. The
787 warning will automatically be disabled when this setting is used, whatever
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100788 the number of processes used. The maximum process ID depends on the machine's
Willy Tarreauae302532014-05-07 19:22:24 +0200789 word size (32 or 64). A better option consists in using the "process" setting
790 of the "stats socket" line to force the process on each line.
Willy Tarreau35b7b162012-10-22 23:17:18 +0200791
Baptiste Assmann5626f482015-08-23 10:00:10 +0200792server-state-base <directory>
793 Specifies the directory prefix to be prepended in front of all servers state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +0200794 file names which do not start with a '/'. See also "server-state-file",
795 "load-server-state-from-file" and "server-state-file-name".
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200796
797server-state-file <file>
798 Specifies the path to the file containing state of servers. If the path starts
799 with a slash ('/'), it is considered absolute, otherwise it is considered
800 relative to the directory specified using "server-state-base" (if set) or to
801 the current directory. Before reloading HAProxy, it is possible to save the
802 servers' current state using the stats command "show servers state". The
803 output of this command must be written in the file pointed by <file>. When
804 starting up, before handling traffic, HAProxy will read, load and apply state
805 for each server found in the file and available in its current running
Baptiste Assmann01c6cc32015-08-23 11:45:29 +0200806 configuration. See also "server-state-base" and "show servers state",
807 "load-server-state-from-file" and "server-state-file-name"
Baptiste Assmann5626f482015-08-23 10:00:10 +0200808
Willy Tarreau610f04b2014-02-13 11:36:41 +0100809ssl-default-bind-ciphers <ciphers>
810 This setting is only available when support for OpenSSL was built in. It sets
811 the default string describing the list of cipher algorithms ("cipher suite")
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300812 that are negotiated during the SSL/TLS handshake for all "bind" lines which
Willy Tarreau610f04b2014-02-13 11:36:41 +0100813 do not explicitly define theirs. The format of the string is defined in
814 "man 1 ciphers" from OpenSSL man pages, and can be for instance a string such
815 as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes). Please check the
816 "bind" keyword for more information.
817
Emeric Brun2c86cbf2014-10-30 15:56:50 +0100818ssl-default-bind-options [<option>]...
819 This setting is only available when support for OpenSSL was built in. It sets
820 default ssl-options to force on all "bind" lines. Please check the "bind"
821 keyword to see available options.
822
823 Example:
824 global
825 ssl-default-bind-options no-sslv3 no-tls-tickets
826
Willy Tarreau610f04b2014-02-13 11:36:41 +0100827ssl-default-server-ciphers <ciphers>
828 This setting is only available when support for OpenSSL was built in. It
829 sets the default string describing the list of cipher algorithms that are
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300830 negotiated during the SSL/TLS handshake with the server, for all "server"
Willy Tarreau610f04b2014-02-13 11:36:41 +0100831 lines which do not explicitly define theirs. The format of the string is
832 defined in "man 1 ciphers". Please check the "server" keyword for more
833 information.
834
Emeric Brun2c86cbf2014-10-30 15:56:50 +0100835ssl-default-server-options [<option>]...
836 This setting is only available when support for OpenSSL was built in. It sets
837 default ssl-options to force on all "server" lines. Please check the "server"
838 keyword to see available options.
839
Remi Gacogne47783ef2015-05-29 15:53:22 +0200840ssl-dh-param-file <file>
841 This setting is only available when support for OpenSSL was built in. It sets
842 the default DH parameters that are used during the SSL/TLS handshake when
843 ephemeral Diffie-Hellman (DHE) key exchange is used, for all "bind" lines
844 which do not explicitely define theirs. It will be overridden by custom DH
845 parameters found in a bind certificate file if any. If custom DH parameters
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200846 are not specified either by using ssl-dh-param-file or by setting them
847 directly in the certificate file, pre-generated DH parameters of the size
848 specified by tune.ssl.default-dh-param will be used. Custom parameters are
849 known to be more secure and therefore their use is recommended.
Remi Gacogne47783ef2015-05-29 15:53:22 +0200850 Custom DH parameters may be generated by using the OpenSSL command
851 "openssl dhparam <size>", where size should be at least 2048, as 1024-bit DH
852 parameters should not be considered secure anymore.
853
Emeric Brun850efd52014-01-29 12:24:34 +0100854ssl-server-verify [none|required]
855 The default behavior for SSL verify on servers side. If specified to 'none',
856 servers certificates are not verified. The default is 'required' except if
857 forced using cmdline option '-dV'.
858
Willy Tarreauabb175f2012-09-24 12:43:26 +0200859stats socket [<address:port>|<path>] [param*]
860 Binds a UNIX socket to <path> or a TCPv4/v6 address to <address:port>.
861 Connections to this socket will return various statistics outputs and even
862 allow some commands to be issued to change some runtime settings. Please
Kevin Decherf949c7202015-10-13 23:26:44 +0200863 consult section 9.2 "Unix Socket commands" of Management Guide for more
864 details.
Willy Tarreau6162db22009-10-10 17:13:00 +0200865
Willy Tarreauabb175f2012-09-24 12:43:26 +0200866 All parameters supported by "bind" lines are supported, for instance to
867 restrict access to some users or their access rights. Please consult
868 section 5.1 for more information.
Willy Tarreaufbee7132007-10-18 13:53:22 +0200869
870stats timeout <timeout, in milliseconds>
871 The default timeout on the stats socket is set to 10 seconds. It is possible
872 to change this value with "stats timeout". The value must be passed in
Willy Tarreaubefdff12007-12-02 22:27:38 +0100873 milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.
Willy Tarreaufbee7132007-10-18 13:53:22 +0200874
875stats maxconn <connections>
876 By default, the stats socket is limited to 10 concurrent connections. It is
877 possible to change this value with "stats maxconn".
878
Willy Tarreau6a06a402007-07-15 20:15:28 +0200879uid <number>
880 Changes the process' user ID to <number>. It is recommended that the user ID
881 is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
882 be started with superuser privileges in order to be able to switch to another
883 one. See also "gid" and "user".
884
885ulimit-n <number>
886 Sets the maximum number of per-process file-descriptors to <number>. By
887 default, it is automatically computed, so it is recommended not to use this
888 option.
889
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100890unix-bind [ prefix <prefix> ] [ mode <mode> ] [ user <user> ] [ uid <uid> ]
891 [ group <group> ] [ gid <gid> ]
892
893 Fixes common settings to UNIX listening sockets declared in "bind" statements.
894 This is mainly used to simplify declaration of those UNIX sockets and reduce
895 the risk of errors, since those settings are most commonly required but are
896 also process-specific. The <prefix> setting can be used to force all socket
897 path to be relative to that directory. This might be needed to access another
898 component's chroot. Note that those paths are resolved before haproxy chroots
899 itself, so they are absolute. The <mode>, <user>, <uid>, <group> and <gid>
900 all have the same meaning as their homonyms used by the "bind" statement. If
901 both are specified, the "bind" statement has priority, meaning that the
902 "unix-bind" settings may be seen as process-wide default settings.
903
Willy Tarreau6a06a402007-07-15 20:15:28 +0200904user <user name>
905 Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.
906 See also "uid" and "group".
907
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +0200908node <name>
909 Only letters, digits, hyphen and underscore are allowed, like in DNS names.
910
911 This statement is useful in HA configurations where two or more processes or
912 servers share the same IP address. By setting a different node-name on all
913 nodes, it becomes easy to immediately spot what server is handling the
914 traffic.
915
916description <text>
917 Add a text that describes the instance.
918
919 Please note that it is required to escape certain characters (# for example)
920 and this text is inserted into a html page so you should avoid using
921 "<" and ">" characters.
922
Thomas Holmesdb04f192015-05-18 13:21:39 +010092351degrees-data-file <file path>
924 The path of the 51Degrees data file to provide device detection services. The
925 file should be unzipped and accessible by HAProxy with relevavnt permissions.
926
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200927 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +0100928 compiled with USE_51DEGREES.
929
93051degrees-property-name-list [<string>]
931 A list of 51Degrees property names to be load from the dataset. A full list
932 of names is available on the 51Degrees website:
933 https://51degrees.com/resources/property-dictionary
934
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200935 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +0100936 compiled with USE_51DEGREES.
937
Dragan Dosen93b38d92015-06-29 16:43:25 +020093851degrees-property-separator <char>
Thomas Holmesdb04f192015-05-18 13:21:39 +0100939 A char that will be appended to every property value in a response header
940 containing 51Degrees results. If not set that will be set as ','.
941
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200942 Please note that this option is only available when haproxy has been
943 compiled with USE_51DEGREES.
944
94551degrees-cache-size <number>
946 Sets the size of the 51Degrees converter cache to <number> entries. This
947 is an LRU cache which reminds previous device detections and their results.
948 By default, this cache is disabled.
949
950 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +0100951 compiled with USE_51DEGREES.
952
Willy Tarreau6a06a402007-07-15 20:15:28 +0200953
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009543.2. Performance tuning
Willy Tarreau6a06a402007-07-15 20:15:28 +0200955-----------------------
956
Willy Tarreau1746eec2014-04-25 10:46:47 +0200957max-spread-checks <delay in milliseconds>
958 By default, haproxy tries to spread the start of health checks across the
959 smallest health check interval of all the servers in a farm. The principle is
960 to avoid hammering services running on the same server. But when using large
961 check intervals (10 seconds or more), the last servers in the farm take some
962 time before starting to be tested, which can be a problem. This parameter is
963 used to enforce an upper bound on delay between the first and the last check,
964 even if the servers' check intervals are larger. When servers run with
965 shorter intervals, their intervals will be respected though.
966
Willy Tarreau6a06a402007-07-15 20:15:28 +0200967maxconn <number>
968 Sets the maximum per-process number of concurrent connections to <number>. It
969 is equivalent to the command-line argument "-n". Proxies will stop accepting
970 connections when this limit is reached. The "ulimit-n" parameter is
Willy Tarreau8274e102014-06-19 15:31:25 +0200971 automatically adjusted according to this value. See also "ulimit-n". Note:
972 the "select" poller cannot reliably use more than 1024 file descriptors on
973 some platforms. If your platform only supports select and reports "select
974 FAILED" on startup, you need to reduce maxconn until it works (slightly
Willy Tarreaud0256482015-01-15 21:45:22 +0100975 below 500 in general). If this value is not set, it will default to the value
976 set in DEFAULT_MAXCONN at build time (reported in haproxy -vv) if no memory
977 limit is enforced, or will be computed based on the memory limit, the buffer
978 size, memory allocated to compression, SSL cache size, and use or not of SSL
979 and the associated maxsslconn (which can also be automatic).
Willy Tarreau6a06a402007-07-15 20:15:28 +0200980
Willy Tarreau81c25d02011-09-07 15:17:21 +0200981maxconnrate <number>
982 Sets the maximum per-process number of connections per second to <number>.
983 Proxies will stop accepting connections when this limit is reached. It can be
984 used to limit the global capacity regardless of each frontend capacity. It is
985 important to note that this can only be used as a service protection measure,
986 as there will not necessarily be a fair share between frontends when the
987 limit is reached, so it's a good idea to also limit each frontend to some
988 value close to its expected share. Also, lowering tune.maxaccept can improve
989 fairness.
990
William Lallemandd85f9172012-11-09 17:05:39 +0100991maxcomprate <number>
992 Sets the maximum per-process input compression rate to <number> kilobytes
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300993 per second. For each session, if the maximum is reached, the compression
William Lallemandd85f9172012-11-09 17:05:39 +0100994 level will be decreased during the session. If the maximum is reached at the
995 beginning of a session, the session will not compress at all. If the maximum
996 is not reached, the compression level will be increased up to
997 tune.comp.maxlevel. A value of zero means there is no limit, this is the
998 default value.
999
William Lallemand072a2bf2012-11-20 17:01:01 +01001000maxcompcpuusage <number>
1001 Sets the maximum CPU usage HAProxy can reach before stopping the compression
1002 for new requests or decreasing the compression level of current requests.
1003 It works like 'maxcomprate' but measures CPU usage instead of incoming data
1004 bandwidth. The value is expressed in percent of the CPU used by haproxy. In
1005 case of multiple processes (nbproc > 1), each process manages its individual
1006 usage. A value of 100 disable the limit. The default value is 100. Setting
1007 a lower value will prevent the compression work from slowing the whole
1008 process down and from introducing high latencies.
1009
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001010maxpipes <number>
1011 Sets the maximum per-process number of pipes to <number>. Currently, pipes
1012 are only used by kernel-based tcp splicing. Since a pipe contains two file
1013 descriptors, the "ulimit-n" value will be increased accordingly. The default
1014 value is maxconn/4, which seems to be more than enough for most heavy usages.
1015 The splice code dynamically allocates and releases pipes, and can fall back
1016 to standard copy, so setting this value too low may only impact performance.
1017
Willy Tarreau93e7c002013-10-07 18:51:07 +02001018maxsessrate <number>
1019 Sets the maximum per-process number of sessions per second to <number>.
1020 Proxies will stop accepting connections when this limit is reached. It can be
1021 used to limit the global capacity regardless of each frontend capacity. It is
1022 important to note that this can only be used as a service protection measure,
1023 as there will not necessarily be a fair share between frontends when the
1024 limit is reached, so it's a good idea to also limit each frontend to some
1025 value close to its expected share. Also, lowering tune.maxaccept can improve
1026 fairness.
1027
Willy Tarreau403edff2012-09-06 11:58:37 +02001028maxsslconn <number>
1029 Sets the maximum per-process number of concurrent SSL connections to
1030 <number>. By default there is no SSL-specific limit, which means that the
1031 global maxconn setting will apply to all connections. Setting this limit
1032 avoids having openssl use too much memory and crash when malloc returns NULL
1033 (since it unfortunately does not reliably check for such conditions). Note
1034 that the limit applies both to incoming and outgoing connections, so one
1035 connection which is deciphered then ciphered accounts for 2 SSL connections.
Willy Tarreaud0256482015-01-15 21:45:22 +01001036 If this value is not set, but a memory limit is enforced, this value will be
1037 automatically computed based on the memory limit, maxconn, the buffer size,
1038 memory allocated to compression, SSL cache size, and use of SSL in either
1039 frontends, backends or both. If neither maxconn nor maxsslconn are specified
1040 when there is a memory limit, haproxy will automatically adjust these values
1041 so that 100% of the connections can be made over SSL with no risk, and will
1042 consider the sides where it is enabled (frontend, backend, both).
Willy Tarreau403edff2012-09-06 11:58:37 +02001043
Willy Tarreaue43d5322013-10-07 20:01:52 +02001044maxsslrate <number>
1045 Sets the maximum per-process number of SSL sessions per second to <number>.
1046 SSL listeners will stop accepting connections when this limit is reached. It
1047 can be used to limit the global SSL CPU usage regardless of each frontend
1048 capacity. It is important to note that this can only be used as a service
1049 protection measure, as there will not necessarily be a fair share between
1050 frontends when the limit is reached, so it's a good idea to also limit each
1051 frontend to some value close to its expected share. It is also important to
1052 note that the sessions are accounted before they enter the SSL stack and not
1053 after, which also protects the stack against bad handshakes. Also, lowering
1054 tune.maxaccept can improve fairness.
1055
William Lallemand9d5f5482012-11-07 16:12:57 +01001056maxzlibmem <number>
1057 Sets the maximum amount of RAM in megabytes per process usable by the zlib.
1058 When the maximum amount is reached, future sessions will not compress as long
1059 as RAM is unavailable. When sets to 0, there is no limit.
William Lallemande3a7d992012-11-20 11:25:20 +01001060 The default value is 0. The value is available in bytes on the UNIX socket
1061 with "show info" on the line "MaxZlibMemUsage", the memory used by zlib is
1062 "ZlibMemUsage" in bytes.
1063
Willy Tarreau6a06a402007-07-15 20:15:28 +02001064noepoll
1065 Disables the use of the "epoll" event polling system on Linux. It is
1066 equivalent to the command-line argument "-de". The next polling system
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001067 used will generally be "poll". See also "nopoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001068
1069nokqueue
1070 Disables the use of the "kqueue" event polling system on BSD. It is
1071 equivalent to the command-line argument "-dk". The next polling system
1072 used will generally be "poll". See also "nopoll".
1073
1074nopoll
1075 Disables the use of the "poll" event polling system. It is equivalent to the
1076 command-line argument "-dp". The next polling system used will be "select".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001077 It should never be needed to disable "poll" since it's available on all
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001078 platforms supported by HAProxy. See also "nokqueue" and "noepoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001079
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001080nosplice
1081 Disables the use of kernel tcp splicing between sockets on Linux. It is
1082 equivalent to the command line argument "-dS". Data will then be copied
1083 using conventional and more portable recv/send calls. Kernel tcp splicing is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001084 limited to some very recent instances of kernel 2.6. Most versions between
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001085 2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not
1086 be used. This option makes it easier to globally disable kernel splicing in
1087 case of doubt. See also "option splice-auto", "option splice-request" and
1088 "option splice-response".
1089
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001090nogetaddrinfo
1091 Disables the use of getaddrinfo(3) for name resolving. It is equivalent to
1092 the command line argument "-dG". Deprecated gethostbyname(3) will be used.
1093
Willy Tarreaufe255b72007-10-14 23:09:26 +02001094spread-checks <0..50, in percent>
Simon Hormand60d6912013-11-25 10:46:36 +09001095 Sometimes it is desirable to avoid sending agent and health checks to
1096 servers at exact intervals, for instance when many logical servers are
1097 located on the same physical server. With the help of this parameter, it
1098 becomes possible to add some randomness in the check interval between 0
1099 and +/- 50%. A value between 2 and 5 seems to show good results. The
1100 default value remains at 0.
Willy Tarreaufe255b72007-10-14 23:09:26 +02001101
Willy Tarreau33cb0652014-12-23 22:52:37 +01001102tune.buffers.limit <number>
1103 Sets a hard limit on the number of buffers which may be allocated per process.
1104 The default value is zero which means unlimited. The minimum non-zero value
1105 will always be greater than "tune.buffers.reserve" and should ideally always
1106 be about twice as large. Forcing this value can be particularly useful to
1107 limit the amount of memory a process may take, while retaining a sane
1108 behaviour. When this limit is reached, sessions which need a buffer wait for
1109 another one to be released by another session. Since buffers are dynamically
1110 allocated and released, the waiting time is very short and not perceptible
1111 provided that limits remain reasonable. In fact sometimes reducing the limit
1112 may even increase performance by increasing the CPU cache's efficiency. Tests
1113 have shown good results on average HTTP traffic with a limit to 1/10 of the
1114 expected global maxconn setting, which also significantly reduces memory
1115 usage. The memory savings come from the fact that a number of connections
1116 will not allocate 2*tune.bufsize. It is best not to touch this value unless
1117 advised to do so by an haproxy core developer.
1118
Willy Tarreau1058ae72014-12-23 22:40:40 +01001119tune.buffers.reserve <number>
1120 Sets the number of buffers which are pre-allocated and reserved for use only
1121 during memory shortage conditions resulting in failed memory allocations. The
1122 minimum value is 2 and is also the default. There is no reason a user would
1123 want to change this value, it's mostly aimed at haproxy core developers.
1124
Willy Tarreau27a674e2009-08-17 07:23:33 +02001125tune.bufsize <number>
1126 Sets the buffer size to this size (in bytes). Lower values allow more
1127 sessions to coexist in the same amount of RAM, and higher values allow some
1128 applications with very large cookies to work. The default value is 16384 and
1129 can be changed at build time. It is strongly recommended not to change this
1130 from the default value, as very low values will break some services such as
1131 statistics, and values larger than default size will increase memory usage,
1132 possibly causing the system to run out of memory. At least the global maxconn
1133 parameter should be decreased by the same factor as this one is increased.
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04001134 If HTTP request is larger than (tune.bufsize - tune.maxrewrite), haproxy will
1135 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
1136 than this size, haproxy will return HTTP 502 (Bad Gateway).
Willy Tarreau27a674e2009-08-17 07:23:33 +02001137
Willy Tarreau43961d52010-10-04 20:39:20 +02001138tune.chksize <number>
1139 Sets the check buffer size to this size (in bytes). Higher values may help
1140 find string or regex patterns in very large pages, though doing so may imply
1141 more memory and CPU usage. The default value is 16384 and can be changed at
1142 build time. It is not recommended to change this value, but to use better
1143 checks whenever possible.
1144
William Lallemandf3747832012-11-09 12:33:10 +01001145tune.comp.maxlevel <number>
1146 Sets the maximum compression level. The compression level affects CPU
1147 usage during compression. This value affects CPU usage during compression.
1148 Each session using compression initializes the compression algorithm with
1149 this value. The default value is 1.
1150
Willy Tarreau193b8c62012-11-22 00:17:38 +01001151tune.http.cookielen <number>
1152 Sets the maximum length of captured cookies. This is the maximum value that
1153 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
1154 will automatically be truncated to this one. It is important not to set too
1155 high a value because all cookie captures still allocate this size whatever
1156 their configured value (they share a same pool). This value is per request
1157 per response, so the memory allocated is twice this value per connection.
1158 When not specified, the limit is set to 63 characters. It is recommended not
1159 to change this value.
1160
Willy Tarreauac1932d2011-10-24 19:14:41 +02001161tune.http.maxhdr <number>
1162 Sets the maximum number of headers in a request. When a request comes with a
1163 number of headers greater than this value (including the first line), it is
1164 rejected with a "400 Bad Request" status code. Similarly, too large responses
1165 are blocked with "502 Bad Gateway". The default value is 101, which is enough
1166 for all usages, considering that the widely deployed Apache server uses the
1167 same limit. It can be useful to push this limit further to temporarily allow
1168 a buggy application to work by the time it gets fixed. Keep in mind that each
1169 new header consumes 32bits of memory for each session, so don't push this
1170 limit too high.
1171
Willy Tarreau7e312732014-02-12 16:35:14 +01001172tune.idletimer <timeout>
1173 Sets the duration after which haproxy will consider that an empty buffer is
1174 probably associated with an idle stream. This is used to optimally adjust
1175 some packet sizes while forwarding large and small data alternatively. The
1176 decision to use splice() or to send large buffers in SSL is modulated by this
1177 parameter. The value is in milliseconds between 0 and 65535. A value of zero
1178 means that haproxy will not try to detect idle streams. The default is 1000,
1179 which seems to correctly detect end user pauses (eg: read a page before
1180 clicking). There should be not reason for changing this value. Please check
1181 tune.ssl.maxrecord below.
1182
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001183tune.lua.forced-yield <number>
1184 This directive forces the Lua engine to execute a yield each <number> of
1185 instructions executed. This permits interruptng a long script and allows the
1186 HAProxy scheduler to process other tasks like accepting connections or
1187 forwarding traffic. The default value is 10000 instructions. If HAProxy often
1188 executes some Lua code but more reactivity is required, this value can be
1189 lowered. If the Lua code is quite long and its result is absolutely required
1190 to process the data, the <number> can be increased.
1191
Willy Tarreau32f61e22015-03-18 17:54:59 +01001192tune.lua.maxmem
1193 Sets the maximum amount of RAM in megabytes per process usable by Lua. By
1194 default it is zero which means unlimited. It is important to set a limit to
1195 ensure that a bug in a script will not result in the system running out of
1196 memory.
1197
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001198tune.lua.session-timeout <timeout>
1199 This is the execution timeout for the Lua sessions. This is useful for
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001200 preventing infinite loops or spending too much time in Lua. This timeout
1201 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
1202 not taked in account. The default timeout is 4s.
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001203
1204tune.lua.task-timeout <timeout>
1205 Purpose is the same as "tune.lua.session-timeout", but this timeout is
1206 dedicated to the tasks. By default, this timeout isn't set because a task may
1207 remain alive during of the lifetime of HAProxy. For example, a task used to
1208 check servers.
1209
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001210tune.lua.service-timeout <timeout>
1211 This is the execution timeout for the Lua services. This is useful for
1212 preventing infinite loops or spending too much time in Lua. This timeout
1213 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
1214 not taked in account. The default timeout is 4s.
1215
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001216tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +01001217 Sets the maximum number of consecutive connections a process may accept in a
1218 row before switching to other work. In single process mode, higher numbers
1219 give better performance at high connection rates. However in multi-process
1220 modes, keeping a bit of fairness between processes generally is better to
1221 increase performance. This value applies individually to each listener, so
1222 that the number of processes a listener is bound to is taken into account.
1223 This value defaults to 64. In multi-process mode, it is divided by twice
1224 the number of processes the listener is bound to. Setting this value to -1
1225 completely disables the limitation. It should normally not be needed to tweak
1226 this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001227
1228tune.maxpollevents <number>
1229 Sets the maximum amount of events that can be processed at once in a call to
1230 the polling system. The default value is adapted to the operating system. It
1231 has been noticed that reducing it below 200 tends to slightly decrease
1232 latency at the expense of network bandwidth, and increasing it above 200
1233 tends to trade latency for slightly increased bandwidth.
1234
Willy Tarreau27a674e2009-08-17 07:23:33 +02001235tune.maxrewrite <number>
1236 Sets the reserved buffer space to this size in bytes. The reserved space is
1237 used for header rewriting or appending. The first reads on sockets will never
1238 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
1239 bufsize, though that does not make much sense since there are rarely large
1240 numbers of headers to add. Setting it too high prevents processing of large
1241 requests or responses. Setting it too low prevents addition of new headers
1242 to already large requests or to POST requests. It is generally wise to set it
1243 to about 1024. It is automatically readjusted to half of bufsize if it is
1244 larger than that. This means you don't have to worry about it when changing
1245 bufsize.
1246
Willy Tarreauf3045d22015-04-29 16:24:50 +02001247tune.pattern.cache-size <number>
1248 Sets the size of the pattern lookup cache to <number> entries. This is an LRU
1249 cache which reminds previous lookups and their results. It is used by ACLs
1250 and maps on slow pattern lookups, namely the ones using the "sub", "reg",
1251 "dir", "dom", "end", "bin" match methods as well as the case-insensitive
1252 strings. It applies to pattern expressions which means that it will be able
1253 to memorize the result of a lookup among all the patterns specified on a
1254 configuration line (including all those loaded from files). It automatically
1255 invalidates entries which are updated using HTTP actions or on the CLI. The
1256 default cache size is set to 10000 entries, which limits its footprint to
1257 about 5 MB on 32-bit systems and 8 MB on 64-bit systems. There is a very low
1258 risk of collision in this cache, which is in the order of the size of the
1259 cache divided by 2^64. Typically, at 10000 requests per second with the
1260 default cache size of 10000 entries, there's 1% chance that a brute force
1261 attack could cause a single collision after 60 years, or 0.1% after 6 years.
1262 This is considered much lower than the risk of a memory corruption caused by
1263 aging components. If this is not acceptable, the cache can be disabled by
1264 setting this parameter to 0.
1265
Willy Tarreaubd9a0a72011-10-23 21:14:29 +02001266tune.pipesize <number>
1267 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
1268 are the default size for the system. But sometimes when using TCP splicing,
1269 it can improve performance to increase pipe sizes, especially if it is
1270 suspected that pipes are not filled and that many calls to splice() are
1271 performed. This has an impact on the kernel's memory footprint, so this must
1272 not be changed if impacts are not understood.
1273
Willy Tarreaue803de22010-01-21 17:43:04 +01001274tune.rcvbuf.client <number>
1275tune.rcvbuf.server <number>
1276 Forces the kernel socket receive buffer size on the client or the server side
1277 to the specified value in bytes. This value applies to all TCP/HTTP frontends
1278 and backends. It should normally never be set, and the default size (0) lets
1279 the kernel autotune this value depending on the amount of available memory.
1280 However it can sometimes help to set it to very low values (eg: 4096) in
1281 order to save kernel memory by preventing it from buffering too large amounts
1282 of received data. Lower values will significantly increase CPU usage though.
1283
Willy Tarreaub22fc302015-12-14 12:04:35 +01001284tune.recv_enough <number>
1285 Haproxy uses some hints to detect that a short read indicates the end of the
1286 socket buffers. One of them is that a read returns more than <recv_enough>
1287 bytes, which defaults to 10136 (7 segments of 1448 each). This default value
1288 may be changed by this setting to better deal with workloads involving lots
1289 of short messages such as telnet or SSH sessions.
1290
Willy Tarreaue803de22010-01-21 17:43:04 +01001291tune.sndbuf.client <number>
1292tune.sndbuf.server <number>
1293 Forces the kernel socket send buffer size on the client or the server side to
1294 the specified value in bytes. This value applies to all TCP/HTTP frontends
1295 and backends. It should normally never be set, and the default size (0) lets
1296 the kernel autotune this value depending on the amount of available memory.
1297 However it can sometimes help to set it to very low values (eg: 4096) in
1298 order to save kernel memory by preventing it from buffering too large amounts
1299 of received data. Lower values will significantly increase CPU usage though.
1300 Another use case is to prevent write timeouts with extremely slow clients due
1301 to the kernel waiting for a large part of the buffer to be read before
1302 notifying haproxy again.
1303
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001304tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +01001305 Sets the size of the global SSL session cache, in a number of blocks. A block
1306 is large enough to contain an encoded session without peer certificate.
1307 An encoded session with peer certificate is stored in multiple blocks
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001308 depending on the size of the peer certificate. A block uses approximately
Emeric Brunaf9619d2012-11-28 18:47:52 +01001309 200 bytes of memory. The default value may be forced at build time, otherwise
1310 defaults to 20000. When the cache is full, the most idle entries are purged
1311 and reassigned. Higher values reduce the occurrence of such a purge, hence
1312 the number of CPU-intensive SSL handshakes by ensuring that all users keep
1313 their session as long as possible. All entries are pre-allocated upon startup
Emeric Brun22890a12012-12-28 14:41:32 +01001314 and are shared between all processes if "nbproc" is greater than 1. Setting
1315 this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001316
Emeric Brun8dc60392014-05-09 13:52:00 +02001317tune.ssl.force-private-cache
1318 This boolean disables SSL session cache sharing between all processes. It
1319 should normally not be used since it will force many renegotiations due to
1320 clients hitting a random process. But it may be required on some operating
1321 systems where none of the SSL cache synchronization method may be used. In
1322 this case, adding a first layer of hash-based load balancing before the SSL
1323 layer might limit the impact of the lack of session sharing.
1324
Emeric Brun4f65bff2012-11-16 15:11:00 +01001325tune.ssl.lifetime <timeout>
1326 Sets how long a cached SSL session may remain valid. This time is expressed
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001327 in seconds and defaults to 300 (5 min). It is important to understand that it
Emeric Brun4f65bff2012-11-16 15:11:00 +01001328 does not guarantee that sessions will last that long, because if the cache is
1329 full, the longest idle sessions will be purged despite their configured
1330 lifetime. The real usefulness of this setting is to prevent sessions from
1331 being used for too long.
1332
Willy Tarreaubfd59462013-02-21 07:46:09 +01001333tune.ssl.maxrecord <number>
1334 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
1335 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
1336 data only once it has received a full record. With large records, it means
1337 that clients might have to download up to 16kB of data before starting to
1338 process them. Limiting the value can improve page load times on browsers
1339 located over high latency or low bandwidth networks. It is suggested to find
1340 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
1341 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
1342 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
1343 2859 gave good results during tests. Use "strace -e trace=write" to find the
Willy Tarreau7e312732014-02-12 16:35:14 +01001344 best value. Haproxy will automatically switch to this setting after an idle
1345 stream has been detected (see tune.idletimer above).
Willy Tarreaubfd59462013-02-21 07:46:09 +01001346
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001347tune.ssl.default-dh-param <number>
1348 Sets the maximum size of the Diffie-Hellman parameters used for generating
1349 the ephemeral/temporary Diffie-Hellman key in case of DHE key exchange. The
1350 final size will try to match the size of the server's RSA (or DSA) key (e.g,
1351 a 2048 bits temporary DH key for a 2048 bits RSA key), but will not exceed
1352 this maximum value. Default value if 1024. Only 1024 or higher values are
1353 allowed. Higher values will increase the CPU load, and values greater than
1354 1024 bits are not supported by Java 7 and earlier clients. This value is not
Remi Gacogne47783ef2015-05-29 15:53:22 +02001355 used if static Diffie-Hellman parameters are supplied either directly
1356 in the certificate file or by using the ssl-dh-param-file parameter.
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001357
Christopher Faulet31af49d2015-06-09 17:29:50 +02001358tune.ssl.ssl-ctx-cache-size <number>
1359 Sets the size of the cache used to store generated certificates to <number>
1360 entries. This is a LRU cache. Because generating a SSL certificate
1361 dynamically is expensive, they are cached. The default cache size is set to
1362 1000 entries.
1363
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001364tune.vars.global-max-size <size>
1365tune.vars.reqres-max-size <size>
1366tune.vars.sess-max-size <size>
1367tune.vars.txn-max-size <size>
1368 These four tunes helps to manage the allowed amount of memory used by the
1369 variables system. "global" limits the memory for all the systems. "sess" limit
1370 the memory by session, "txn" limits the memory by transaction and "reqres"
1371 limits the memory for each request or response processing. during the
1372 accounting, "sess" embbed "txn" and "txn" embed "reqres".
1373
1374 By example, we considers that "tune.vars.sess-max-size" is fixed to 100,
1375 "tune.vars.txn-max-size" is fixed to 100, "tune.vars.reqres-max-size" is
1376 also fixed to 100. If we create a variable "txn.var" that contains 100 bytes,
1377 we cannot create any more variable in the other contexts.
1378
William Lallemanda509e4c2012-11-07 16:54:34 +01001379tune.zlib.memlevel <number>
1380 Sets the memLevel parameter in zlib initialization for each session. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001381 defines how much memory should be allocated for the internal compression
William Lallemanda509e4c2012-11-07 16:54:34 +01001382 state. A value of 1 uses minimum memory but is slow and reduces compression
1383 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
1384 between 1 and 9. The default value is 8.
1385
1386tune.zlib.windowsize <number>
1387 Sets the window size (the size of the history buffer) as a parameter of the
1388 zlib initialization for each session. Larger values of this parameter result
1389 in better compression at the expense of memory usage. Can be a value between
1390 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001391
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013923.3. Debugging
1393--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02001394
1395debug
1396 Enables debug mode which dumps to stdout all exchanges, and disables forking
1397 into background. It is the equivalent of the command-line argument "-d". It
1398 should never be used in a production configuration since it may prevent full
1399 system startup.
1400
1401quiet
1402 Do not display any message during startup. It is equivalent to the command-
1403 line argument "-q".
1404
Emeric Brunf099e792010-09-27 12:05:28 +02001405
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010014063.4. Userlists
1407--------------
1408It is possible to control access to frontend/backend/listen sections or to
1409http stats by allowing only authenticated and authorized users. To do this,
1410it is required to create at least one userlist and to define users.
1411
1412userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +01001413 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001414 used to store authentication & authorization data for independent customers.
1415
1416group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +01001417 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001418 attach users to this group by using a comma separated list of names
1419 proceeded by "users" keyword.
1420
Cyril Bontéf0c60612010-02-06 14:44:47 +01001421user <username> [password|insecure-password <password>]
1422 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001423 Adds user <username> to the current userlist. Both secure (encrypted) and
1424 insecure (unencrypted) passwords can be used. Encrypted passwords are
Cyril Bonté78caf842010-03-10 22:41:43 +01001425 evaluated using the crypt(3) function so depending of the system's
1426 capabilities, different algorithms are supported. For example modern Glibc
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001427 based Linux system supports MD5, SHA-256, SHA-512 and of course classic,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001428 DES-based method of encrypting passwords.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001429
1430
1431 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +01001432 userlist L1
1433 group G1 users tiger,scott
1434 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001435
Cyril Bontéf0c60612010-02-06 14:44:47 +01001436 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
1437 user scott insecure-password elgato
1438 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001439
Cyril Bontéf0c60612010-02-06 14:44:47 +01001440 userlist L2
1441 group G1
1442 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001443
Cyril Bontéf0c60612010-02-06 14:44:47 +01001444 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
1445 user scott insecure-password elgato groups G1,G2
1446 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001447
1448 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001449
Emeric Brunf099e792010-09-27 12:05:28 +02001450
14513.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02001452----------
Emeric Brun94900952015-06-11 18:25:54 +02001453It is possible to propagate entries of any data-types in stick-tables between
1454several haproxy instances over TCP connections in a multi-master fashion. Each
1455instance pushes its local updates and insertions to remote peers. The pushed
1456values overwrite remote ones without aggregation. Interrupted exchanges are
1457automatically detected and recovered from the last known point.
1458In addition, during a soft restart, the old process connects to the new one
1459using such a TCP connection to push all its entries before the new process
1460tries to connect to other peers. That ensures very fast replication during a
1461reload, it typically takes a fraction of a second even for large tables.
1462Note that Server IDs are used to identify servers remotely, so it is important
1463that configurations look similar or at least that the same IDs are forced on
1464each server on all participants.
Emeric Brunf099e792010-09-27 12:05:28 +02001465
1466peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001467 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02001468 which is referenced by one or more stick-tables.
1469
Willy Tarreau77e4bd12015-05-01 20:02:17 +02001470disabled
1471 Disables a peers section. It disables both listening and any synchronization
1472 related to this section. This is provided to disable synchronization of stick
1473 tables without having to comment out all "peers" references.
1474
1475enable
1476 This re-enables a disabled peers section which was previously disabled.
1477
Emeric Brunf099e792010-09-27 12:05:28 +02001478peer <peername> <ip>:<port>
1479 Defines a peer inside a peers section.
1480 If <peername> is set to the local peer name (by default hostname, or forced
1481 using "-L" command line option), haproxy will listen for incoming remote peer
1482 connection on <ip>:<port>. Otherwise, <ip>:<port> defines where to connect to
1483 to join the remote peer, and <peername> is used at the protocol level to
1484 identify and validate the remote peer on the server side.
1485
1486 During a soft restart, local peer <ip>:<port> is used by the old instance to
1487 connect the new one and initiate a complete replication (teaching process).
1488
1489 It is strongly recommended to have the exact same peers declaration on all
1490 peers and to only rely on the "-L" command line argument to change the local
1491 peer name. This makes it easier to maintain coherent configuration files
1492 across all peers.
1493
William Lallemandb2f07452015-05-12 14:27:13 +02001494 You may want to reference some environment variables in the address
1495 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01001496
Cyril Bontédc4d9032012-04-08 21:57:39 +02001497 Example:
Emeric Brunf099e792010-09-27 12:05:28 +02001498 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001499 peer haproxy1 192.168.0.1:1024
1500 peer haproxy2 192.168.0.2:1024
1501 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02001502
1503 backend mybackend
1504 mode tcp
1505 balance roundrobin
1506 stick-table type ip size 20k peers mypeers
1507 stick on src
1508
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001509 server srv1 192.168.0.30:80
1510 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02001511
1512
Simon Horman51a1cf62015-02-03 13:00:44 +090015133.6. Mailers
1514------------
1515It is possible to send email alerts when the state of servers changes.
1516If configured email alerts are sent to each mailer that is configured
1517in a mailers section. Email is sent to mailers using SMTP.
1518
Pieter Baauw386a1272015-08-16 15:26:24 +02001519mailers <mailersect>
Simon Horman51a1cf62015-02-03 13:00:44 +09001520 Creates a new mailer list with the name <mailersect>. It is an
1521 independent section which is referenced by one or more proxies.
1522
1523mailer <mailername> <ip>:<port>
1524 Defines a mailer inside a mailers section.
1525
1526 Example:
1527 mailers mymailers
1528 mailer smtp1 192.168.0.1:587
1529 mailer smtp2 192.168.0.2:587
1530
1531 backend mybackend
1532 mode tcp
1533 balance roundrobin
1534
1535 email-alert mailers mymailers
1536 email-alert from test1@horms.org
1537 email-alert to test2@horms.org
1538
1539 server srv1 192.168.0.30:80
1540 server srv2 192.168.0.31:80
1541
1542
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015434. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02001544----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001545
Willy Tarreau6a06a402007-07-15 20:15:28 +02001546Proxy configuration can be located in a set of sections :
William Lallemand6e62fb62015-04-28 16:55:23 +02001547 - defaults [<name>]
Willy Tarreau6a06a402007-07-15 20:15:28 +02001548 - frontend <name>
1549 - backend <name>
1550 - listen <name>
1551
1552A "defaults" section sets default parameters for all other sections following
1553its declaration. Those default parameters are reset by the next "defaults"
1554section. See below for the list of parameters which can be set in a "defaults"
Willy Tarreau0ba27502007-12-24 16:55:16 +01001555section. The name is optional but its use is encouraged for better readability.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001556
1557A "frontend" section describes a set of listening sockets accepting client
1558connections.
1559
1560A "backend" section describes a set of servers to which the proxy will connect
1561to forward incoming connections.
1562
1563A "listen" section defines a complete proxy with its frontend and backend
1564parts combined in one section. It is generally useful for TCP-only traffic.
1565
Willy Tarreau0ba27502007-12-24 16:55:16 +01001566All proxy names must be formed from upper and lower case letters, digits,
1567'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
1568case-sensitive, which means that "www" and "WWW" are two different proxies.
1569
1570Historically, all proxy names could overlap, it just caused troubles in the
1571logs. Since the introduction of content switching, it is mandatory that two
1572proxies with overlapping capabilities (frontend/backend) have different names.
1573However, it is still permitted that a frontend and a backend share the same
1574name, as this configuration seems to be commonly encountered.
1575
1576Right now, two major proxy modes are supported : "tcp", also known as layer 4,
1577and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001578bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01001579protocol, and can interact with it by allowing, blocking, switching, adding,
1580modifying, or removing arbitrary contents in requests or responses, based on
1581arbitrary criteria.
1582
Willy Tarreau70dffda2014-01-30 03:07:23 +01001583In HTTP mode, the processing applied to requests and responses flowing over
1584a connection depends in the combination of the frontend's HTTP options and
1585the backend's. HAProxy supports 5 connection modes :
1586
1587 - KAL : keep alive ("option http-keep-alive") which is the default mode : all
1588 requests and responses are processed, and connections remain open but idle
1589 between responses and new requests.
1590
1591 - TUN: tunnel ("option http-tunnel") : this was the default mode for versions
1592 1.0 to 1.5-dev21 : only the first request and response are processed, and
1593 everything else is forwarded with no analysis at all. This mode should not
1594 be used as it creates lots of trouble with logging and HTTP processing.
1595
1596 - PCL: passive close ("option httpclose") : exactly the same as tunnel mode,
1597 but with "Connection: close" appended in both directions to try to make
1598 both ends close after the first request/response exchange.
1599
1600 - SCL: server close ("option http-server-close") : the server-facing
1601 connection is closed after the end of the response is received, but the
1602 client-facing connection remains open.
1603
1604 - FCL: forced close ("option forceclose") : the connection is actively closed
1605 after the end of the response.
1606
1607The effective mode that will be applied to a connection passing through a
1608frontend and a backend can be determined by both proxy modes according to the
1609following matrix, but in short, the modes are symmetric, keep-alive is the
1610weakest option and force close is the strongest.
1611
1612 Backend mode
1613
1614 | KAL | TUN | PCL | SCL | FCL
1615 ----+-----+-----+-----+-----+----
1616 KAL | KAL | TUN | PCL | SCL | FCL
1617 ----+-----+-----+-----+-----+----
1618 TUN | TUN | TUN | PCL | SCL | FCL
1619 Frontend ----+-----+-----+-----+-----+----
1620 mode PCL | PCL | PCL | PCL | FCL | FCL
1621 ----+-----+-----+-----+-----+----
1622 SCL | SCL | SCL | FCL | SCL | FCL
1623 ----+-----+-----+-----+-----+----
1624 FCL | FCL | FCL | FCL | FCL | FCL
1625
Willy Tarreau0ba27502007-12-24 16:55:16 +01001626
Willy Tarreau70dffda2014-01-30 03:07:23 +01001627
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016284.1. Proxy keywords matrix
1629--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001630
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001631The following list of keywords is supported. Most of them may only be used in a
1632limited set of section types. Some of them are marked as "deprecated" because
1633they are inherited from an old syntax which may be confusing or functionally
1634limited, and there are new recommended keywords to replace them. Keywords
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001635marked with "(*)" can be optionally inverted using the "no" prefix, eg. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001636option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02001637and must be disabled for a specific instance. Such options may also be prefixed
1638with "default" in order to restore default settings regardless of what has been
1639specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001640
Willy Tarreau6a06a402007-07-15 20:15:28 +02001641
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001642 keyword defaults frontend listen backend
1643------------------------------------+----------+----------+---------+---------
1644acl - X X X
Willy Tarreau294d0f02015-08-10 19:40:12 +02001645appsession - - - -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001646backlog X X X -
1647balance X - X X
1648bind - X X -
1649bind-process X X X X
1650block - X X X
1651capture cookie - X X -
1652capture request header - X X -
1653capture response header - X X -
1654clitimeout (deprecated) X X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02001655compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001656contimeout (deprecated) X - X X
1657cookie X - X X
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02001658declare capture - X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001659default-server X - X X
1660default_backend X X X -
1661description - X X X
1662disabled X X X X
1663dispatch - - X X
Simon Horman51a1cf62015-02-03 13:00:44 +09001664email-alert from X X X X
Simon Horman64e34162015-02-06 11:11:57 +09001665email-alert level X X X X
Simon Horman51a1cf62015-02-03 13:00:44 +09001666email-alert mailers X X X X
1667email-alert myhostname X X X X
1668email-alert to X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001669enabled X X X X
1670errorfile X X X X
1671errorloc X X X X
1672errorloc302 X X X X
1673-- keyword -------------------------- defaults - frontend - listen -- backend -
1674errorloc303 X X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001675force-persist - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001676fullconn X - X X
1677grace X X X X
1678hash-type X - X X
1679http-check disable-on-404 X - X X
Willy Tarreaubd741542010-03-16 18:46:54 +01001680http-check expect - - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02001681http-check send-state X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001682http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02001683http-response - X X X
Willy Tarreau30631952015-08-06 15:05:24 +02001684http-reuse X - X X
Baptiste Assmann2c42ef52013-10-09 21:57:02 +02001685http-send-name-header - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001686id - X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001687ignore-persist - X X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001688load-server-state-from-file X - X X
William Lallemand0f99e342011-10-12 17:50:54 +02001689log (*) X X X X
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01001690log-format X X X -
Dragan Dosen7ad31542015-09-28 17:16:47 +02001691log-format-sd X X X -
Willy Tarreau094af4e2015-01-07 15:03:42 +01001692log-tag X X X X
Willy Tarreauc35362a2014-04-25 13:58:37 +02001693max-keep-alive-queue X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001694maxconn X X X -
1695mode X X X X
1696monitor fail - X X -
1697monitor-net X X X -
1698monitor-uri X X X -
1699option abortonclose (*) X - X X
1700option accept-invalid-http-request (*) X X X -
1701option accept-invalid-http-response (*) X - X X
1702option allbackups (*) X - X X
1703option checkcache (*) X - X X
1704option clitcpka (*) X X X -
1705option contstats (*) X X X -
1706option dontlog-normal (*) X X X -
1707option dontlognull (*) X X X -
1708option forceclose (*) X X X X
1709-- keyword -------------------------- defaults - frontend - listen -- backend -
1710option forwardfor X X X X
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02001711option http-buffer-request (*) X X X X
Willy Tarreau82649f92015-05-01 22:40:51 +02001712option http-ignore-probes (*) X X X -
Willy Tarreau16bfb022010-01-16 19:48:41 +01001713option http-keep-alive (*) X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02001714option http-no-delay (*) X X X X
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02001715option http-pretend-keepalive (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001716option http-server-close (*) X X X X
Willy Tarreau02bce8b2014-01-30 00:15:28 +01001717option http-tunnel (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001718option http-use-proxy-header (*) X X X -
1719option httpchk X - X X
1720option httpclose (*) X X X X
1721option httplog X X X X
1722option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001723option independent-streams (*) X X X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02001724option ldap-check X - X X
Simon Horman98637e52014-06-20 12:30:16 +09001725option external-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001726option log-health-checks (*) X - X X
1727option log-separate-errors (*) X X X -
1728option logasap (*) X X X -
1729option mysql-check X - X X
1730option nolinger (*) X X X X
1731option originalto X X X X
1732option persist (*) X - X X
Baptiste Assmann809e22a2015-10-12 20:22:55 +02001733option pgsql-check X - X X
1734option prefer-last-server (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001735option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02001736option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001737option smtpchk X - X X
1738option socket-stats (*) X X X -
1739option splice-auto (*) X X X X
1740option splice-request (*) X X X X
1741option splice-response (*) X X X X
1742option srvtcpka (*) X - X X
1743option ssl-hello-chk X - X X
1744-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreaued179852013-12-16 01:07:00 +01001745option tcp-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001746option tcp-smart-accept (*) X X X -
1747option tcp-smart-connect (*) X - X X
1748option tcpka X X X X
1749option tcplog X X X X
1750option transparent (*) X - X X
Simon Horman98637e52014-06-20 12:30:16 +09001751external-check command X - X X
1752external-check path X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001753persist rdp-cookie X - X X
1754rate-limit sessions X X X -
1755redirect - X X X
1756redisp (deprecated) X - X X
1757redispatch (deprecated) X - X X
1758reqadd - X X X
1759reqallow - X X X
1760reqdel - X X X
1761reqdeny - X X X
1762reqiallow - X X X
1763reqidel - X X X
1764reqideny - X X X
1765reqipass - X X X
1766reqirep - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001767reqitarpit - X X X
1768reqpass - X X X
1769reqrep - X X X
1770-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001771reqtarpit - X X X
1772retries X - X X
1773rspadd - X X X
1774rspdel - X X X
1775rspdeny - X X X
1776rspidel - X X X
1777rspideny - X X X
1778rspirep - X X X
1779rsprep - X X X
1780server - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001781server-state-file-name X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001782source X - X X
1783srvtimeout (deprecated) X - X X
Baptiste Assmann5a549212015-10-12 20:30:24 +02001784stats admin - X X X
1785stats auth X X X X
1786stats enable X X X X
1787stats hide-version X X X X
1788stats http-request - X X X
1789stats realm X X X X
1790stats refresh X X X X
1791stats scope X X X X
1792stats show-desc X X X X
1793stats show-legends X X X X
1794stats show-node X X X X
1795stats uri X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001796-- keyword -------------------------- defaults - frontend - listen -- backend -
1797stick match - - X X
1798stick on - - X X
1799stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02001800stick store-response - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001801stick-table - - X X
Willy Tarreau938c7fe2014-04-25 14:21:39 +02001802tcp-check connect - - X X
1803tcp-check expect - - X X
1804tcp-check send - - X X
1805tcp-check send-binary - - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02001806tcp-request connection - X X -
1807tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02001808tcp-request inspect-delay - X X X
Emeric Brun0a3b67f2010-09-24 15:34:53 +02001809tcp-response content - - X X
1810tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001811timeout check X - X X
1812timeout client X X X -
Willy Tarreau05cdd962014-05-10 14:30:07 +02001813timeout client-fin X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001814timeout clitimeout (deprecated) X X X -
1815timeout connect X - X X
1816timeout contimeout (deprecated) X - X X
1817timeout http-keep-alive X X X X
1818timeout http-request X X X X
1819timeout queue X - X X
1820timeout server X - X X
Willy Tarreau05cdd962014-05-10 14:30:07 +02001821timeout server-fin X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001822timeout srvtimeout (deprecated) X - X X
1823timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02001824timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001825transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01001826unique-id-format X X X -
1827unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001828use_backend - X X -
Willy Tarreau4a5cade2012-04-05 21:09:48 +02001829use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001830------------------------------------+----------+----------+---------+---------
1831 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02001832
Willy Tarreau0ba27502007-12-24 16:55:16 +01001833
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018344.2. Alphabetically sorted keywords reference
1835---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001836
1837This section provides a description of each keyword and its usage.
1838
1839
1840acl <aclname> <criterion> [flags] [operator] <value> ...
1841 Declare or complete an access list.
1842 May be used in sections : defaults | frontend | listen | backend
1843 no | yes | yes | yes
1844 Example:
1845 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
1846 acl invalid_src src_port 0:1023
1847 acl local_dst hdr(host) -i localhost
1848
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001849 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001850
1851
Cyril Bontéb21570a2009-11-29 20:04:48 +01001852appsession <cookie> len <length> timeout <holdtime>
1853 [request-learn] [prefix] [mode <path-parameters|query-string>]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001854 Define session stickiness on an existing application cookie.
1855 May be used in sections : defaults | frontend | listen | backend
1856 no | no | yes | yes
1857 Arguments :
1858 <cookie> this is the name of the cookie used by the application and which
1859 HAProxy will have to learn for each new session.
1860
Cyril Bontéb21570a2009-11-29 20:04:48 +01001861 <length> this is the max number of characters that will be memorized and
Willy Tarreau0ba27502007-12-24 16:55:16 +01001862 checked in each cookie value.
1863
1864 <holdtime> this is the time after which the cookie will be removed from
1865 memory if unused. If no unit is specified, this time is in
1866 milliseconds.
1867
Cyril Bontébf47aeb2009-10-15 00:15:40 +02001868 request-learn
1869 If this option is specified, then haproxy will be able to learn
1870 the cookie found in the request in case the server does not
1871 specify any in response. This is typically what happens with
1872 PHPSESSID cookies, or when haproxy's session expires before
1873 the application's session and the correct server is selected.
1874 It is recommended to specify this option to improve reliability.
1875
Cyril Bontéb21570a2009-11-29 20:04:48 +01001876 prefix When this option is specified, haproxy will match on the cookie
1877 prefix (or URL parameter prefix). The appsession value is the
1878 data following this prefix.
1879
1880 Example :
1881 appsession ASPSESSIONID len 64 timeout 3h prefix
1882
1883 This will match the cookie ASPSESSIONIDXXXX=XXXXX,
1884 the appsession value will be XXXX=XXXXX.
1885
1886 mode This option allows to change the URL parser mode.
1887 2 modes are currently supported :
1888 - path-parameters :
1889 The parser looks for the appsession in the path parameters
1890 part (each parameter is separated by a semi-colon), which is
1891 convenient for JSESSIONID for example.
1892 This is the default mode if the option is not set.
1893 - query-string :
1894 In this mode, the parser will look for the appsession in the
1895 query string.
1896
Willy Tarreau294d0f02015-08-10 19:40:12 +02001897 As of version 1.6, appsessions was removed. It is more flexible and more
1898 convenient to use stick-tables instead, and stick-tables support multi-master
1899 replication and data conservation across reloads, which appsessions did not.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001900
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01001901 See also : "cookie", "capture cookie", "balance", "stick", "stick-table",
1902 "ignore-persist", "nbproc" and "bind-process".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001903
1904
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01001905backlog <conns>
1906 Give hints to the system about the approximate listen backlog desired size
1907 May be used in sections : defaults | frontend | listen | backend
1908 yes | yes | yes | no
1909 Arguments :
1910 <conns> is the number of pending connections. Depending on the operating
1911 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02001912 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01001913
1914 In order to protect against SYN flood attacks, one solution is to increase
1915 the system's SYN backlog size. Depending on the system, sometimes it is just
1916 tunable via a system parameter, sometimes it is not adjustable at all, and
1917 sometimes the system relies on hints given by the application at the time of
1918 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
1919 to the listen() syscall. On systems which can make use of this value, it can
1920 sometimes be useful to be able to specify a different value, hence this
1921 backlog parameter.
1922
1923 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
1924 used as a hint and the system accepts up to the smallest greater power of
1925 two, and never more than some limits (usually 32768).
1926
1927 See also : "maxconn" and the target operating system's tuning guide.
1928
1929
Willy Tarreau0ba27502007-12-24 16:55:16 +01001930balance <algorithm> [ <arguments> ]
Willy Tarreau226071e2014-04-10 11:55:45 +02001931balance url_param <param> [check_post]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001932 Define the load balancing algorithm to be used in a backend.
1933 May be used in sections : defaults | frontend | listen | backend
1934 yes | no | yes | yes
1935 Arguments :
1936 <algorithm> is the algorithm used to select a server when doing load
1937 balancing. This only applies when no persistence information
1938 is available, or when a connection is redispatched to another
1939 server. <algorithm> may be one of the following :
1940
1941 roundrobin Each server is used in turns, according to their weights.
1942 This is the smoothest and fairest algorithm when the server's
1943 processing time remains equally distributed. This algorithm
1944 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02001945 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08001946 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02001947 large farms, when a server becomes up after having been down
1948 for a very short time, it may sometimes take a few hundreds
1949 requests for it to be re-integrated into the farm and start
1950 receiving traffic. This is normal, though very rare. It is
1951 indicated here in case you would have the chance to observe
1952 it, so that you don't worry.
1953
1954 static-rr Each server is used in turns, according to their weights.
1955 This algorithm is as similar to roundrobin except that it is
1956 static, which means that changing a server's weight on the
1957 fly will have no effect. On the other hand, it has no design
1958 limitation on the number of servers, and when a server goes
1959 up, it is always immediately reintroduced into the farm, once
1960 the full map is recomputed. It also uses slightly less CPU to
1961 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01001962
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01001963 leastconn The server with the lowest number of connections receives the
1964 connection. Round-robin is performed within groups of servers
1965 of the same load to ensure that all servers will be used. Use
1966 of this algorithm is recommended where very long sessions are
1967 expected, such as LDAP, SQL, TSE, etc... but is not very well
1968 suited for protocols using short sessions such as HTTP. This
1969 algorithm is dynamic, which means that server weights may be
1970 adjusted on the fly for slow starts for instance.
1971
Willy Tarreauf09c6602012-02-13 17:12:08 +01001972 first The first server with available connection slots receives the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001973 connection. The servers are chosen from the lowest numeric
Willy Tarreauf09c6602012-02-13 17:12:08 +01001974 identifier to the highest (see server parameter "id"), which
1975 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02001976 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01001977 not make sense to use this algorithm without setting maxconn.
1978 The purpose of this algorithm is to always use the smallest
1979 number of servers so that extra servers can be powered off
1980 during non-intensive hours. This algorithm ignores the server
1981 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02001982 or IMAP than HTTP, though it can be useful there too. In
1983 order to use this algorithm efficiently, it is recommended
1984 that a cloud controller regularly checks server usage to turn
1985 them off when unused, and regularly checks backend queue to
1986 turn new servers on when the queue inflates. Alternatively,
1987 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01001988
Willy Tarreau0ba27502007-12-24 16:55:16 +01001989 source The source IP address is hashed and divided by the total
1990 weight of the running servers to designate which server will
1991 receive the request. This ensures that the same client IP
1992 address will always reach the same server as long as no
1993 server goes down or up. If the hash result changes due to the
1994 number of running servers changing, many clients will be
1995 directed to a different server. This algorithm is generally
1996 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001997 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01001998 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001999 static by default, which means that changing a server's
2000 weight on the fly will have no effect, but this can be
2001 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002002
Oskar Stolc8dc41842012-05-19 10:19:54 +01002003 uri This algorithm hashes either the left part of the URI (before
2004 the question mark) or the whole URI (if the "whole" parameter
2005 is present) and divides the hash value by the total weight of
2006 the running servers. The result designates which server will
2007 receive the request. This ensures that the same URI will
2008 always be directed to the same server as long as no server
2009 goes up or down. This is used with proxy caches and
2010 anti-virus proxies in order to maximize the cache hit rate.
2011 Note that this algorithm may only be used in an HTTP backend.
2012 This algorithm is static by default, which means that
2013 changing a server's weight on the fly will have no effect,
2014 but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002015
Oskar Stolc8dc41842012-05-19 10:19:54 +01002016 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002017 "depth", both followed by a positive integer number. These
2018 options may be helpful when it is needed to balance servers
2019 based on the beginning of the URI only. The "len" parameter
2020 indicates that the algorithm should only consider that many
2021 characters at the beginning of the URI to compute the hash.
2022 Note that having "len" set to 1 rarely makes sense since most
2023 URIs start with a leading "/".
2024
2025 The "depth" parameter indicates the maximum directory depth
2026 to be used to compute the hash. One level is counted for each
2027 slash in the request. If both parameters are specified, the
2028 evaluation stops when either is reached.
2029
Willy Tarreau0ba27502007-12-24 16:55:16 +01002030 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002031 the query string of each HTTP GET request.
2032
2033 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02002034 request entity will be searched for the parameter argument,
2035 when it is not found in a query string after a question mark
Willy Tarreau226071e2014-04-10 11:55:45 +02002036 ('?') in the URL. The message body will only start to be
2037 analyzed once either the advertised amount of data has been
2038 received or the request buffer is full. In the unlikely event
2039 that chunked encoding is used, only the first chunk is
Cyril Bontédc4d9032012-04-08 21:57:39 +02002040 scanned. Parameter values separated by a chunk boundary, may
Willy Tarreau226071e2014-04-10 11:55:45 +02002041 be randomly balanced if at all. This keyword used to support
2042 an optional <max_wait> parameter which is now ignored.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002043
2044 If the parameter is found followed by an equal sign ('=') and
2045 a value, then the value is hashed and divided by the total
2046 weight of the running servers. The result designates which
2047 server will receive the request.
2048
2049 This is used to track user identifiers in requests and ensure
2050 that a same user ID will always be sent to the same server as
2051 long as no server goes up or down. If no value is found or if
2052 the parameter is not found, then a round robin algorithm is
2053 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002054 backend. This algorithm is static by default, which means
2055 that changing a server's weight on the fly will have no
2056 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002057
Cyril Bontédc4d9032012-04-08 21:57:39 +02002058 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
2059 request. Just as with the equivalent ACL 'hdr()' function,
2060 the header name in parenthesis is not case sensitive. If the
2061 header is absent or if it does not contain any value, the
2062 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01002063
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002064 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01002065 reducing the hash algorithm to the main domain part with some
2066 specific headers such as 'Host'. For instance, in the Host
2067 value "haproxy.1wt.eu", only "1wt" will be considered.
2068
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002069 This algorithm is static by default, which means that
2070 changing a server's weight on the fly will have no effect,
2071 but this can be changed using "hash-type".
2072
Emeric Brun736aa232009-06-30 17:56:00 +02002073 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02002074 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02002075 The RDP cookie <name> (or "mstshash" if omitted) will be
2076 looked up and hashed for each incoming TCP request. Just as
2077 with the equivalent ACL 'req_rdp_cookie()' function, the name
2078 is not case-sensitive. This mechanism is useful as a degraded
2079 persistence mode, as it makes it possible to always send the
2080 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002081 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02002082 used instead.
2083
2084 Note that for this to work, the frontend must ensure that an
2085 RDP cookie is already present in the request buffer. For this
2086 you must use 'tcp-request content accept' rule combined with
2087 a 'req_rdp_cookie_cnt' ACL.
2088
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002089 This algorithm is static by default, which means that
2090 changing a server's weight on the fly will have no effect,
2091 but this can be changed using "hash-type".
2092
Cyril Bontédc4d9032012-04-08 21:57:39 +02002093 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09002094
Willy Tarreau0ba27502007-12-24 16:55:16 +01002095 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002096 algorithms. Right now, only "url_param" and "uri" support an
2097 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002098
Willy Tarreau3cd9af22009-03-15 14:06:41 +01002099 The load balancing algorithm of a backend is set to roundrobin when no other
2100 algorithm, mode nor option have been set. The algorithm may only be set once
2101 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002102
2103 Examples :
2104 balance roundrobin
2105 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002106 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01002107 balance hdr(User-Agent)
2108 balance hdr(host)
2109 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002110
2111 Note: the following caveats and limitations on using the "check_post"
2112 extension with "url_param" must be considered :
2113
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002114 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002115 to determine if the parameters will be found in the body or entity which
2116 may contain binary data. Therefore another method may be required to
2117 restrict consideration of POST requests that have no URL parameters in
2118 the body. (see acl reqideny http_end)
2119
2120 - using a <max_wait> value larger than the request buffer size does not
2121 make sense and is useless. The buffer size is set at build time, and
2122 defaults to 16 kB.
2123
2124 - Content-Encoding is not supported, the parameter search will probably
2125 fail; and load balancing will fall back to Round Robin.
2126
2127 - Expect: 100-continue is not supported, load balancing will fall back to
2128 Round Robin.
2129
2130 - Transfer-Encoding (RFC2616 3.6.1) is only supported in the first chunk.
2131 If the entire parameter value is not present in the first chunk, the
2132 selection of server is undefined (actually, defined by how little
2133 actually appeared in the first chunk).
2134
2135 - This feature does not support generation of a 100, 411 or 501 response.
2136
2137 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002138 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002139 white space or control characters are found, indicating the end of what
2140 might be a URL parameter list. This is probably not a concern with SGML
2141 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002142
Willy Tarreau294d0f02015-08-10 19:40:12 +02002143 See also : "dispatch", "cookie", "transparent", "hash-type" and "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002144
2145
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002146bind [<address>]:<port_range> [, ...] [param*]
2147bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002148 Define one or several listening addresses and/or ports in a frontend.
2149 May be used in sections : defaults | frontend | listen | backend
2150 no | yes | yes | no
2151 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002152 <address> is optional and can be a host name, an IPv4 address, an IPv6
2153 address, or '*'. It designates the address the frontend will
2154 listen on. If unset, all IPv4 addresses of the system will be
2155 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01002156 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01002157 Optionally, an address family prefix may be used before the
2158 address to force the family regardless of the address format,
2159 which can be useful to specify a path to a unix socket with
2160 no slash ('/'). Currently supported prefixes are :
2161 - 'ipv4@' -> address is always IPv4
2162 - 'ipv6@' -> address is always IPv6
2163 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau70f72e02014-07-08 00:37:50 +02002164 - 'abns@' -> address is in abstract namespace (Linux only).
2165 Note: since abstract sockets are not "rebindable", they
2166 do not cope well with multi-process mode during
2167 soft-restart, so it is better to avoid them if
2168 nbproc is greater than 1. The effect is that if the
2169 new process fails to start, only one of the old ones
2170 will be able to rebind to the socket.
Willy Tarreau40aa0702013-03-10 23:51:38 +01002171 - 'fd@<n>' -> use file descriptor <n> inherited from the
2172 parent. The fd must be bound and may or may not already
2173 be listening.
William Lallemandb2f07452015-05-12 14:27:13 +02002174 You may want to reference some environment variables in the
2175 address parameter, see section 2.3 about environment
2176 variables.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002177
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002178 <port_range> is either a unique TCP port, or a port range for which the
2179 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002180 above. The port is mandatory for TCP listeners. Note that in
2181 the case of an IPv6 address, the port is always the number
2182 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002183 - a numerical port (ex: '80')
2184 - a dash-delimited ports range explicitly stating the lower
2185 and upper bounds (ex: '2000-2100') which are included in
2186 the range.
2187
2188 Particular care must be taken against port ranges, because
2189 every <address:port> couple consumes one socket (= a file
2190 descriptor), so it's easy to consume lots of descriptors
2191 with a simple range, and to run out of sockets. Also, each
2192 <address:port> couple must be used only once among all
2193 instances running on a same system. Please note that binding
2194 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002195 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002196 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002197
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002198 <path> is a UNIX socket path beginning with a slash ('/'). This is
2199 alternative to the TCP listening port. Haproxy will then
2200 receive UNIX connections on the socket located at this place.
2201 The path must begin with a slash and by default is absolute.
2202 It can be relative to the prefix defined by "unix-bind" in
2203 the global section. Note that the total length of the prefix
2204 followed by the socket path cannot exceed some system limits
2205 for UNIX sockets, which commonly are set to 107 characters.
2206
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002207 <param*> is a list of parameters common to all sockets declared on the
2208 same line. These numerous parameters depend on OS and build
2209 options and have a complete section dedicated to them. Please
2210 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002211
Willy Tarreau0ba27502007-12-24 16:55:16 +01002212 It is possible to specify a list of address:port combinations delimited by
2213 commas. The frontend will then listen on all of these addresses. There is no
2214 fixed limit to the number of addresses and ports which can be listened on in
2215 a frontend, as well as there is no limit to the number of "bind" statements
2216 in a frontend.
2217
2218 Example :
2219 listen http_proxy
2220 bind :80,:443
2221 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002222 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01002223
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002224 listen http_https_proxy
2225 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02002226 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002227
Willy Tarreau24709282013-03-10 21:32:12 +01002228 listen http_https_proxy_explicit
2229 bind ipv6@:80
2230 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
2231 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
2232
Willy Tarreaudad36a32013-03-11 01:20:04 +01002233 listen external_bind_app1
William Lallemandb2f07452015-05-12 14:27:13 +02002234 bind "fd@${FD_APP1}"
Willy Tarreaudad36a32013-03-11 01:20:04 +01002235
Willy Tarreau55dcaf62015-09-27 15:03:15 +02002236 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
2237 sun_path length is used for the address length. Some other programs
2238 such as socat use the string length only by default. Pass the option
2239 ",unix-tightsocklen=0" to any abstract socket definition in socat to
2240 make it compatible with HAProxy's.
2241
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002242 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002243 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002244
2245
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002246bind-process [ all | odd | even | <number 1-64>[-<number 1-64>] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002247 Limit visibility of an instance to a certain set of processes numbers.
2248 May be used in sections : defaults | frontend | listen | backend
2249 yes | yes | yes | yes
2250 Arguments :
2251 all All process will see this instance. This is the default. It
2252 may be used to override a default value.
2253
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002254 odd This instance will be enabled on processes 1,3,5,...63. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002255 option may be combined with other numbers.
2256
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002257 even This instance will be enabled on processes 2,4,6,...64. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002258 option may be combined with other numbers. Do not use it
2259 with less than 2 processes otherwise some instances might be
2260 missing from all processes.
2261
Willy Tarreau110ecc12012-11-15 17:50:01 +01002262 number The instance will be enabled on this process number or range,
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002263 whose values must all be between 1 and 32 or 64 depending on
Willy Tarreau102df612014-05-07 23:56:38 +02002264 the machine's word size. If a proxy is bound to process
2265 numbers greater than the configured global.nbproc, it will
2266 either be forced to process #1 if a single process was
2267 specified, or to all processes otherwise.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002268
2269 This keyword limits binding of certain instances to certain processes. This
2270 is useful in order not to have too many processes listening to the same
2271 ports. For instance, on a dual-core machine, it might make sense to set
2272 'nbproc 2' in the global section, then distributes the listeners among 'odd'
2273 and 'even' instances.
2274
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002275 At the moment, it is not possible to reference more than 32 or 64 processes
2276 using this keyword, but this should be more than enough for most setups.
2277 Please note that 'all' really means all processes regardless of the machine's
2278 word size, and is not limited to the first 32 or 64.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002279
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002280 Each "bind" line may further be limited to a subset of the proxy's processes,
2281 please consult the "process" bind keyword in section 5.1.
2282
Willy Tarreaub369a042014-09-16 13:21:03 +02002283 When a frontend has no explicit "bind-process" line, it tries to bind to all
2284 the processes referenced by its "bind" lines. That means that frontends can
2285 easily adapt to their listeners' processes.
2286
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002287 If some backends are referenced by frontends bound to other processes, the
2288 backend automatically inherits the frontend's processes.
2289
2290 Example :
2291 listen app_ip1
2292 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002293 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002294
2295 listen app_ip2
2296 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002297 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002298
2299 listen management
2300 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002301 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002302
Willy Tarreau110ecc12012-11-15 17:50:01 +01002303 listen management
2304 bind 10.0.0.4:80
2305 bind-process 1-4
2306
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002307 See also : "nbproc" in global section, and "process" in section 5.1.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002308
2309
Willy Tarreau0ba27502007-12-24 16:55:16 +01002310block { if | unless } <condition>
2311 Block a layer 7 request if/unless a condition is matched
2312 May be used in sections : defaults | frontend | listen | backend
2313 no | yes | yes | yes
2314
2315 The HTTP request will be blocked very early in the layer 7 processing
2316 if/unless <condition> is matched. A 403 error will be returned if the request
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002317 is blocked. The condition has to reference ACLs (see section 7). This is
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02002318 typically used to deny access to certain sensitive resources if some
Willy Tarreau0ba27502007-12-24 16:55:16 +01002319 conditions are met or not met. There is no fixed limit to the number of
2320 "block" statements per instance.
2321
2322 Example:
2323 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2324 acl invalid_src src_port 0:1023
2325 acl local_dst hdr(host) -i localhost
2326 block if invalid_src || local_dst
2327
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002328 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002329
2330
2331capture cookie <name> len <length>
2332 Capture and log a cookie in the request and in the response.
2333 May be used in sections : defaults | frontend | listen | backend
2334 no | yes | yes | no
2335 Arguments :
2336 <name> is the beginning of the name of the cookie to capture. In order
2337 to match the exact name, simply suffix the name with an equal
2338 sign ('='). The full name will appear in the logs, which is
2339 useful with application servers which adjust both the cookie name
2340 and value (eg: ASPSESSIONXXXXX).
2341
2342 <length> is the maximum number of characters to report in the logs, which
2343 include the cookie name, the equal sign and the value, all in the
2344 standard "name=value" form. The string will be truncated on the
2345 right if it exceeds <length>.
2346
2347 Only the first cookie is captured. Both the "cookie" request headers and the
2348 "set-cookie" response headers are monitored. This is particularly useful to
2349 check for application bugs causing session crossing or stealing between
2350 users, because generally the user's cookies can only change on a login page.
2351
2352 When the cookie was not presented by the client, the associated log column
2353 will report "-". When a request does not cause a cookie to be assigned by the
2354 server, a "-" is reported in the response column.
2355
2356 The capture is performed in the frontend only because it is necessary that
2357 the log format does not change for a given frontend depending on the
2358 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01002359 "capture cookie" statement in a frontend. The maximum capture length is set
2360 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
2361 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002362
2363 Example:
2364 capture cookie ASPSESSION len 32
2365
2366 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002367 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002368
2369
2370capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01002371 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002372 May be used in sections : defaults | frontend | listen | backend
2373 no | yes | yes | no
2374 Arguments :
2375 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002376 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01002377 appear in the requests, with the first letter of each word in
2378 upper case. The header name will not appear in the logs, only the
2379 value is reported, but the position in the logs is respected.
2380
2381 <length> is the maximum number of characters to extract from the value and
2382 report in the logs. The string will be truncated on the right if
2383 it exceeds <length>.
2384
Willy Tarreau4460d032012-11-21 23:37:37 +01002385 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01002386 value will be added to the logs between braces ('{}'). If multiple headers
2387 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002388 in the same order they were declared in the configuration. Non-existent
2389 headers will be logged just as an empty string. Common uses for request
2390 header captures include the "Host" field in virtual hosting environments, the
2391 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002392 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002393 environments to find where the request came from.
2394
2395 Note that when capturing headers such as "User-agent", some spaces may be
2396 logged, making the log analysis more difficult. Thus be careful about what
2397 you log if you know your log parser is not smart enough to rely on the
2398 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002399
Willy Tarreau0900abb2012-11-22 00:21:46 +01002400 There is no limit to the number of captured request headers nor to their
2401 length, though it is wise to keep them low to limit memory usage per session.
2402 In order to keep log format consistent for a same frontend, header captures
2403 can only be declared in a frontend. It is not possible to specify a capture
2404 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002405
2406 Example:
2407 capture request header Host len 15
2408 capture request header X-Forwarded-For len 15
Cyril Bontéd1b0f7c2015-10-26 22:37:39 +01002409 capture request header Referer len 15
Willy Tarreau0ba27502007-12-24 16:55:16 +01002410
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002411 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01002412 about logging.
2413
2414
2415capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01002416 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002417 May be used in sections : defaults | frontend | listen | backend
2418 no | yes | yes | no
2419 Arguments :
2420 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002421 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01002422 appear in the response, with the first letter of each word in
2423 upper case. The header name will not appear in the logs, only the
2424 value is reported, but the position in the logs is respected.
2425
2426 <length> is the maximum number of characters to extract from the value and
2427 report in the logs. The string will be truncated on the right if
2428 it exceeds <length>.
2429
Willy Tarreau4460d032012-11-21 23:37:37 +01002430 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01002431 result will be added to the logs between braces ('{}') after the captured
2432 request headers. If multiple headers are captured, they will be delimited by
2433 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002434 the configuration. Non-existent headers will be logged just as an empty
2435 string. Common uses for response header captures include the "Content-length"
2436 header which indicates how many bytes are expected to be returned, the
2437 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002438
Willy Tarreau0900abb2012-11-22 00:21:46 +01002439 There is no limit to the number of captured response headers nor to their
2440 length, though it is wise to keep them low to limit memory usage per session.
2441 In order to keep log format consistent for a same frontend, header captures
2442 can only be declared in a frontend. It is not possible to specify a capture
2443 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002444
2445 Example:
2446 capture response header Content-length len 9
2447 capture response header Location len 15
2448
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002449 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01002450 about logging.
2451
2452
Cyril Bontéf0c60612010-02-06 14:44:47 +01002453clitimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002454 Set the maximum inactivity time on the client side.
2455 May be used in sections : defaults | frontend | listen | backend
2456 yes | yes | yes | no
2457 Arguments :
2458 <timeout> is the timeout value is specified in milliseconds by default, but
2459 can be in any other unit if the number is suffixed by the unit,
2460 as explained at the top of this document.
2461
2462 The inactivity timeout applies when the client is expected to acknowledge or
2463 send data. In HTTP mode, this timeout is particularly important to consider
2464 during the first phase, when the client sends the request, and during the
2465 response while it is reading data sent by the server. The value is specified
2466 in milliseconds by default, but can be in any other unit if the number is
2467 suffixed by the unit, as specified at the top of this document. In TCP mode
2468 (and to a lesser extent, in HTTP mode), it is highly recommended that the
2469 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002470 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01002471 losses by specifying timeouts that are slightly above multiples of 3 seconds
2472 (eg: 4 or 5 seconds).
2473
2474 This parameter is specific to frontends, but can be specified once for all in
2475 "defaults" sections. This is in fact one of the easiest solutions not to
2476 forget about it. An unspecified timeout results in an infinite timeout, which
2477 is not recommended. Such a usage is accepted and works but reports a warning
2478 during startup because it may results in accumulation of expired sessions in
2479 the system if the system's timeouts are not configured either.
2480
2481 This parameter is provided for compatibility but is currently deprecated.
2482 Please use "timeout client" instead.
2483
Willy Tarreau036fae02008-01-06 13:24:40 +01002484 See also : "timeout client", "timeout http-request", "timeout server", and
2485 "srvtimeout".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002486
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002487compression algo <algorithm> ...
2488compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02002489compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02002490 Enable HTTP compression.
2491 May be used in sections : defaults | frontend | listen | backend
2492 yes | yes | yes | yes
2493 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002494 algo is followed by the list of supported compression algorithms.
2495 type is followed by the list of MIME types that will be compressed.
2496 offload makes haproxy work as a compression offloader only (see notes).
2497
2498 The currently supported algorithms are :
Willy Tarreauc91840a2015-03-28 17:00:39 +01002499 identity this is mostly for debugging, and it was useful for developing
2500 the compression feature. Identity does not apply any change on
2501 data.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002502
Willy Tarreauc91840a2015-03-28 17:00:39 +01002503 gzip applies gzip compression. This setting is only available when
Baptiste Assmannf085d632015-12-21 17:57:32 +01002504 support for zlib or libslz was built in.
Willy Tarreauc91840a2015-03-28 17:00:39 +01002505
2506 deflate same as "gzip", but with deflate algorithm and zlib format.
2507 Note that this algorithm has ambiguous support on many
2508 browsers and no support at all from recent ones. It is
2509 strongly recommended not to use it for anything else than
2510 experimentation. This setting is only available when support
Baptiste Assmannf085d632015-12-21 17:57:32 +01002511 for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002512
Willy Tarreauc91840a2015-03-28 17:00:39 +01002513 raw-deflate same as "deflate" without the zlib wrapper, and used as an
2514 alternative when the browser wants "deflate". All major
2515 browsers understand it and despite violating the standards,
2516 it is known to work better than "deflate", at least on MSIE
2517 and some versions of Safari. Do not use it in conjunction
2518 with "deflate", use either one or the other since both react
2519 to the same Accept-Encoding token. This setting is only
Baptiste Assmannf085d632015-12-21 17:57:32 +01002520 available when support for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002521
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04002522 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002523 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04002524 If backend servers support HTTP compression, these directives
2525 will be no-op: haproxy will see the compressed response and will not
2526 compress again. If backend servers do not support HTTP compression and
2527 there is Accept-Encoding header in request, haproxy will compress the
2528 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02002529
2530 The "offload" setting makes haproxy remove the Accept-Encoding header to
2531 prevent backend servers from compressing responses. It is strongly
2532 recommended not to do this because this means that all the compression work
2533 will be done on the single point where haproxy is located. However in some
2534 deployment scenarios, haproxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04002535 with broken HTTP compression implementation which can't be turned off.
2536 In that case haproxy can be used to prevent that gateway from emitting
2537 invalid payloads. In this case, simply removing the header in the
2538 configuration does not work because it applies before the header is parsed,
2539 so that prevents haproxy from compressing. The "offload" setting should
Willy Tarreauffea9fd2014-07-12 16:37:02 +02002540 then be used for such scenarios. Note: for now, the "offload" setting is
2541 ignored when set in a defaults section.
William Lallemand82fe75c2012-10-23 10:25:10 +02002542
William Lallemand05097442012-11-20 12:14:28 +01002543 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002544 * the request does not advertise a supported compression algorithm in the
2545 "Accept-Encoding" header
2546 * the response message is not HTTP/1.1
William Lallemandd3002612012-11-26 14:34:47 +01002547 * HTTP status code is not 200
William Lallemand8bb4e342013-12-10 17:28:48 +01002548 * response header "Transfer-Encoding" contains "chunked" (Temporary
2549 Workaround)
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002550 * response contain neither a "Content-Length" header nor a
2551 "Transfer-Encoding" whose last value is "chunked"
2552 * response contains a "Content-Type" header whose first value starts with
2553 "multipart"
2554 * the response contains the "no-transform" value in the "Cache-control"
2555 header
2556 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
2557 and later
2558 * The response contains a "Content-Encoding" header, indicating that the
2559 response is already compressed (see compression offload)
William Lallemand05097442012-11-20 12:14:28 +01002560
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002561 Note: The compression does not rewrite Etag headers, and does not emit the
2562 Warning header.
William Lallemand05097442012-11-20 12:14:28 +01002563
William Lallemand82fe75c2012-10-23 10:25:10 +02002564 Examples :
2565 compression algo gzip
2566 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01002567
Cyril Bontéf0c60612010-02-06 14:44:47 +01002568contimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002569 Set the maximum time to wait for a connection attempt to a server to succeed.
2570 May be used in sections : defaults | frontend | listen | backend
2571 yes | no | yes | yes
2572 Arguments :
2573 <timeout> is the timeout value is specified in milliseconds by default, but
2574 can be in any other unit if the number is suffixed by the unit,
2575 as explained at the top of this document.
2576
2577 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002578 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01002579 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01002580 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
2581 connect timeout also presets the queue timeout to the same value if this one
2582 has not been specified. Historically, the contimeout was also used to set the
2583 tarpit timeout in a listen section, which is not possible in a pure frontend.
2584
2585 This parameter is specific to backends, but can be specified once for all in
2586 "defaults" sections. This is in fact one of the easiest solutions not to
2587 forget about it. An unspecified timeout results in an infinite timeout, which
2588 is not recommended. Such a usage is accepted and works but reports a warning
2589 during startup because it may results in accumulation of failed sessions in
2590 the system if the system's timeouts are not configured either.
2591
2592 This parameter is provided for backwards compatibility but is currently
2593 deprecated. Please use "timeout connect", "timeout queue" or "timeout tarpit"
2594 instead.
2595
2596 See also : "timeout connect", "timeout queue", "timeout tarpit",
2597 "timeout server", "contimeout".
2598
2599
Willy Tarreau55165fe2009-05-10 12:02:55 +02002600cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02002601 [ postonly ] [ preserve ] [ httponly ] [ secure ]
2602 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002603 Enable cookie-based persistence in a backend.
2604 May be used in sections : defaults | frontend | listen | backend
2605 yes | no | yes | yes
2606 Arguments :
2607 <name> is the name of the cookie which will be monitored, modified or
2608 inserted in order to bring persistence. This cookie is sent to
2609 the client via a "Set-Cookie" header in the response, and is
2610 brought back by the client in a "Cookie" header in all requests.
2611 Special care should be taken to choose a name which does not
2612 conflict with any likely application cookie. Also, if the same
2613 backends are subject to be used by the same clients (eg:
2614 HTTP/HTTPS), care should be taken to use different cookie names
2615 between all backends if persistence between them is not desired.
2616
2617 rewrite This keyword indicates that the cookie will be provided by the
2618 server and that haproxy will have to modify its value to set the
2619 server's identifier in it. This mode is handy when the management
2620 of complex combinations of "Set-cookie" and "Cache-control"
2621 headers is left to the application. The application can then
2622 decide whether or not it is appropriate to emit a persistence
2623 cookie. Since all responses should be monitored, this mode only
2624 works in HTTP close mode. Unless the application behaviour is
2625 very complex and/or broken, it is advised not to start with this
2626 mode for new deployments. This keyword is incompatible with
2627 "insert" and "prefix".
2628
2629 insert This keyword indicates that the persistence cookie will have to
Willy Tarreaua79094d2010-08-31 22:54:15 +02002630 be inserted by haproxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002631
Willy Tarreaua79094d2010-08-31 22:54:15 +02002632 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002633 server. When used without the "preserve" option, if the server
2634 emits a cookie with the same name, it will be remove before
2635 processing. For this reason, this mode can be used to upgrade
2636 existing configurations running in the "rewrite" mode. The cookie
2637 will only be a session cookie and will not be stored on the
2638 client's disk. By default, unless the "indirect" option is added,
2639 the server will see the cookies emitted by the client. Due to
2640 caching effects, it is generally wise to add the "nocache" or
2641 "postonly" keywords (see below). The "insert" keyword is not
2642 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002643
2644 prefix This keyword indicates that instead of relying on a dedicated
2645 cookie for the persistence, an existing one will be completed.
2646 This may be needed in some specific environments where the client
2647 does not support more than one single cookie and the application
2648 already needs it. In this case, whenever the server sets a cookie
2649 named <name>, it will be prefixed with the server's identifier
2650 and a delimiter. The prefix will be removed from all client
2651 requests so that the server still finds the cookie it emitted.
2652 Since all requests and responses are subject to being modified,
2653 this mode requires the HTTP close mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02002654 not compatible with "rewrite" and "insert". Note: it is highly
2655 recommended not to use "indirect" with "prefix", otherwise server
2656 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002657
Willy Tarreaua79094d2010-08-31 22:54:15 +02002658 indirect When this option is specified, no cookie will be emitted to a
2659 client which already has a valid one for the server which has
2660 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002661 it will be removed, unless the "preserve" option is also set. In
2662 "insert" mode, this will additionally remove cookies from the
2663 requests transmitted to the server, making the persistence
2664 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02002665 Note: it is highly recommended not to use "indirect" with
2666 "prefix", otherwise server cookie updates would not be sent to
2667 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002668
2669 nocache This option is recommended in conjunction with the insert mode
2670 when there is a cache between the client and HAProxy, as it
2671 ensures that a cacheable response will be tagged non-cacheable if
2672 a cookie needs to be inserted. This is important because if all
2673 persistence cookies are added on a cacheable home page for
2674 instance, then all customers will then fetch the page from an
2675 outer cache and will all share the same persistence cookie,
2676 leading to one server receiving much more traffic than others.
2677 See also the "insert" and "postonly" options.
2678
2679 postonly This option ensures that cookie insertion will only be performed
2680 on responses to POST requests. It is an alternative to the
2681 "nocache" option, because POST responses are not cacheable, so
2682 this ensures that the persistence cookie will never get cached.
2683 Since most sites do not need any sort of persistence before the
2684 first POST which generally is a login request, this is a very
2685 efficient method to optimize caching without risking to find a
2686 persistence cookie in the cache.
2687 See also the "insert" and "nocache" options.
2688
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002689 preserve This option may only be used with "insert" and/or "indirect". It
2690 allows the server to emit the persistence cookie itself. In this
2691 case, if a cookie is found in the response, haproxy will leave it
2692 untouched. This is useful in order to end persistence after a
2693 logout request for instance. For this, the server just has to
2694 emit a cookie with an invalid value (eg: empty) or with a date in
2695 the past. By combining this mechanism with the "disable-on-404"
2696 check option, it is possible to perform a completely graceful
2697 shutdown because users will definitely leave the server after
2698 they logout.
2699
Willy Tarreau4992dd22012-05-31 21:02:17 +02002700 httponly This option tells haproxy to add an "HttpOnly" cookie attribute
2701 when a cookie is inserted. This attribute is used so that a
2702 user agent doesn't share the cookie with non-HTTP components.
2703 Please check RFC6265 for more information on this attribute.
2704
2705 secure This option tells haproxy to add a "Secure" cookie attribute when
2706 a cookie is inserted. This attribute is used so that a user agent
2707 never emits this cookie over non-secure channels, which means
2708 that a cookie learned with this flag will be presented only over
2709 SSL/TLS connections. Please check RFC6265 for more information on
2710 this attribute.
2711
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002712 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002713 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01002714 name. If the domain begins with a dot, the browser is allowed to
2715 use it for any host ending with that name. It is also possible to
2716 specify several domain names by invoking this option multiple
2717 times. Some browsers might have small limits on the number of
2718 domains, so be careful when doing that. For the record, sending
2719 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002720
Willy Tarreau996a92c2010-10-13 19:30:47 +02002721 maxidle This option allows inserted cookies to be ignored after some idle
2722 time. It only works with insert-mode cookies. When a cookie is
2723 sent to the client, the date this cookie was emitted is sent too.
2724 Upon further presentations of this cookie, if the date is older
2725 than the delay indicated by the parameter (in seconds), it will
2726 be ignored. Otherwise, it will be refreshed if needed when the
2727 response is sent to the client. This is particularly useful to
2728 prevent users who never close their browsers from remaining for
2729 too long on the same server (eg: after a farm size change). When
2730 this option is set and a cookie has no date, it is always
2731 accepted, but gets refreshed in the response. This maintains the
2732 ability for admins to access their sites. Cookies that have a
2733 date in the future further than 24 hours are ignored. Doing so
2734 lets admins fix timezone issues without risking kicking users off
2735 the site.
2736
2737 maxlife This option allows inserted cookies to be ignored after some life
2738 time, whether they're in use or not. It only works with insert
2739 mode cookies. When a cookie is first sent to the client, the date
2740 this cookie was emitted is sent too. Upon further presentations
2741 of this cookie, if the date is older than the delay indicated by
2742 the parameter (in seconds), it will be ignored. If the cookie in
2743 the request has no date, it is accepted and a date will be set.
2744 Cookies that have a date in the future further than 24 hours are
2745 ignored. Doing so lets admins fix timezone issues without risking
2746 kicking users off the site. Contrary to maxidle, this value is
2747 not refreshed, only the first visit date counts. Both maxidle and
2748 maxlife may be used at the time. This is particularly useful to
2749 prevent users who never close their browsers from remaining for
2750 too long on the same server (eg: after a farm size change). This
2751 is stronger than the maxidle method in that it forces a
2752 redispatch after some absolute delay.
2753
Willy Tarreau0ba27502007-12-24 16:55:16 +01002754 There can be only one persistence cookie per HTTP backend, and it can be
2755 declared in a defaults section. The value of the cookie will be the value
2756 indicated after the "cookie" keyword in a "server" statement. If no cookie
2757 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002758
Willy Tarreau0ba27502007-12-24 16:55:16 +01002759 Examples :
2760 cookie JSESSIONID prefix
2761 cookie SRV insert indirect nocache
2762 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02002763 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01002764
Willy Tarreau294d0f02015-08-10 19:40:12 +02002765 See also : "balance source", "capture cookie", "server" and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002766
Willy Tarreau983e01e2010-01-11 18:42:06 +01002767
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02002768declare capture [ request | response ] len <length>
2769 Declares a capture slot.
2770 May be used in sections : defaults | frontend | listen | backend
2771 no | yes | yes | no
2772 Arguments:
2773 <length> is the length allowed for the capture.
2774
2775 This declaration is only available in the frontend or listen section, but the
2776 reserved slot can be used in the backends. The "request" keyword allocates a
2777 capture slot for use in the request, and "response" allocates a capture slot
2778 for use in the response.
2779
2780 See also: "capture-req", "capture-res" (sample converters),
Baptiste Assmann5ac425c2015-10-21 23:13:46 +02002781 "capture.req.hdr", "capture.res.hdr" (sample fetches),
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02002782 "http-request capture" and "http-response capture".
2783
2784
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002785default-server [param*]
2786 Change default options for a server in a backend
2787 May be used in sections : defaults | frontend | listen | backend
2788 yes | no | yes | yes
2789 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01002790 <param*> is a list of parameters for this server. The "default-server"
2791 keyword accepts an important number of options and has a complete
2792 section dedicated to it. Please refer to section 5 for more
2793 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002794
Willy Tarreau983e01e2010-01-11 18:42:06 +01002795 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002796 default-server inter 1000 weight 13
2797
2798 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01002799
Willy Tarreau983e01e2010-01-11 18:42:06 +01002800
Willy Tarreau0ba27502007-12-24 16:55:16 +01002801default_backend <backend>
2802 Specify the backend to use when no "use_backend" rule has been matched.
2803 May be used in sections : defaults | frontend | listen | backend
2804 yes | yes | yes | no
2805 Arguments :
2806 <backend> is the name of the backend to use.
2807
2808 When doing content-switching between frontend and backends using the
2809 "use_backend" keyword, it is often useful to indicate which backend will be
2810 used when no rule has matched. It generally is the dynamic backend which
2811 will catch all undetermined requests.
2812
Willy Tarreau0ba27502007-12-24 16:55:16 +01002813 Example :
2814
2815 use_backend dynamic if url_dyn
2816 use_backend static if url_css url_img extension_img
2817 default_backend dynamic
2818
Willy Tarreau98d04852015-05-26 12:18:29 +02002819 See also : "use_backend"
Willy Tarreau2769aa02007-12-27 18:26:09 +01002820
Willy Tarreau0ba27502007-12-24 16:55:16 +01002821
Baptiste Assmann27f51342013-10-09 06:51:49 +02002822description <string>
2823 Describe a listen, frontend or backend.
2824 May be used in sections : defaults | frontend | listen | backend
2825 no | yes | yes | yes
2826 Arguments : string
2827
2828 Allows to add a sentence to describe the related object in the HAProxy HTML
2829 stats page. The description will be printed on the right of the object name
2830 it describes.
2831 No need to backslash spaces in the <string> arguments.
2832
2833
Willy Tarreau0ba27502007-12-24 16:55:16 +01002834disabled
2835 Disable a proxy, frontend or backend.
2836 May be used in sections : defaults | frontend | listen | backend
2837 yes | yes | yes | yes
2838 Arguments : none
2839
2840 The "disabled" keyword is used to disable an instance, mainly in order to
2841 liberate a listening port or to temporarily disable a service. The instance
2842 will still be created and its configuration will be checked, but it will be
2843 created in the "stopped" state and will appear as such in the statistics. It
2844 will not receive any traffic nor will it send any health-checks or logs. It
2845 is possible to disable many instances at once by adding the "disabled"
2846 keyword in a "defaults" section.
2847
2848 See also : "enabled"
2849
2850
Willy Tarreau5ce94572010-06-07 14:35:41 +02002851dispatch <address>:<port>
2852 Set a default server address
2853 May be used in sections : defaults | frontend | listen | backend
2854 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02002855 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02002856
2857 <address> is the IPv4 address of the default server. Alternatively, a
2858 resolvable hostname is supported, but this name will be resolved
2859 during start-up.
2860
2861 <ports> is a mandatory port specification. All connections will be sent
2862 to this port, and it is not permitted to use port offsets as is
2863 possible with normal servers.
2864
Willy Tarreau787aed52011-04-15 06:45:37 +02002865 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02002866 server can take the connection. In the past it was used to forward non
2867 persistent connections to an auxiliary load balancer. Due to its simple
2868 syntax, it has also been used for simple TCP relays. It is recommended not to
2869 use it for more clarity, and to use the "server" directive instead.
2870
2871 See also : "server"
2872
2873
Willy Tarreau0ba27502007-12-24 16:55:16 +01002874enabled
2875 Enable a proxy, frontend or backend.
2876 May be used in sections : defaults | frontend | listen | backend
2877 yes | yes | yes | yes
2878 Arguments : none
2879
2880 The "enabled" keyword is used to explicitly enable an instance, when the
2881 defaults has been set to "disabled". This is very rarely used.
2882
2883 See also : "disabled"
2884
2885
2886errorfile <code> <file>
2887 Return a file contents instead of errors generated by HAProxy
2888 May be used in sections : defaults | frontend | listen | backend
2889 yes | yes | yes | yes
2890 Arguments :
2891 <code> is the HTTP status code. Currently, HAProxy is capable of
CJ Ess108b1dd2015-04-07 12:03:37 -04002892 generating codes 200, 400, 403, 405, 408, 429, 500, 502, 503, and
2893 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002894
2895 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002896 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01002897 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01002898 error pages, and to use absolute paths, since files are read
2899 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002900
2901 It is important to understand that this keyword is not meant to rewrite
2902 errors returned by the server, but errors detected and returned by HAProxy.
2903 This is why the list of supported errors is limited to a small set.
2904
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002905 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2906
Willy Tarreau0ba27502007-12-24 16:55:16 +01002907 The files are returned verbatim on the TCP socket. This allows any trick such
2908 as redirections to another URL or site, as well as tricks to clean cookies,
2909 force enable or disable caching, etc... The package provides default error
2910 files returning the same contents as default errors.
2911
Willy Tarreau59140a22009-02-22 12:02:30 +01002912 The files should not exceed the configured buffer size (BUFSIZE), which
2913 generally is 8 or 16 kB, otherwise they will be truncated. It is also wise
2914 not to put any reference to local contents (eg: images) in order to avoid
2915 loops between the client and HAProxy when all servers are down, causing an
2916 error to be returned instead of an image. For better HTTP compliance, it is
2917 recommended that all header lines end with CR-LF and not LF alone.
2918
Willy Tarreau0ba27502007-12-24 16:55:16 +01002919 The files are read at the same time as the configuration and kept in memory.
2920 For this reason, the errors continue to be returned even when the process is
2921 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01002922 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01002923 403 status code and interrogating a blocked URL.
2924
2925 See also : "errorloc", "errorloc302", "errorloc303"
2926
Willy Tarreau59140a22009-02-22 12:02:30 +01002927 Example :
2928 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
Willy Tarreau2705a612014-05-23 17:38:34 +02002929 errorfile 408 /dev/null # workaround Chrome pre-connect bug
Willy Tarreau59140a22009-02-22 12:02:30 +01002930 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
2931 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
2932
Willy Tarreau2769aa02007-12-27 18:26:09 +01002933
2934errorloc <code> <url>
2935errorloc302 <code> <url>
2936 Return an HTTP redirection to a URL instead of errors generated by HAProxy
2937 May be used in sections : defaults | frontend | listen | backend
2938 yes | yes | yes | yes
2939 Arguments :
2940 <code> is the HTTP status code. Currently, HAProxy is capable of
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002941 generating codes 200, 400, 403, 408, 500, 502, 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002942
2943 <url> it is the exact contents of the "Location" header. It may contain
2944 either a relative URI to an error page hosted on the same site,
2945 or an absolute URI designating an error page on another site.
2946 Special care should be given to relative URIs to avoid redirect
2947 loops if the URI itself may generate the same error (eg: 500).
2948
2949 It is important to understand that this keyword is not meant to rewrite
2950 errors returned by the server, but errors detected and returned by HAProxy.
2951 This is why the list of supported errors is limited to a small set.
2952
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002953 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2954
Willy Tarreau2769aa02007-12-27 18:26:09 +01002955 Note that both keyword return the HTTP 302 status code, which tells the
2956 client to fetch the designated URL using the same HTTP method. This can be
2957 quite problematic in case of non-GET methods such as POST, because the URL
2958 sent to the client might not be allowed for something other than GET. To
2959 workaround this problem, please use "errorloc303" which send the HTTP 303
2960 status code, indicating to the client that the URL must be fetched with a GET
2961 request.
2962
2963 See also : "errorfile", "errorloc303"
2964
2965
2966errorloc303 <code> <url>
2967 Return an HTTP redirection to a URL instead of errors generated by HAProxy
2968 May be used in sections : defaults | frontend | listen | backend
2969 yes | yes | yes | yes
2970 Arguments :
2971 <code> is the HTTP status code. Currently, HAProxy is capable of
2972 generating codes 400, 403, 408, 500, 502, 503, and 504.
2973
2974 <url> it is the exact contents of the "Location" header. It may contain
2975 either a relative URI to an error page hosted on the same site,
2976 or an absolute URI designating an error page on another site.
2977 Special care should be given to relative URIs to avoid redirect
2978 loops if the URI itself may generate the same error (eg: 500).
2979
2980 It is important to understand that this keyword is not meant to rewrite
2981 errors returned by the server, but errors detected and returned by HAProxy.
2982 This is why the list of supported errors is limited to a small set.
2983
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002984 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2985
Willy Tarreau2769aa02007-12-27 18:26:09 +01002986 Note that both keyword return the HTTP 303 status code, which tells the
2987 client to fetch the designated URL using the same HTTP GET method. This
2988 solves the usual problems associated with "errorloc" and the 302 code. It is
2989 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002990 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002991
2992 See also : "errorfile", "errorloc", "errorloc302"
2993
2994
Simon Horman51a1cf62015-02-03 13:00:44 +09002995email-alert from <emailaddr>
2996 Declare the from email address to be used in both the envelope and header
2997 of email alerts. This is the address that email alerts are sent from.
2998 May be used in sections: defaults | frontend | listen | backend
2999 yes | yes | yes | yes
3000
3001 Arguments :
3002
3003 <emailaddr> is the from email address to use when sending email alerts
3004
3005 Also requires "email-alert mailers" and "email-alert to" to be set
3006 and if so sending email alerts is enabled for the proxy.
3007
Simon Horman64e34162015-02-06 11:11:57 +09003008 See also : "email-alert level", "email-alert mailers",
Cyril Bonté307ee1e2015-09-28 23:16:06 +02003009 "email-alert myhostname", "email-alert to", section 3.6 about
3010 mailers.
Simon Horman64e34162015-02-06 11:11:57 +09003011
3012
3013email-alert level <level>
3014 Declare the maximum log level of messages for which email alerts will be
3015 sent. This acts as a filter on the sending of email alerts.
3016 May be used in sections: defaults | frontend | listen | backend
3017 yes | yes | yes | yes
3018
3019 Arguments :
3020
3021 <level> One of the 8 syslog levels:
3022 emerg alert crit err warning notice info debug
3023 The above syslog levels are ordered from lowest to highest.
3024
3025 By default level is alert
3026
3027 Also requires "email-alert from", "email-alert mailers" and
3028 "email-alert to" to be set and if so sending email alerts is enabled
3029 for the proxy.
3030
Simon Horman1421e212015-04-30 13:10:35 +09003031 Alerts are sent when :
3032
3033 * An un-paused server is marked as down and <level> is alert or lower
3034 * A paused server is marked as down and <level> is notice or lower
3035 * A server is marked as up or enters the drain state and <level>
3036 is notice or lower
3037 * "option log-health-checks" is enabled, <level> is info or lower,
3038 and a health check status update occurs
3039
Simon Horman64e34162015-02-06 11:11:57 +09003040 See also : "email-alert from", "email-alert mailers",
3041 "email-alert myhostname", "email-alert to",
Simon Horman51a1cf62015-02-03 13:00:44 +09003042 section 3.6 about mailers.
3043
3044
3045email-alert mailers <mailersect>
3046 Declare the mailers to be used when sending email alerts
3047 May be used in sections: defaults | frontend | listen | backend
3048 yes | yes | yes | yes
3049
3050 Arguments :
3051
3052 <mailersect> is the name of the mailers section to send email alerts.
3053
3054 Also requires "email-alert from" and "email-alert to" to be set
3055 and if so sending email alerts is enabled for the proxy.
3056
Simon Horman64e34162015-02-06 11:11:57 +09003057 See also : "email-alert from", "email-alert level", "email-alert myhostname",
3058 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003059
3060
3061email-alert myhostname <hostname>
3062 Declare the to hostname address to be used when communicating with
3063 mailers.
3064 May be used in sections: defaults | frontend | listen | backend
3065 yes | yes | yes | yes
3066
3067 Arguments :
3068
Baptiste Assmann738bad92015-12-21 15:27:53 +01003069 <hostname> is the hostname to use when communicating with mailers
Simon Horman51a1cf62015-02-03 13:00:44 +09003070
3071 By default the systems hostname is used.
3072
3073 Also requires "email-alert from", "email-alert mailers" and
3074 "email-alert to" to be set and if so sending email alerts is enabled
3075 for the proxy.
3076
Simon Horman64e34162015-02-06 11:11:57 +09003077 See also : "email-alert from", "email-alert level", "email-alert mailers",
3078 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003079
3080
3081email-alert to <emailaddr>
3082 Declare both the recipent address in the envelope and to address in the
3083 header of email alerts. This is the address that email alerts are sent to.
3084 May be used in sections: defaults | frontend | listen | backend
3085 yes | yes | yes | yes
3086
3087 Arguments :
3088
3089 <emailaddr> is the to email address to use when sending email alerts
3090
3091 Also requires "email-alert mailers" and "email-alert to" to be set
3092 and if so sending email alerts is enabled for the proxy.
3093
Simon Horman64e34162015-02-06 11:11:57 +09003094 See also : "email-alert from", "email-alert level", "email-alert mailers",
Simon Horman51a1cf62015-02-03 13:00:44 +09003095 "email-alert myhostname", section 3.6 about mailers.
3096
3097
Willy Tarreau4de91492010-01-22 19:10:05 +01003098force-persist { if | unless } <condition>
3099 Declare a condition to force persistence on down servers
3100 May be used in sections: defaults | frontend | listen | backend
3101 no | yes | yes | yes
3102
3103 By default, requests are not dispatched to down servers. It is possible to
3104 force this using "option persist", but it is unconditional and redispatches
3105 to a valid server if "option redispatch" is set. That leaves with very little
3106 possibilities to force some requests to reach a server which is artificially
3107 marked down for maintenance operations.
3108
3109 The "force-persist" statement allows one to declare various ACL-based
3110 conditions which, when met, will cause a request to ignore the down status of
3111 a server and still try to connect to it. That makes it possible to start a
3112 server, still replying an error to the health checks, and run a specially
3113 configured browser to test the service. Among the handy methods, one could
3114 use a specific source IP address, or a specific cookie. The cookie also has
3115 the advantage that it can easily be added/removed on the browser from a test
3116 page. Once the service is validated, it is then possible to open the service
3117 to the world by returning a valid response to health checks.
3118
3119 The forced persistence is enabled when an "if" condition is met, or unless an
3120 "unless" condition is met. The final redispatch is always disabled when this
3121 is used.
3122
Cyril Bonté0d4bf012010-04-25 23:21:46 +02003123 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02003124 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01003125
3126
Willy Tarreau2769aa02007-12-27 18:26:09 +01003127fullconn <conns>
3128 Specify at what backend load the servers will reach their maxconn
3129 May be used in sections : defaults | frontend | listen | backend
3130 yes | no | yes | yes
3131 Arguments :
3132 <conns> is the number of connections on the backend which will make the
3133 servers use the maximal number of connections.
3134
Willy Tarreau198a7442008-01-17 12:05:32 +01003135 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01003136 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01003137 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01003138 load. The server will then always accept at least <minconn> connections,
3139 never more than <maxconn>, and the limit will be on the ramp between both
3140 values when the backend has less than <conns> concurrent connections. This
3141 makes it possible to limit the load on the servers during normal loads, but
3142 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003143 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003144
Willy Tarreaufbb78422011-06-05 15:38:35 +02003145 Since it's hard to get this value right, haproxy automatically sets it to
3146 10% of the sum of the maxconns of all frontends that may branch to this
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01003147 backend (based on "use_backend" and "default_backend" rules). That way it's
3148 safe to leave it unset. However, "use_backend" involving dynamic names are
3149 not counted since there is no way to know if they could match or not.
Willy Tarreaufbb78422011-06-05 15:38:35 +02003150
Willy Tarreau2769aa02007-12-27 18:26:09 +01003151 Example :
3152 # The servers will accept between 100 and 1000 concurrent connections each
3153 # and the maximum of 1000 will be reached when the backend reaches 10000
3154 # connections.
3155 backend dynamic
3156 fullconn 10000
3157 server srv1 dyn1:80 minconn 100 maxconn 1000
3158 server srv2 dyn2:80 minconn 100 maxconn 1000
3159
3160 See also : "maxconn", "server"
3161
3162
3163grace <time>
3164 Maintain a proxy operational for some time after a soft stop
3165 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01003166 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01003167 Arguments :
3168 <time> is the time (by default in milliseconds) for which the instance
3169 will remain operational with the frontend sockets still listening
3170 when a soft-stop is received via the SIGUSR1 signal.
3171
3172 This may be used to ensure that the services disappear in a certain order.
3173 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003174 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01003175 needed by the equipment to detect the failure.
3176
3177 Note that currently, there is very little benefit in using this parameter,
3178 and it may in fact complicate the soft-reconfiguration process more than
3179 simplify it.
3180
Willy Tarreau0ba27502007-12-24 16:55:16 +01003181
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003182hash-type <method> <function> <modifier>
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003183 Specify a method to use for mapping hashes to servers
3184 May be used in sections : defaults | frontend | listen | backend
3185 yes | no | yes | yes
3186 Arguments :
Bhaskar98634f02013-10-29 23:30:51 -04003187 <method> is the method used to select a server from the hash computed by
3188 the <function> :
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003189
Bhaskar98634f02013-10-29 23:30:51 -04003190 map-based the hash table is a static array containing all alive servers.
3191 The hashes will be very smooth, will consider weights, but
3192 will be static in that weight changes while a server is up
3193 will be ignored. This means that there will be no slow start.
3194 Also, since a server is selected by its position in the array,
3195 most mappings are changed when the server count changes. This
3196 means that when a server goes up or down, or when a server is
3197 added to a farm, most connections will be redistributed to
3198 different servers. This can be inconvenient with caches for
3199 instance.
Willy Tarreau798a39c2010-11-24 15:04:29 +01003200
Bhaskar98634f02013-10-29 23:30:51 -04003201 consistent the hash table is a tree filled with many occurrences of each
3202 server. The hash key is looked up in the tree and the closest
3203 server is chosen. This hash is dynamic, it supports changing
3204 weights while the servers are up, so it is compatible with the
3205 slow start feature. It has the advantage that when a server
3206 goes up or down, only its associations are moved. When a
3207 server is added to the farm, only a few part of the mappings
3208 are redistributed, making it an ideal method for caches.
3209 However, due to its principle, the distribution will never be
3210 very smooth and it may sometimes be necessary to adjust a
3211 server's weight or its ID to get a more balanced distribution.
3212 In order to get the same distribution on multiple load
3213 balancers, it is important that all servers have the exact
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003214 same IDs. Note: consistent hash uses sdbm and avalanche if no
3215 hash function is specified.
Bhaskar98634f02013-10-29 23:30:51 -04003216
3217 <function> is the hash function to be used :
3218
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003219 sdbm this function was created initially for sdbm (a public-domain
Bhaskar98634f02013-10-29 23:30:51 -04003220 reimplementation of ndbm) database library. It was found to do
3221 well in scrambling bits, causing better distribution of the keys
3222 and fewer splits. It also happens to be a good general hashing
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003223 function with good distribution, unless the total server weight
3224 is a multiple of 64, in which case applying the avalanche
3225 modifier may help.
Bhaskar98634f02013-10-29 23:30:51 -04003226
3227 djb2 this function was first proposed by Dan Bernstein many years ago
3228 on comp.lang.c. Studies have shown that for certain workload this
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003229 function provides a better distribution than sdbm. It generally
3230 works well with text-based inputs though it can perform extremely
3231 poorly with numeric-only input or when the total server weight is
3232 a multiple of 33, unless the avalanche modifier is also used.
3233
Willy Tarreaua0f42712013-11-14 14:30:35 +01003234 wt6 this function was designed for haproxy while testing other
3235 functions in the past. It is not as smooth as the other ones, but
3236 is much less sensible to the input data set or to the number of
3237 servers. It can make sense as an alternative to sdbm+avalanche or
3238 djb2+avalanche for consistent hashing or when hashing on numeric
3239 data such as a source IP address or a visitor identifier in a URL
3240 parameter.
3241
Willy Tarreau324f07f2015-01-20 19:44:50 +01003242 crc32 this is the most common CRC32 implementation as used in Ethernet,
3243 gzip, PNG, etc. It is slower than the other ones but may provide
3244 a better distribution or less predictable results especially when
3245 used on strings.
3246
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003247 <modifier> indicates an optional method applied after hashing the key :
3248
3249 avalanche This directive indicates that the result from the hash
3250 function above should not be used in its raw form but that
3251 a 4-byte full avalanche hash must be applied first. The
3252 purpose of this step is to mix the resulting bits from the
3253 previous hash in order to avoid any undesired effect when
3254 the input contains some limited values or when the number of
3255 servers is a multiple of one of the hash's components (64
3256 for SDBM, 33 for DJB2). Enabling avalanche tends to make the
3257 result less predictable, but it's also not as smooth as when
3258 using the original function. Some testing might be needed
3259 with some workloads. This hash is one of the many proposed
3260 by Bob Jenkins.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003261
Bhaskar98634f02013-10-29 23:30:51 -04003262 The default hash type is "map-based" and is recommended for most usages. The
3263 default function is "sdbm", the selection of a function should be based on
3264 the range of the values being hashed.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003265
3266 See also : "balance", "server"
3267
3268
Willy Tarreau0ba27502007-12-24 16:55:16 +01003269http-check disable-on-404
3270 Enable a maintenance mode upon HTTP/404 response to health-checks
3271 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01003272 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01003273 Arguments : none
3274
3275 When this option is set, a server which returns an HTTP code 404 will be
3276 excluded from further load-balancing, but will still receive persistent
3277 connections. This provides a very convenient method for Web administrators
3278 to perform a graceful shutdown of their servers. It is also important to note
3279 that a server which is detected as failed while it was in this mode will not
3280 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
3281 will immediately be reinserted into the farm. The status on the stats page
3282 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01003283 option only works in conjunction with the "httpchk" option. If this option
3284 is used with "http-check expect", then it has precedence over it so that 404
3285 responses will still be considered as soft-stop.
3286
3287 See also : "option httpchk", "http-check expect"
3288
3289
3290http-check expect [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003291 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01003292 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02003293 yes | no | yes | yes
Willy Tarreaubd741542010-03-16 18:46:54 +01003294 Arguments :
3295 <match> is a keyword indicating how to look for a specific pattern in the
3296 response. The keyword may be one of "status", "rstatus",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003297 "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01003298 exclamation mark ("!") to negate the match. Spaces are allowed
3299 between the exclamation mark and the keyword. See below for more
3300 details on the supported keywords.
3301
3302 <pattern> is the pattern to look for. It may be a string or a regular
3303 expression. If the pattern contains spaces, they must be escaped
3304 with the usual backslash ('\').
3305
3306 By default, "option httpchk" considers that response statuses 2xx and 3xx
3307 are valid, and that others are invalid. When "http-check expect" is used,
3308 it defines what is considered valid or invalid. Only one "http-check"
3309 statement is supported in a backend. If a server fails to respond or times
3310 out, the check obviously fails. The available matches are :
3311
3312 status <string> : test the exact string match for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003313 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003314 response's status code is exactly this string. If the
3315 "status" keyword is prefixed with "!", then the response
3316 will be considered invalid if the status code matches.
3317
3318 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003319 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003320 response's status code matches the expression. If the
3321 "rstatus" keyword is prefixed with "!", then the response
3322 will be considered invalid if the status code matches.
3323 This is mostly used to check for multiple codes.
3324
3325 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003326 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003327 response's body contains this exact string. If the
3328 "string" keyword is prefixed with "!", then the response
3329 will be considered invalid if the body contains this
3330 string. This can be used to look for a mandatory word at
3331 the end of a dynamic page, or to detect a failure when a
3332 specific error appears on the check page (eg: a stack
3333 trace).
3334
3335 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04003336 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01003337 response's body matches this expression. If the "rstring"
3338 keyword is prefixed with "!", then the response will be
3339 considered invalid if the body matches the expression.
3340 This can be used to look for a mandatory word at the end
3341 of a dynamic page, or to detect a failure when a specific
3342 error appears on the check page (eg: a stack trace).
3343
3344 It is important to note that the responses will be limited to a certain size
3345 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
3346 Thus, too large responses may not contain the mandatory pattern when using
3347 "string" or "rstring". If a large response is absolutely required, it is
3348 possible to change the default max size by setting the global variable.
3349 However, it is worth keeping in mind that parsing very large responses can
3350 waste some CPU cycles, especially when regular expressions are used, and that
3351 it is always better to focus the checks on smaller resources.
3352
Cyril Bonté32602d22015-01-30 00:07:07 +01003353 Also "http-check expect" doesn't support HTTP keep-alive. Keep in mind that it
3354 will automatically append a "Connection: close" header, meaning that this
3355 header should not be present in the request provided by "option httpchk".
3356
Willy Tarreaubd741542010-03-16 18:46:54 +01003357 Last, if "http-check expect" is combined with "http-check disable-on-404",
3358 then this last one has precedence when the server responds with 404.
3359
3360 Examples :
3361 # only accept status 200 as valid
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003362 http-check expect status 200
Willy Tarreaubd741542010-03-16 18:46:54 +01003363
3364 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003365 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01003366
3367 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003368 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01003369
3370 # check that we have a correct hexadecimal tag before /html
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01003371 http-check expect rstring <!--tag:[0-9a-f]*</html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01003372
Willy Tarreaubd741542010-03-16 18:46:54 +01003373 See also : "option httpchk", "http-check disable-on-404"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003374
3375
Willy Tarreauef781042010-01-27 11:53:01 +01003376http-check send-state
3377 Enable emission of a state header with HTTP health checks
3378 May be used in sections : defaults | frontend | listen | backend
3379 yes | no | yes | yes
3380 Arguments : none
3381
3382 When this option is set, haproxy will systematically send a special header
3383 "X-Haproxy-Server-State" with a list of parameters indicating to each server
3384 how they are seen by haproxy. This can be used for instance when a server is
3385 manipulated without access to haproxy and the operator needs to know whether
3386 haproxy still sees it up or not, or if the server is the last one in a farm.
3387
3388 The header is composed of fields delimited by semi-colons, the first of which
3389 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
3390 checks on the total number before transition, just as appears in the stats
3391 interface. Next headers are in the form "<variable>=<value>", indicating in
3392 no specific order some values available in the stats interface :
Joseph Lynch514061c2015-01-15 17:52:59 -08003393 - a variable "address", containing the address of the backend server.
3394 This corresponds to the <address> field in the server declaration. For
3395 unix domain sockets, it will read "unix".
3396
3397 - a variable "port", containing the port of the backend server. This
3398 corresponds to the <port> field in the server declaration. For unix
3399 domain sockets, it will read "unix".
3400
Willy Tarreauef781042010-01-27 11:53:01 +01003401 - a variable "name", containing the name of the backend followed by a slash
3402 ("/") then the name of the server. This can be used when a server is
3403 checked in multiple backends.
3404
3405 - a variable "node" containing the name of the haproxy node, as set in the
3406 global "node" variable, otherwise the system's hostname if unspecified.
3407
3408 - a variable "weight" indicating the weight of the server, a slash ("/")
3409 and the total weight of the farm (just counting usable servers). This
3410 helps to know if other servers are available to handle the load when this
3411 one fails.
3412
3413 - a variable "scur" indicating the current number of concurrent connections
3414 on the server, followed by a slash ("/") then the total number of
3415 connections on all servers of the same backend.
3416
3417 - a variable "qcur" indicating the current number of requests in the
3418 server's queue.
3419
3420 Example of a header received by the application server :
3421 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
3422 scur=13/22; qcur=0
3423
3424 See also : "option httpchk", "http-check disable-on-404"
3425
Willy Tarreauccbcc372012-12-27 12:37:57 +01003426http-request { allow | deny | tarpit | auth [realm <realm>] | redirect <rule> |
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003427 add-header <name> <fmt> | set-header <name> <fmt> |
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003428 capture <sample> [ len <length> | id <id> ] |
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003429 del-header <name> | set-nice <nice> | set-log-level <level> |
Sasha Pachev218f0642014-06-16 12:05:59 -06003430 replace-header <name> <match-regex> <replace-fmt> |
3431 replace-value <name> <match-regex> <replace-fmt> |
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01003432 set-method <fmt> | set-path <fmt> | set-query <fmt> |
3433 set-uri <fmt> | set-tos <tos> | set-mark <mark> |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003434 add-acl(<file name>) <key fmt> |
3435 del-acl(<file name>) <key fmt> |
3436 del-map(<file name>) <key fmt> |
Baptiste Assmannbb7e86a2014-09-03 18:29:47 +02003437 set-map(<file name>) <key fmt> <value fmt> |
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003438 set-var(<var name>) <expr> |
Thierry FOURNIER90da1912015-03-05 11:17:06 +01003439 { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] |
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02003440 sc-inc-gpc0(<sc-id>) |
Thierry FOURNIER236657b2015-08-19 08:25:14 +02003441 sc-set-gpt0(<sc-id>) <int> |
Willy Tarreau2d392c22015-08-24 01:43:45 +02003442 silent-drop |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003443 }
Cyril Bontéf0c60612010-02-06 14:44:47 +01003444 [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003445 Access control for Layer 7 requests
3446
3447 May be used in sections: defaults | frontend | listen | backend
3448 no | yes | yes | yes
3449
Willy Tarreau20b0de52012-12-24 15:45:22 +01003450 The http-request statement defines a set of rules which apply to layer 7
3451 processing. The rules are evaluated in their declaration order when they are
3452 met in a frontend, listen or backend section. Any rule may optionally be
3453 followed by an ACL-based condition, in which case it will only be evaluated
3454 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003455
Willy Tarreau20b0de52012-12-24 15:45:22 +01003456 The first keyword is the rule's action. Currently supported actions include :
3457 - "allow" : this stops the evaluation of the rules and lets the request
3458 pass the check. No further "http-request" rules are evaluated.
3459
3460 - "deny" : this stops the evaluation of the rules and immediately rejects
3461 the request and emits an HTTP 403 error. No further "http-request" rules
3462 are evaluated.
3463
Willy Tarreauccbcc372012-12-27 12:37:57 +01003464 - "tarpit" : this stops the evaluation of the rules and immediately blocks
3465 the request without responding for a delay specified by "timeout tarpit"
3466 or "timeout connect" if the former is not set. After that delay, if the
3467 client is still connected, an HTTP error 500 is returned so that the
3468 client does not suspect it has been tarpitted. Logs will report the flags
3469 "PT". The goal of the tarpit rule is to slow down robots during an attack
3470 when they're limited on the number of concurrent requests. It can be very
3471 efficient against very dumb robots, and will significantly reduce the
3472 load on firewalls compared to a "deny" rule. But when facing "correctly"
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003473 developed robots, it can make things worse by forcing haproxy and the
Willy Tarreau2d392c22015-08-24 01:43:45 +02003474 front firewall to support insane number of concurrent connections. See
3475 also the "silent-drop" action below.
Willy Tarreauccbcc372012-12-27 12:37:57 +01003476
Willy Tarreau20b0de52012-12-24 15:45:22 +01003477 - "auth" : this stops the evaluation of the rules and immediately responds
3478 with an HTTP 401 or 407 error code to invite the user to present a valid
3479 user name and password. No further "http-request" rules are evaluated. An
3480 optional "realm" parameter is supported, it sets the authentication realm
3481 that is returned with the response (typically the application's name).
3482
Willy Tarreau81499eb2012-12-27 12:19:02 +01003483 - "redirect" : this performs an HTTP redirection based on a redirect rule.
3484 This is exactly the same as the "redirect" statement except that it
3485 inserts a redirect rule which can be processed in the middle of other
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01003486 "http-request" rules and that these rules use the "log-format" strings.
3487 See the "redirect" keyword for the rule's syntax.
Willy Tarreau81499eb2012-12-27 12:19:02 +01003488
Willy Tarreau20b0de52012-12-24 15:45:22 +01003489 - "add-header" appends an HTTP header field whose name is specified in
3490 <name> and whose value is defined by <fmt> which follows the log-format
3491 rules (see Custom Log Format in section 8.2.4). This is particularly
3492 useful to pass connection-specific information to the server (eg: the
3493 client's SSL certificate), or to combine several headers into one. This
3494 rule is not final, so it is possible to add other similar rules. Note
3495 that header addition is performed immediately, so one rule might reuse
3496 the resulting header from a previous rule.
3497
3498 - "set-header" does the same as "add-header" except that the header name
3499 is first removed if it existed. This is useful when passing security
3500 information to the server, where the header must not be manipulated by
Willy Tarreau85603282015-01-21 20:39:27 +01003501 external users. Note that the new value is computed before the removal so
3502 it is possible to concatenate a value to an existing header.
Willy Tarreau20b0de52012-12-24 15:45:22 +01003503
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003504 - "del-header" removes all HTTP header fields whose name is specified in
3505 <name>.
3506
Sasha Pachev218f0642014-06-16 12:05:59 -06003507 - "replace-header" matches the regular expression in all occurrences of
3508 header field <name> according to <match-regex>, and replaces them with
3509 the <replace-fmt> argument. Format characters are allowed in replace-fmt
3510 and work like in <fmt> arguments in "add-header". The match is only
3511 case-sensitive. It is important to understand that this action only
3512 considers whole header lines, regardless of the number of values they
3513 may contain. This usage is suited to headers naturally containing commas
3514 in their value, such as If-Modified-Since and so on.
3515
3516 Example:
3517
3518 http-request replace-header Cookie foo=([^;]*);(.*) foo=\1;ip=%bi;\2
3519
3520 applied to:
3521
3522 Cookie: foo=foobar; expires=Tue, 14-Jun-2016 01:40:45 GMT;
3523
3524 outputs:
3525
3526 Cookie: foo=foobar;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT;
3527
3528 assuming the backend IP is 192.168.1.20
3529
3530 - "replace-value" works like "replace-header" except that it matches the
3531 regex against every comma-delimited value of the header field <name>
3532 instead of the entire header. This is suited for all headers which are
3533 allowed to carry more than one value. An example could be the Accept
3534 header.
3535
3536 Example:
3537
3538 http-request replace-value X-Forwarded-For ^192\.168\.(.*)$ 172.16.\1
3539
3540 applied to:
3541
3542 X-Forwarded-For: 192.168.10.1, 192.168.13.24, 10.0.0.37
3543
3544 outputs:
3545
3546 X-Forwarded-For: 172.16.10.1, 172.16.13.24, 10.0.0.37
3547
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01003548 - "set-method" rewrites the request method with the result of the
3549 evaluation of format string <fmt>. There should be very few valid reasons
3550 for having to do so as this is more likely to break something than to fix
3551 it.
3552
3553 - "set-path" rewrites the request path with the result of the evaluation of
3554 format string <fmt>. The query string, if any, is left intact. If a
3555 scheme and authority is found before the path, they are left intact as
3556 well. If the request doesn't have a path ("*"), this one is replaced with
3557 the format. This can be used to prepend a directory component in front of
3558 a path for example. See also "set-query" and "set-uri".
3559
3560 Example :
3561 # prepend the host name before the path
3562 http-request set-path /%[hdr(host)]%[path]
3563
3564 - "set-query" rewrites the request's query string which appears after the
3565 first question mark ("?") with the result of the evaluation of format
3566 string <fmt>. The part prior to the question mark is left intact. If the
3567 request doesn't contain a question mark and the new value is not empty,
3568 then one is added at the end of the URI, followed by the new value. If
3569 a question mark was present, it will never be removed even if the value
3570 is empty. This can be used to add or remove parameters from the query
3571 string. See also "set-query" and "set-uri".
3572
3573 Example :
3574 # replace "%3D" with "=" in the query string
3575 http-request set-query %[query,regsub(%3D,=,g)]
3576
3577 - "set-uri" rewrites the request URI with the result of the evaluation of
3578 format string <fmt>. The scheme, authority, path and query string are all
3579 replaced at once. This can be used to rewrite hosts in front of proxies,
3580 or to perform complex modifications to the URI such as moving parts
3581 between the path and the query string. See also "set-path" and
3582 "set-query".
3583
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003584 - "set-nice" sets the "nice" factor of the current request being processed.
3585 It only has effect against the other requests being processed at the same
3586 time. The default value is 0, unless altered by the "nice" setting on the
3587 "bind" line. The accepted range is -1024..1024. The higher the value, the
3588 nicest the request will be. Lower values will make the request more
3589 important than other ones. This can be useful to improve the speed of
3590 some requests, or lower the priority of non-important requests. Using
3591 this setting without prior experimentation can cause some major slowdown.
3592
Willy Tarreau9a355ec2013-06-11 17:45:46 +02003593 - "set-log-level" is used to change the log level of the current request
3594 when a certain condition is met. Valid levels are the 8 syslog levels
3595 (see the "log" keyword) plus the special level "silent" which disables
3596 logging for this request. This rule is not final so the last matching
3597 rule wins. This rule can be useful to disable health checks coming from
3598 another equipment.
3599
Willy Tarreau42cf39e2013-06-11 18:51:32 +02003600 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
3601 the client to the value passed in <tos> on platforms which support this.
3602 This value represents the whole 8 bits of the IP TOS field, and can be
3603 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
3604 that only the 6 higher bits are used in DSCP or TOS, and the two lower
3605 bits are always 0. This can be used to adjust some routing behaviour on
3606 border routers based on some information from the request. See RFC 2474,
3607 2597, 3260 and 4594 for more information.
3608
Willy Tarreau51347ed2013-06-11 19:34:13 +02003609 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
3610 client to the value passed in <mark> on platforms which support it. This
3611 value is an unsigned 32 bit value which can be matched by netfilter and
3612 by the routing table. It can be expressed both in decimal or hexadecimal
3613 format (prefixed by "0x"). This can be useful to force certain packets to
3614 take a different route (for example a cheaper network path for bulk
3615 downloads). This works on Linux kernels 2.6.32 and above and requires
3616 admin privileges.
3617
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003618 - "add-acl" is used to add a new entry into an ACL. The ACL must be loaded
3619 from a file (even a dummy empty file). The file name of the ACL to be
3620 updated is passed between parentheses. It takes one argument: <key fmt>,
3621 which follows log-format rules, to collect content of the new entry. It
3622 performs a lookup in the ACL before insertion, to avoid duplicated (or
3623 more) values. This lookup is done by a linear search and can be expensive
3624 with large lists! It is the equivalent of the "add acl" command from the
3625 stats socket, but can be triggered by an HTTP request.
3626
3627 - "del-acl" is used to delete an entry from an ACL. The ACL must be loaded
3628 from a file (even a dummy empty file). The file name of the ACL to be
3629 updated is passed between parentheses. It takes one argument: <key fmt>,
3630 which follows log-format rules, to collect content of the entry to delete.
3631 It is the equivalent of the "del acl" command from the stats socket, but
3632 can be triggered by an HTTP request.
3633
3634 - "del-map" is used to delete an entry from a MAP. The MAP must be loaded
3635 from a file (even a dummy empty file). The file name of the MAP to be
3636 updated is passed between parentheses. It takes one argument: <key fmt>,
3637 which follows log-format rules, to collect content of the entry to delete.
3638 It takes one argument: "file name" It is the equivalent of the "del map"
3639 command from the stats socket, but can be triggered by an HTTP request.
3640
3641 - "set-map" is used to add a new entry into a MAP. The MAP must be loaded
3642 from a file (even a dummy empty file). The file name of the MAP to be
3643 updated is passed between parentheses. It takes 2 arguments: <key fmt>,
3644 which follows log-format rules, used to collect MAP key, and <value fmt>,
3645 which follows log-format rules, used to collect content for the new entry.
3646 It performs a lookup in the MAP before insertion, to avoid duplicated (or
3647 more) values. This lookup is done by a linear search and can be expensive
3648 with large lists! It is the equivalent of the "set map" command from the
3649 stats socket, but can be triggered by an HTTP request.
3650
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003651 - capture <sample> [ len <length> | id <id> ] :
Willy Tarreaua9083d02015-05-08 15:27:59 +02003652 captures sample expression <sample> from the request buffer, and converts
3653 it to a string of at most <len> characters. The resulting string is
3654 stored into the next request "capture" slot, so it will possibly appear
3655 next to some captured HTTP headers. It will then automatically appear in
3656 the logs, and it will be possible to extract it using sample fetch rules
3657 to feed it into headers or anything. The length should be limited given
3658 that this size will be allocated for each capture during the whole
3659 session life. Please check section 7.3 (Fetching samples) and "capture
3660 request header" for more information.
3661
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003662 If the keyword "id" is used instead of "len", the action tries to store
3663 the captured string in a previously declared capture slot. This is useful
3664 to run captures in backends. The slot id can be declared by a previous
3665 directive "http-request capture" or with the "declare capture" keyword.
Baptiste Assmanne9544932015-11-03 23:31:35 +01003666 If the slot <id> doesn't exist, then HAProxy fails parsing the
3667 configuration to prevent unexpected behavior at run time.
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02003668
Willy Tarreau09448f72014-06-25 18:12:15 +02003669 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
3670 enables tracking of sticky counters from current request. These rules
3671 do not stop evaluation and do not change default action. Three sets of
3672 counters may be simultaneously tracked by the same connection. The first
3673 "track-sc0" rule executed enables tracking of the counters of the
3674 specified table as the first set. The first "track-sc1" rule executed
3675 enables tracking of the counters of the specified table as the second
3676 set. The first "track-sc2" rule executed enables tracking of the
3677 counters of the specified table as the third set. It is a recommended
3678 practice to use the first set of counters for the per-frontend counters
3679 and the second set for the per-backend ones. But this is just a
3680 guideline, all may be used everywhere.
3681
3682 These actions take one or two arguments :
3683 <key> is mandatory, and is a sample expression rule as described
3684 in section 7.3. It describes what elements of the incoming
3685 request or connection will be analysed, extracted, combined,
3686 and used to select which table entry to update the counters.
3687
3688 <table> is an optional table to be used instead of the default one,
3689 which is the stick-table declared in the current proxy. All
3690 the counters for the matches and updates for the key will
3691 then be performed in that table until the session ends.
3692
3693 Once a "track-sc*" rule is executed, the key is looked up in the table
3694 and if it is not found, an entry is allocated for it. Then a pointer to
3695 that entry is kept during all the session's life, and this entry's
3696 counters are updated as often as possible, every time the session's
3697 counters are updated, and also systematically when the session ends.
3698 Counters are only updated for events that happen after the tracking has
3699 been started. As an exception, connection counters and request counters
3700 are systematically updated so that they reflect useful information.
3701
3702 If the entry tracks concurrent connection counters, one connection is
3703 counted for as long as the entry is tracked, and the entry will not
3704 expire during that time. Tracking counters also provides a performance
3705 advantage over just checking the keys, because only one table lookup is
3706 performed for all ACL checks that make use of it.
3707
Thierry FOURNIER236657b2015-08-19 08:25:14 +02003708 - sc-set-gpt0(<sc-id>) <int> :
3709 This action sets the GPT0 tag according to the sticky counter designated
3710 by <sc-id> and the value of <int>. The expected result is a boolean. If
3711 an error occurs, this action silently fails and the actions evaluation
3712 continues.
3713
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02003714 - sc-inc-gpc0(<sc-id>):
3715 This action increments the GPC0 counter according with the sticky counter
3716 designated by <sc-id>. If an error occurs, this action silently fails and
3717 the actions evaluation continues.
3718
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003719 - set-var(<var-name>) <expr> :
3720 Is used to set the contents of a variable. The variable is declared
3721 inline.
3722
3723 <var-name> The name of the variable starts by an indication about its
3724 scope. The allowed scopes are:
3725 "sess" : the variable is shared with all the session,
3726 "txn" : the variable is shared with all the transaction
3727 (request and response)
3728 "req" : the variable is shared only during the request
3729 processing
3730 "res" : the variable is shared only during the response
3731 processing.
3732 This prefix is followed by a name. The separator is a '.'.
3733 The name may only contain characters 'a-z', 'A-Z', '0-9',
3734 and '_'.
3735
3736 <expr> Is a standard HAProxy expression formed by a sample-fetch
3737 followed by some converters.
3738
3739 Example:
3740
3741 http-request set-var(req.my_var) req.fhdr(user-agent),lower
3742
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02003743 - set-src <expr> :
3744 Is used to set the source IP address to the value of specified
3745 expression. Useful when a proxy in front of HAProxy rewrites source IP,
3746 but provides the correct IP in a HTTP header; or you want to mask
3747 source IP for privacy.
3748
3749 <expr> Is a standard HAProxy expression formed by a sample-fetch
3750 followed by some converters.
3751
3752 Example:
3753
3754 http-request set-src hdr(x-forwarded-for)
3755 http-request set-src src,ipmask(24)
3756
3757 When set-src is successful, the source port is set to 0.
3758
Willy Tarreau2d392c22015-08-24 01:43:45 +02003759 - "silent-drop" : this stops the evaluation of the rules and makes the
3760 client-facing connection suddenly disappear using a system-dependant way
3761 that tries to prevent the client from being notified. The effect it then
3762 that the client still sees an established connection while there's none
3763 on HAProxy. The purpose is to achieve a comparable effect to "tarpit"
3764 except that it doesn't use any local resource at all on the machine
3765 running HAProxy. It can resist much higher loads than "tarpit", and slow
3766 down stronger attackers. It is important to undestand the impact of using
3767 this mechanism. All stateful equipments placed between the client and
3768 HAProxy (firewalls, proxies, load balancers) will also keep the
3769 established connection for a long time and may suffer from this action.
3770 On modern Linux systems running with enough privileges, the TCP_REPAIR
3771 socket option is used to block the emission of a TCP reset. On other
3772 systems, the socket's TTL is reduced to 1 so that the TCP reset doesn't
3773 pass the first router, though it's still delivered to local networks. Do
3774 not use it unless you fully understand how it works.
3775
Willy Tarreau20b0de52012-12-24 15:45:22 +01003776 There is no limit to the number of http-request statements per instance.
3777
3778 It is important to know that http-request rules are processed very early in
3779 the HTTP processing, just after "block" rules and before "reqdel" or "reqrep"
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08003780 or "reqadd" rules. That way, headers added by "add-header"/"set-header" are
3781 visible by almost all further ACL rules.
3782
3783 Using "reqadd"/"reqdel"/"reqrep" to manipulate request headers is discouraged
3784 in newer versions (>= 1.5). But if you need to use regular expression to
3785 delete headers, you can still use "reqdel". Also please use
3786 "http-request deny/allow/tarpit" instead of "reqdeny"/"reqpass"/"reqtarpit".
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003787
3788 Example:
Cyril Bonté78caf842010-03-10 22:41:43 +01003789 acl nagios src 192.168.129.3
3790 acl local_net src 192.168.0.0/16
3791 acl auth_ok http_auth(L1)
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003792
Cyril Bonté78caf842010-03-10 22:41:43 +01003793 http-request allow if nagios
3794 http-request allow if local_net auth_ok
3795 http-request auth realm Gimme if local_net auth_ok
3796 http-request deny
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003797
Cyril Bonté78caf842010-03-10 22:41:43 +01003798 Example:
3799 acl auth_ok http_auth_group(L1) G1
Cyril Bonté78caf842010-03-10 22:41:43 +01003800 http-request auth unless auth_ok
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003801
Willy Tarreau20b0de52012-12-24 15:45:22 +01003802 Example:
3803 http-request set-header X-Haproxy-Current-Date %T
3804 http-request set-header X-SSL %[ssl_fc]
Willy Tarreaufca42612015-08-27 17:15:05 +02003805 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id,hex]
Willy Tarreau20b0de52012-12-24 15:45:22 +01003806 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
3807 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
3808 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
3809 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
3810 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
3811 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
3812
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003813 Example:
3814 acl key req.hdr(X-Add-Acl-Key) -m found
3815 acl add path /addacl
3816 acl del path /delacl
3817
3818 acl myhost hdr(Host) -f myhost.lst
3819
3820 http-request add-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key add
3821 http-request del-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key del
3822
3823 Example:
3824 acl value req.hdr(X-Value) -m found
3825 acl setmap path /setmap
3826 acl delmap path /delmap
3827
3828 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
3829
3830 http-request set-map(map.lst) %[src] %[req.hdr(X-Value)] if setmap value
3831 http-request del-map(map.lst) %[src] if delmap
3832
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02003833 See also : "stats http-request", section 3.4 about userlists and section 7
3834 about ACL usage.
Willy Tarreauef781042010-01-27 11:53:01 +01003835
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003836http-response { allow | deny | add-header <name> <fmt> | set-nice <nice> |
Willy Tarreau51d861a2015-05-22 17:30:48 +02003837 capture <sample> id <id> | redirect <rule> |
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003838 set-header <name> <fmt> | del-header <name> |
Sasha Pachev218f0642014-06-16 12:05:59 -06003839 replace-header <name> <regex-match> <replace-fmt> |
3840 replace-value <name> <regex-match> <replace-fmt> |
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02003841 set-status <status> |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003842 set-log-level <level> | set-mark <mark> | set-tos <tos> |
3843 add-acl(<file name>) <key fmt> |
3844 del-acl(<file name>) <key fmt> |
3845 del-map(<file name>) <key fmt> |
Thierry FOURNIER90da1912015-03-05 11:17:06 +01003846 set-map(<file name>) <key fmt> <value fmt> |
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02003847 set-var(<var-name>) <expr> |
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02003848 sc-inc-gpc0(<sc-id>) |
Thierry FOURNIER236657b2015-08-19 08:25:14 +02003849 sc-set-gpt0(<sc-id>) <int> |
Willy Tarreau2d392c22015-08-24 01:43:45 +02003850 silent-drop |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003851 }
Lukas Tribus2dd1d1a2013-06-19 23:34:41 +02003852 [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02003853 Access control for Layer 7 responses
3854
3855 May be used in sections: defaults | frontend | listen | backend
3856 no | yes | yes | yes
3857
3858 The http-response statement defines a set of rules which apply to layer 7
3859 processing. The rules are evaluated in their declaration order when they are
3860 met in a frontend, listen or backend section. Any rule may optionally be
3861 followed by an ACL-based condition, in which case it will only be evaluated
3862 if the condition is true. Since these rules apply on responses, the backend
3863 rules are applied first, followed by the frontend's rules.
3864
3865 The first keyword is the rule's action. Currently supported actions include :
3866 - "allow" : this stops the evaluation of the rules and lets the response
3867 pass the check. No further "http-response" rules are evaluated for the
3868 current section.
3869
3870 - "deny" : this stops the evaluation of the rules and immediately rejects
3871 the response and emits an HTTP 502 error. No further "http-response"
3872 rules are evaluated.
3873
3874 - "add-header" appends an HTTP header field whose name is specified in
3875 <name> and whose value is defined by <fmt> which follows the log-format
3876 rules (see Custom Log Format in section 8.2.4). This may be used to send
3877 a cookie to a client for example, or to pass some internal information.
3878 This rule is not final, so it is possible to add other similar rules.
3879 Note that header addition is performed immediately, so one rule might
3880 reuse the resulting header from a previous rule.
3881
3882 - "set-header" does the same as "add-header" except that the header name
3883 is first removed if it existed. This is useful when passing security
3884 information to the server, where the header must not be manipulated by
3885 external users.
3886
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003887 - "del-header" removes all HTTP header fields whose name is specified in
3888 <name>.
3889
Sasha Pachev218f0642014-06-16 12:05:59 -06003890 - "replace-header" matches the regular expression in all occurrences of
3891 header field <name> according to <match-regex>, and replaces them with
3892 the <replace-fmt> argument. Format characters are allowed in replace-fmt
3893 and work like in <fmt> arguments in "add-header". The match is only
3894 case-sensitive. It is important to understand that this action only
3895 considers whole header lines, regardless of the number of values they
3896 may contain. This usage is suited to headers naturally containing commas
3897 in their value, such as Set-Cookie, Expires and so on.
3898
3899 Example:
3900
3901 http-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
3902
3903 applied to:
3904
3905 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
3906
3907 outputs:
3908
3909 Set-Cookie: C=1;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT
3910
3911 assuming the backend IP is 192.168.1.20.
3912
3913 - "replace-value" works like "replace-header" except that it matches the
3914 regex against every comma-delimited value of the header field <name>
3915 instead of the entire header. This is suited for all headers which are
3916 allowed to carry more than one value. An example could be the Accept
3917 header.
3918
3919 Example:
3920
3921 http-response replace-value Cache-control ^public$ private
3922
3923 applied to:
3924
3925 Cache-Control: max-age=3600, public
3926
3927 outputs:
3928
3929 Cache-Control: max-age=3600, private
3930
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02003931 - "set-status" replaces the response status code with <status> which must
3932 be an integer between 100 and 999. Note that the reason is automatically
3933 adapted to the new code.
3934
3935 Example:
3936
3937 # return "431 Request Header Fields Too Large"
3938 http-response set-status 431
3939
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003940 - "set-nice" sets the "nice" factor of the current request being processed.
3941 It only has effect against the other requests being processed at the same
3942 time. The default value is 0, unless altered by the "nice" setting on the
3943 "bind" line. The accepted range is -1024..1024. The higher the value, the
3944 nicest the request will be. Lower values will make the request more
3945 important than other ones. This can be useful to improve the speed of
3946 some requests, or lower the priority of non-important requests. Using
3947 this setting without prior experimentation can cause some major slowdown.
3948
Willy Tarreau9a355ec2013-06-11 17:45:46 +02003949 - "set-log-level" is used to change the log level of the current request
3950 when a certain condition is met. Valid levels are the 8 syslog levels
3951 (see the "log" keyword) plus the special level "silent" which disables
3952 logging for this request. This rule is not final so the last matching
3953 rule wins. This rule can be useful to disable health checks coming from
3954 another equipment.
3955
Willy Tarreau42cf39e2013-06-11 18:51:32 +02003956 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
3957 the client to the value passed in <tos> on platforms which support this.
3958 This value represents the whole 8 bits of the IP TOS field, and can be
3959 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
3960 that only the 6 higher bits are used in DSCP or TOS, and the two lower
3961 bits are always 0. This can be used to adjust some routing behaviour on
3962 border routers based on some information from the request. See RFC 2474,
3963 2597, 3260 and 4594 for more information.
3964
Willy Tarreau51347ed2013-06-11 19:34:13 +02003965 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
3966 client to the value passed in <mark> on platforms which support it. This
3967 value is an unsigned 32 bit value which can be matched by netfilter and
3968 by the routing table. It can be expressed both in decimal or hexadecimal
3969 format (prefixed by "0x"). This can be useful to force certain packets to
3970 take a different route (for example a cheaper network path for bulk
3971 downloads). This works on Linux kernels 2.6.32 and above and requires
3972 admin privileges.
3973
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003974 - "add-acl" is used to add a new entry into an ACL. The ACL must be loaded
3975 from a file (even a dummy empty file). The file name of the ACL to be
3976 updated is passed between parentheses. It takes one argument: <key fmt>,
3977 which follows log-format rules, to collect content of the new entry. It
3978 performs a lookup in the ACL before insertion, to avoid duplicated (or
3979 more) values. This lookup is done by a linear search and can be expensive
3980 with large lists! It is the equivalent of the "add acl" command from the
3981 stats socket, but can be triggered by an HTTP response.
3982
3983 - "del-acl" is used to delete an entry from an ACL. The ACL must be loaded
3984 from a file (even a dummy empty file). The file name of the ACL to be
3985 updated is passed between parentheses. It takes one argument: <key fmt>,
3986 which follows log-format rules, to collect content of the entry to delete.
3987 It is the equivalent of the "del acl" command from the stats socket, but
3988 can be triggered by an HTTP response.
3989
3990 - "del-map" is used to delete an entry from a MAP. The MAP must be loaded
3991 from a file (even a dummy empty file). The file name of the MAP to be
3992 updated is passed between parentheses. It takes one argument: <key fmt>,
3993 which follows log-format rules, to collect content of the entry to delete.
3994 It takes one argument: "file name" It is the equivalent of the "del map"
3995 command from the stats socket, but can be triggered by an HTTP response.
3996
3997 - "set-map" is used to add a new entry into a MAP. The MAP must be loaded
3998 from a file (even a dummy empty file). The file name of the MAP to be
3999 updated is passed between parentheses. It takes 2 arguments: <key fmt>,
4000 which follows log-format rules, used to collect MAP key, and <value fmt>,
4001 which follows log-format rules, used to collect content for the new entry.
4002 It performs a lookup in the MAP before insertion, to avoid duplicated (or
4003 more) values. This lookup is done by a linear search and can be expensive
4004 with large lists! It is the equivalent of the "set map" command from the
4005 stats socket, but can be triggered by an HTTP response.
4006
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02004007 - capture <sample> id <id> :
4008 captures sample expression <sample> from the response buffer, and converts
4009 it to a string. The resulting string is stored into the next request
4010 "capture" slot, so it will possibly appear next to some captured HTTP
4011 headers. It will then automatically appear in the logs, and it will be
4012 possible to extract it using sample fetch rules to feed it into headers or
4013 anything. Please check section 7.3 (Fetching samples) and "capture
4014 response header" for more information.
4015
4016 The keyword "id" is the id of the capture slot which is used for storing
4017 the string. The capture slot must be defined in an associated frontend.
4018 This is useful to run captures in backends. The slot id can be declared by
4019 a previous directive "http-response capture" or with the "declare capture"
4020 keyword.
Baptiste Assmanne9544932015-11-03 23:31:35 +01004021 If the slot <id> doesn't exist, then HAProxy fails parsing the
4022 configuration to prevent unexpected behavior at run time.
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02004023
Willy Tarreau51d861a2015-05-22 17:30:48 +02004024 - "redirect" : this performs an HTTP redirection based on a redirect rule.
4025 This supports a format string similarly to "http-request redirect" rules,
4026 with the exception that only the "location" type of redirect is possible
4027 on the response. See the "redirect" keyword for the rule's syntax. When
4028 a redirect rule is applied during a response, connections to the server
4029 are closed so that no data can be forwarded from the server to the client.
4030
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004031 - set-var(<var-name>) expr:
4032 Is used to set the contents of a variable. The variable is declared
4033 inline.
4034
4035 <var-name> The name of the variable starts by an indication about its
4036 scope. The allowed scopes are:
4037 "sess" : the variable is shared with all the session,
4038 "txn" : the variable is shared with all the transaction
4039 (request and response)
4040 "req" : the variable is shared only during the request
4041 processing
4042 "res" : the variable is shared only during the response
4043 processing.
4044 This prefix is followed by a name. The separator is a '.'.
4045 The name may only contain characters 'a-z', 'A-Z', '0-9',
4046 and '_'.
4047
4048 <expr> Is a standard HAProxy expression formed by a sample-fetch
4049 followed by some converters.
4050
4051 Example:
4052
4053 http-response set-var(sess.last_redir) res.hdr(location)
4054
Thierry FOURNIER236657b2015-08-19 08:25:14 +02004055 - sc-set-gpt0(<sc-id>) <int> :
4056 This action sets the GPT0 tag according to the sticky counter designated
4057 by <sc-id> and the value of <int>. The expected result is a boolean. If
4058 an error occurs, this action silently fails and the actions evaluation
4059 continues.
4060
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02004061 - sc-inc-gpc0(<sc-id>):
4062 This action increments the GPC0 counter according with the sticky counter
4063 designated by <sc-id>. If an error occurs, this action silently fails and
4064 the actions evaluation continues.
4065
Willy Tarreau2d392c22015-08-24 01:43:45 +02004066 - "silent-drop" : this stops the evaluation of the rules and makes the
4067 client-facing connection suddenly disappear using a system-dependant way
4068 that tries to prevent the client from being notified. The effect it then
4069 that the client still sees an established connection while there's none
4070 on HAProxy. The purpose is to achieve a comparable effect to "tarpit"
4071 except that it doesn't use any local resource at all on the machine
4072 running HAProxy. It can resist much higher loads than "tarpit", and slow
4073 down stronger attackers. It is important to undestand the impact of using
4074 this mechanism. All stateful equipments placed between the client and
4075 HAProxy (firewalls, proxies, load balancers) will also keep the
4076 established connection for a long time and may suffer from this action.
4077 On modern Linux systems running with enough privileges, the TCP_REPAIR
4078 socket option is used to block the emission of a TCP reset. On other
4079 systems, the socket's TTL is reduced to 1 so that the TCP reset doesn't
4080 pass the first router, though it's still delivered to local networks. Do
4081 not use it unless you fully understand how it works.
4082
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004083 There is no limit to the number of http-response statements per instance.
4084
Godbach09250262013-07-02 01:19:15 +08004085 It is important to know that http-response rules are processed very early in
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08004086 the HTTP processing, before "rspdel" or "rsprep" or "rspadd" rules. That way,
4087 headers added by "add-header"/"set-header" are visible by almost all further ACL
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004088 rules.
4089
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08004090 Using "rspadd"/"rspdel"/"rsprep" to manipulate request headers is discouraged
4091 in newer versions (>= 1.5). But if you need to use regular expression to
4092 delete headers, you can still use "rspdel". Also please use
4093 "http-response deny" instead of "rspdeny".
4094
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004095 Example:
4096 acl key_acl res.hdr(X-Acl-Key) -m found
4097
4098 acl myhost hdr(Host) -f myhost.lst
4099
4100 http-response add-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
4101 http-response del-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
4102
4103 Example:
4104 acl value res.hdr(X-Value) -m found
4105
4106 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
4107
4108 http-response set-map(map.lst) %[src] %[res.hdr(X-Value)] if value
4109 http-response del-map(map.lst) %[src] if ! value
4110
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004111 See also : "http-request", section 3.4 about userlists and section 7 about
4112 ACL usage.
4113
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02004114
Willy Tarreau30631952015-08-06 15:05:24 +02004115http-reuse { never | safe | aggressive | always }
4116 Declare how idle HTTP connections may be shared between requests
4117
4118 May be used in sections: defaults | frontend | listen | backend
4119 yes | no | yes | yes
4120
4121 By default, a connection established between haproxy and the backend server
4122 belongs to the session that initiated it. The downside is that between the
4123 response and the next request, the connection remains idle and is not used.
4124 In many cases for performance reasons it is desirable to make it possible to
4125 reuse these idle connections to serve other requests from different sessions.
4126 This directive allows to tune this behaviour.
4127
4128 The argument indicates the desired connection reuse strategy :
4129
4130 - "never" : idle connections are never shared between sessions. This is
4131 the default choice. It may be enforced to cancel a different
4132 strategy inherited from a defaults section or for
4133 troubleshooting. For example, if an old bogus application
4134 considers that multiple requests over the same connection come
4135 from the same client and it is not possible to fix the
4136 application, it may be desirable to disable connection sharing
4137 in a single backend. An example of such an application could
4138 be an old haproxy using cookie insertion in tunnel mode and
4139 not checking any request past the first one.
4140
4141 - "safe" : this is the recommended strategy. The first request of a
4142 session is always sent over its own connection, and only
4143 subsequent requests may be dispatched over other existing
4144 connections. This ensures that in case the server closes the
4145 connection when the request is being sent, the browser can
4146 decide to silently retry it. Since it is exactly equivalent to
4147 regular keep-alive, there should be no side effects.
4148
4149 - "aggressive" : this mode may be useful in webservices environments where
4150 all servers are not necessarily known and where it would be
4151 appreciable to deliver most first requests over existing
4152 connections. In this case, first requests are only delivered
4153 over existing connections that have been reused at least once,
4154 proving that the server correctly supports connection reuse.
4155 It should only be used when it's sure that the client can
4156 retry a failed request once in a while and where the benefit
4157 of aggressive connection reuse significantly outweights the
4158 downsides of rare connection failures.
4159
4160 - "always" : this mode is only recommended when the path to the server is
4161 known for never breaking existing connections quickly after
4162 releasing them. It allows the first request of a session to be
4163 sent to an existing connection. This can provide a significant
4164 performance increase over the "safe" strategy when the backend
4165 is a cache farm, since such components tend to show a
4166 consistent behaviour and will benefit from the connection
4167 sharing. It is recommended that the "http-keep-alive" timeout
4168 remains low in this mode so that no dead connections remain
4169 usable. In most cases, this will lead to the same performance
4170 gains as "aggressive" but with more risks. It should only be
4171 used when it improves the situation over "aggressive".
4172
4173 When http connection sharing is enabled, a great care is taken to respect the
4174 connection properties and compatiblities. Specifically :
4175 - connections made with "usesrc" followed by a client-dependant value
4176 ("client", "clientip", "hdr_ip") are marked private and never shared ;
4177
4178 - connections sent to a server with a TLS SNI extension are marked private
4179 and are never shared ;
4180
4181 - connections receiving a status code 401 or 407 expect some authentication
4182 to be sent in return. Due to certain bogus authentication schemes (such
4183 as NTLM) relying on the connection, these connections are marked private
4184 and are never shared ;
4185
4186 No connection pool is involved, once a session dies, the last idle connection
4187 it was attached to is deleted at the same time. This ensures that connections
4188 may not last after all sessions are closed.
4189
4190 Note: connection reuse improves the accuracy of the "server maxconn" setting,
4191 because almost no new connection will be established while idle connections
4192 remain available. This is particularly true with the "always" strategy.
4193
4194 See also : "option http-keep-alive", "server maxconn"
4195
4196
Mark Lamourinec2247f02012-01-04 13:02:01 -05004197http-send-name-header [<header>]
4198 Add the server name to a request. Use the header string given by <header>
4199
4200 May be used in sections: defaults | frontend | listen | backend
4201 yes | no | yes | yes
4202
4203 Arguments :
4204
4205 <header> The header string to use to send the server name
4206
4207 The "http-send-name-header" statement causes the name of the target
4208 server to be added to the headers of an HTTP request. The name
4209 is added with the header string proved.
4210
4211 See also : "server"
4212
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01004213id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02004214 Set a persistent ID to a proxy.
4215 May be used in sections : defaults | frontend | listen | backend
4216 no | yes | yes | yes
4217 Arguments : none
4218
4219 Set a persistent ID for the proxy. This ID must be unique and positive.
4220 An unused ID will automatically be assigned if unset. The first assigned
4221 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01004222
4223
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004224ignore-persist { if | unless } <condition>
4225 Declare a condition to ignore persistence
4226 May be used in sections: defaults | frontend | listen | backend
4227 no | yes | yes | yes
4228
4229 By default, when cookie persistence is enabled, every requests containing
4230 the cookie are unconditionally persistent (assuming the target server is up
4231 and running).
4232
4233 The "ignore-persist" statement allows one to declare various ACL-based
4234 conditions which, when met, will cause a request to ignore persistence.
4235 This is sometimes useful to load balance requests for static files, which
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03004236 often don't require persistence. This can also be used to fully disable
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004237 persistence for a specific User-Agent (for example, some web crawler bots).
4238
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004239 The persistence is ignored when an "if" condition is met, or unless an
4240 "unless" condition is met.
4241
4242 See also : "force-persist", "cookie", and section 7 about ACL usage.
4243
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004244load-server-state-from-file { global | local | none }
4245 Allow seamless reload of HAProxy
4246 May be used in sections: defaults | frontend | listen | backend
4247 yes | no | yes | yes
4248
4249 This directive points HAProxy to a file where server state from previous
4250 running process has been saved. That way, when starting up, before handling
4251 traffic, the new process can apply old states to servers exactly has if no
4252 reload occured. The purpose of the "load-server-state-from-file" directive is
4253 to tell haproxy which file to use. For now, only 2 arguments to either prevent
4254 loading state or load states from a file containing all backends and servers.
4255 The state file can be generated by running the command "show servers state"
4256 over the stats socket and redirect output.
4257
4258 The format of the file is versionned and is very specific. To understand it,
4259 please read the documentation of the "show servers state" command (chapter
Kevin Decherf949c7202015-10-13 23:26:44 +02004260 9.2 of Management Guide).
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02004261
4262 Arguments:
4263 global load the content of the file pointed by the global directive
4264 named "server-state-file".
4265
4266 local load the content of the file pointed by the directive
4267 "server-state-file-name" if set. If not set, then the backend
4268 name is used as a file name.
4269
4270 none don't load any stat for this backend
4271
4272 Notes:
4273 - server's IP address is not updated unless DNS resolution is enabled on
4274 the server. It means that if a server IP address has been changed using
4275 the stat socket, this information won't be re-applied after reloading.
4276
4277 - server's weight is applied from previous running process unless it has
4278 has changed between previous and new configuration files.
4279
4280 Example 1:
4281
4282 Minimal configuration:
4283
4284 global
4285 stats socket /tmp/socket
4286 server-state-file /tmp/server_state
4287
4288 defaults
4289 load-server-state-from-file global
4290
4291 backend bk
4292 server s1 127.0.0.1:22 check weight 11
4293 server s2 127.0.0.1:22 check weight 12
4294
4295 Then one can run :
4296
4297 socat /tmp/socket - <<< "show servers state" > /tmp/server_state
4298
4299 Content of the file /tmp/server_state would be like this:
4300
4301 1
4302 # <field names skipped for the doc example>
4303 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
4304 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
4305
4306 Example 2:
4307
4308 Minimal configuration:
4309
4310 global
4311 stats socket /tmp/socket
4312 server-state-base /etc/haproxy/states
4313
4314 defaults
4315 load-server-state-from-file local
4316
4317 backend bk
4318 server s1 127.0.0.1:22 check weight 11
4319 server s2 127.0.0.1:22 check weight 12
4320
4321 Then one can run :
4322
4323 socat /tmp/socket - <<< "show servers state bk" > /etc/haproxy/states/bk
4324
4325 Content of the file /etc/haproxy/states/bk would be like this:
4326
4327 1
4328 # <field names skipped for the doc example>
4329 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
4330 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
4331
4332 See also: "server-state-file", "server-state-file-name", and
4333 "show servers state"
4334
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004335
Willy Tarreau2769aa02007-12-27 18:26:09 +01004336log global
Willy Tarreau18324f52014-06-27 18:10:07 +02004337log <address> [len <length>] <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02004338no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01004339 Enable per-instance logging of events and traffic.
4340 May be used in sections : defaults | frontend | listen | backend
4341 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02004342
4343 Prefix :
4344 no should be used when the logger list must be flushed. For example,
4345 if you don't want to inherit from the default logger list. This
4346 prefix does not allow arguments.
4347
Willy Tarreau2769aa02007-12-27 18:26:09 +01004348 Arguments :
4349 global should be used when the instance's logging parameters are the
4350 same as the global ones. This is the most common usage. "global"
4351 replaces <address>, <facility> and <level> with those of the log
4352 entries found in the "global" section. Only one "log global"
4353 statement may be used per instance, and this form takes no other
4354 parameter.
4355
4356 <address> indicates where to send the logs. It takes the same format as
4357 for the "global" section's logs, and can be one of :
4358
4359 - An IPv4 address optionally followed by a colon (':') and a UDP
4360 port. If no port is specified, 514 is used by default (the
4361 standard syslog port).
4362
David du Colombier24bb5f52011-03-17 10:40:23 +01004363 - An IPv6 address followed by a colon (':') and optionally a UDP
4364 port. If no port is specified, 514 is used by default (the
4365 standard syslog port).
4366
Willy Tarreau2769aa02007-12-27 18:26:09 +01004367 - A filesystem path to a UNIX domain socket, keeping in mind
4368 considerations for chroot (be sure the path is accessible
4369 inside the chroot) and uid/gid (be sure the path is
4370 appropriately writeable).
4371
William Lallemandb2f07452015-05-12 14:27:13 +02004372 You may want to reference some environment variables in the
4373 address parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01004374
Willy Tarreau18324f52014-06-27 18:10:07 +02004375 <length> is an optional maximum line length. Log lines larger than this
4376 value will be truncated before being sent. The reason is that
4377 syslog servers act differently on log line length. All servers
4378 support the default value of 1024, but some servers simply drop
4379 larger lines while others do log them. If a server supports long
4380 lines, it may make sense to set this value here in order to avoid
4381 truncating long lines. Similarly, if a server drops long lines,
4382 it is preferable to truncate them before sending them. Accepted
4383 values are 80 to 65535 inclusive. The default value of 1024 is
4384 generally fine for all standard usages. Some specific cases of
4385 long captures or JSON-formated logs may require larger values.
4386
Willy Tarreau2769aa02007-12-27 18:26:09 +01004387 <facility> must be one of the 24 standard syslog facilities :
4388
4389 kern user mail daemon auth syslog lpr news
4390 uucp cron auth2 ftp ntp audit alert cron2
4391 local0 local1 local2 local3 local4 local5 local6 local7
4392
4393 <level> is optional and can be specified to filter outgoing messages. By
4394 default, all messages are sent. If a level is specified, only
4395 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02004396 will be sent. An optional minimum level can be specified. If it
4397 is set, logs emitted with a more severe level than this one will
4398 be capped to this level. This is used to avoid sending "emerg"
4399 messages on all terminals on some default syslog configurations.
4400 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01004401
4402 emerg alert crit err warning notice info debug
4403
William Lallemand0f99e342011-10-12 17:50:54 +02004404 It is important to keep in mind that it is the frontend which decides what to
4405 log from a connection, and that in case of content switching, the log entries
4406 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01004407
4408 However, backend log declaration define how and where servers status changes
4409 will be logged. Level "notice" will be used to indicate a server going up,
4410 "warning" will be used for termination signals and definitive service
4411 termination, and "alert" will be used for when a server goes down.
4412
4413 Note : According to RFC3164, messages are truncated to 1024 bytes before
4414 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004415
4416 Example :
4417 log global
Willy Tarreauf7edefa2009-05-10 17:20:05 +02004418 log 127.0.0.1:514 local0 notice # only send important events
4419 log 127.0.0.1:514 local0 notice notice # same but limit output level
William Lallemandb2f07452015-05-12 14:27:13 +02004420 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
Willy Tarreaudad36a32013-03-11 01:20:04 +01004421
Willy Tarreau2769aa02007-12-27 18:26:09 +01004422
William Lallemand48940402012-01-30 16:47:22 +01004423log-format <string>
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01004424 Specifies the log format string to use for traffic logs
4425 May be used in sections: defaults | frontend | listen | backend
4426 yes | yes | yes | no
William Lallemand48940402012-01-30 16:47:22 +01004427
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01004428 This directive specifies the log format string that will be used for all logs
4429 resulting from traffic passing through the frontend using this line. If the
4430 directive is used in a defaults section, all subsequent frontends will use
4431 the same log format. Please see section 8.2.4 which covers the log format
4432 string in depth.
William Lallemand48940402012-01-30 16:47:22 +01004433
Dragan Dosen7ad31542015-09-28 17:16:47 +02004434log-format-sd <string>
4435 Specifies the RFC5424 structured-data log format string
4436 May be used in sections: defaults | frontend | listen | backend
4437 yes | yes | yes | no
4438
4439 This directive specifies the RFC5424 structured-data log format string that
4440 will be used for all logs resulting from traffic passing through the frontend
4441 using this line. If the directive is used in a defaults section, all
4442 subsequent frontends will use the same log format. Please see section 8.2.4
4443 which covers the log format string in depth.
4444
4445 See https://tools.ietf.org/html/rfc5424#section-6.3 for more information
4446 about the RFC5424 structured-data part.
4447
4448 Note : This log format string will be used only for loggers that have set
4449 log format to "rfc5424".
4450
4451 Example :
4452 log-format-sd [exampleSDID@1234\ bytes=\"%B\"\ status=\"%ST\"]
4453
4454
Willy Tarreau094af4e2015-01-07 15:03:42 +01004455log-tag <string>
4456 Specifies the log tag to use for all outgoing logs
4457 May be used in sections: defaults | frontend | listen | backend
4458 yes | yes | yes | yes
4459
4460 Sets the tag field in the syslog header to this string. It defaults to the
4461 log-tag set in the global section, otherwise the program name as launched
4462 from the command line, which usually is "haproxy". Sometimes it can be useful
4463 to differentiate between multiple processes running on the same host, or to
4464 differentiate customer instances running in the same process. In the backend,
4465 logs about servers up/down will use this tag. As a hint, it can be convenient
4466 to set a log-tag related to a hosted customer in a defaults section then put
4467 all the frontends and backends for that customer, then start another customer
4468 in a new defaults section. See also the global "log-tag" directive.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004469
Willy Tarreauc35362a2014-04-25 13:58:37 +02004470max-keep-alive-queue <value>
4471 Set the maximum server queue size for maintaining keep-alive connections
4472 May be used in sections: defaults | frontend | listen | backend
4473 yes | no | yes | yes
4474
4475 HTTP keep-alive tries to reuse the same server connection whenever possible,
4476 but sometimes it can be counter-productive, for example if a server has a lot
4477 of connections while other ones are idle. This is especially true for static
4478 servers.
4479
4480 The purpose of this setting is to set a threshold on the number of queued
4481 connections at which haproxy stops trying to reuse the same server and prefers
4482 to find another one. The default value, -1, means there is no limit. A value
4483 of zero means that keep-alive requests will never be queued. For very close
4484 servers which can be reached with a low latency and which are not sensible to
4485 breaking keep-alive, a low value is recommended (eg: local static server can
4486 use a value of 10 or less). For remote servers suffering from a high latency,
4487 higher values might be needed to cover for the latency and/or the cost of
4488 picking a different server.
4489
4490 Note that this has no impact on responses which are maintained to the same
4491 server consecutively to a 401 response. They will still go to the same server
4492 even if they have to be queued.
4493
4494 See also : "option http-server-close", "option prefer-last-server", server
4495 "maxconn" and cookie persistence.
4496
4497
Willy Tarreau2769aa02007-12-27 18:26:09 +01004498maxconn <conns>
4499 Fix the maximum number of concurrent connections on a frontend
4500 May be used in sections : defaults | frontend | listen | backend
4501 yes | yes | yes | no
4502 Arguments :
4503 <conns> is the maximum number of concurrent connections the frontend will
4504 accept to serve. Excess connections will be queued by the system
4505 in the socket's listen queue and will be served once a connection
4506 closes.
4507
4508 If the system supports it, it can be useful on big sites to raise this limit
4509 very high so that haproxy manages connection queues, instead of leaving the
4510 clients with unanswered connection attempts. This value should not exceed the
4511 global maxconn. Also, keep in mind that a connection contains two buffers
4512 of 8kB each, as well as some other data resulting in about 17 kB of RAM being
4513 consumed per established connection. That means that a medium system equipped
4514 with 1GB of RAM can withstand around 40000-50000 concurrent connections if
4515 properly tuned.
4516
4517 Also, when <conns> is set to large values, it is possible that the servers
4518 are not sized to accept such loads, and for this reason it is generally wise
4519 to assign them some reasonable connection limits.
4520
Vincent Bernat6341be52012-06-27 17:18:30 +02004521 By default, this value is set to 2000.
4522
Willy Tarreau2769aa02007-12-27 18:26:09 +01004523 See also : "server", global section's "maxconn", "fullconn"
4524
4525
4526mode { tcp|http|health }
4527 Set the running mode or protocol of the instance
4528 May be used in sections : defaults | frontend | listen | backend
4529 yes | yes | yes | yes
4530 Arguments :
4531 tcp The instance will work in pure TCP mode. A full-duplex connection
4532 will be established between clients and servers, and no layer 7
4533 examination will be performed. This is the default mode. It
4534 should be used for SSL, SSH, SMTP, ...
4535
4536 http The instance will work in HTTP mode. The client request will be
4537 analyzed in depth before connecting to any server. Any request
4538 which is not RFC-compliant will be rejected. Layer 7 filtering,
4539 processing and switching will be possible. This is the mode which
4540 brings HAProxy most of its value.
4541
4542 health The instance will work in "health" mode. It will just reply "OK"
Willy Tarreau82569f92012-09-27 23:48:56 +02004543 to incoming connections and close the connection. Alternatively,
4544 If the "httpchk" option is set, "HTTP/1.0 200 OK" will be sent
4545 instead. Nothing will be logged in either case. This mode is used
4546 to reply to external components health checks. This mode is
4547 deprecated and should not be used anymore as it is possible to do
4548 the same and even better by combining TCP or HTTP modes with the
4549 "monitor" keyword.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004550
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004551 When doing content switching, it is mandatory that the frontend and the
4552 backend are in the same mode (generally HTTP), otherwise the configuration
4553 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004554
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004555 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01004556 defaults http_instances
4557 mode http
4558
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004559 See also : "monitor", "monitor-net"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004560
Willy Tarreau0ba27502007-12-24 16:55:16 +01004561
Cyril Bontéf0c60612010-02-06 14:44:47 +01004562monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01004563 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004564 May be used in sections : defaults | frontend | listen | backend
4565 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01004566 Arguments :
4567 if <cond> the monitor request will fail if the condition is satisfied,
4568 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004569 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01004570 are met, for instance a low number of servers both in a
4571 backend and its backup.
4572
4573 unless <cond> the monitor request will succeed only if the condition is
4574 satisfied, and will fail otherwise. Such a condition may be
4575 based on a test on the presence of a minimum number of active
4576 servers in a list of backends.
4577
4578 This statement adds a condition which can force the response to a monitor
4579 request to report a failure. By default, when an external component queries
4580 the URI dedicated to monitoring, a 200 response is returned. When one of the
4581 conditions above is met, haproxy will return 503 instead of 200. This is
4582 very useful to report a site failure to an external component which may base
4583 routing advertisements between multiple sites on the availability reported by
4584 haproxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02004585 criterion. Note that "monitor fail" only works in HTTP mode. Both status
4586 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004587
4588 Example:
4589 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01004590 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01004591 acl site_dead nbsrv(dynamic) lt 2
4592 acl site_dead nbsrv(static) lt 2
4593 monitor-uri /site_alive
4594 monitor fail if site_dead
4595
Willy Tarreauae94d4d2011-05-11 16:28:49 +02004596 See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004597
4598
4599monitor-net <source>
4600 Declare a source network which is limited to monitor requests
4601 May be used in sections : defaults | frontend | listen | backend
4602 yes | yes | yes | no
4603 Arguments :
4604 <source> is the source IPv4 address or network which will only be able to
4605 get monitor responses to any request. It can be either an IPv4
4606 address, a host name, or an address followed by a slash ('/')
4607 followed by a mask.
4608
4609 In TCP mode, any connection coming from a source matching <source> will cause
4610 the connection to be immediately closed without any log. This allows another
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004611 equipment to probe the port and verify that it is still listening, without
Willy Tarreau2769aa02007-12-27 18:26:09 +01004612 forwarding the connection to a remote server.
4613
4614 In HTTP mode, a connection coming from a source matching <source> will be
4615 accepted, the following response will be sent without waiting for a request,
4616 then the connection will be closed : "HTTP/1.0 200 OK". This is normally
4617 enough for any front-end HTTP probe to detect that the service is UP and
Willy Tarreau82569f92012-09-27 23:48:56 +02004618 running without forwarding the request to a backend server. Note that this
4619 response is sent in raw format, without any transformation. This is important
4620 as it means that it will not be SSL-encrypted on SSL listeners.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004621
Willy Tarreau82569f92012-09-27 23:48:56 +02004622 Monitor requests are processed very early, just after tcp-request connection
4623 ACLs which are the only ones able to block them. These connections are short
4624 lived and never wait for any data from the client. They cannot be logged, and
4625 it is the intended purpose. They are only used to report HAProxy's health to
4626 an upper component, nothing more. Please note that "monitor fail" rules do
4627 not apply to connections intercepted by "monitor-net".
Willy Tarreau2769aa02007-12-27 18:26:09 +01004628
Willy Tarreau95cd2832010-03-04 23:36:33 +01004629 Last, please note that only one "monitor-net" statement can be specified in
4630 a frontend. If more than one is found, only the last one will be considered.
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004631
Willy Tarreau2769aa02007-12-27 18:26:09 +01004632 Example :
4633 # addresses .252 and .253 are just probing us.
4634 frontend www
4635 monitor-net 192.168.0.252/31
4636
4637 See also : "monitor fail", "monitor-uri"
4638
4639
4640monitor-uri <uri>
4641 Intercept a URI used by external components' monitor requests
4642 May be used in sections : defaults | frontend | listen | backend
4643 yes | yes | yes | no
4644 Arguments :
4645 <uri> is the exact URI which we want to intercept to return HAProxy's
4646 health status instead of forwarding the request.
4647
4648 When an HTTP request referencing <uri> will be received on a frontend,
4649 HAProxy will not forward it nor log it, but instead will return either
4650 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
4651 conditions defined with "monitor fail". This is normally enough for any
4652 front-end HTTP probe to detect that the service is UP and running without
4653 forwarding the request to a backend server. Note that the HTTP method, the
4654 version and all headers are ignored, but the request must at least be valid
4655 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
4656
4657 Monitor requests are processed very early. It is not possible to block nor
4658 divert them using ACLs. They cannot be logged either, and it is the intended
4659 purpose. They are only used to report HAProxy's health to an upper component,
4660 nothing more. However, it is possible to add any number of conditions using
4661 "monitor fail" and ACLs so that the result can be adjusted to whatever check
4662 can be imagined (most often the number of available servers in a backend).
4663
4664 Example :
4665 # Use /haproxy_test to report haproxy's status
4666 frontend www
4667 mode http
4668 monitor-uri /haproxy_test
4669
4670 See also : "monitor fail", "monitor-net"
4671
Willy Tarreau0ba27502007-12-24 16:55:16 +01004672
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004673option abortonclose
4674no option abortonclose
4675 Enable or disable early dropping of aborted requests pending in queues.
4676 May be used in sections : defaults | frontend | listen | backend
4677 yes | no | yes | yes
4678 Arguments : none
4679
4680 In presence of very high loads, the servers will take some time to respond.
4681 The per-instance connection queue will inflate, and the response time will
4682 increase respective to the size of the queue times the average per-session
4683 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01004684 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004685 the queue, and slowing down other users, and the servers as well, because the
4686 request will eventually be served, then aborted at the first error
4687 encountered while delivering the response.
4688
4689 As there is no way to distinguish between a full STOP and a simple output
4690 close on the client side, HTTP agents should be conservative and consider
4691 that the client might only have closed its output channel while waiting for
4692 the response. However, this introduces risks of congestion when lots of users
4693 do the same, and is completely useless nowadays because probably no client at
4694 all will close the session while waiting for the response. Some HTTP agents
Willy Tarreaud72758d2010-01-12 10:42:19 +01004695 support this behaviour (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004696 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01004697 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004698 of being the single component to break rare but valid traffic is extremely
4699 low, which adds to the temptation to be able to abort a session early while
4700 still not served and not pollute the servers.
4701
4702 In HAProxy, the user can choose the desired behaviour using the option
4703 "abortonclose". By default (without the option) the behaviour is HTTP
4704 compliant and aborted requests will be served. But when the option is
4705 specified, a session with an incoming channel closed will be aborted while
4706 it is still possible, either pending in the queue for a connection slot, or
4707 during the connection establishment if the server has not yet acknowledged
4708 the connection request. This considerably reduces the queue size and the load
4709 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01004710 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004711
4712 If this option has been enabled in a "defaults" section, it can be disabled
4713 in a specific instance by prepending the "no" keyword before it.
4714
4715 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
4716
4717
Willy Tarreau4076a152009-04-02 15:18:36 +02004718option accept-invalid-http-request
4719no option accept-invalid-http-request
4720 Enable or disable relaxing of HTTP request parsing
4721 May be used in sections : defaults | frontend | listen | backend
4722 yes | yes | yes | no
4723 Arguments : none
4724
Willy Tarreau91852eb2015-05-01 13:26:00 +02004725 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02004726 means that invalid characters in header names are not permitted and cause an
4727 error to be returned to the client. This is the desired behaviour as such
4728 forbidden characters are essentially used to build attacks exploiting server
4729 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
4730 server will emit invalid header names for whatever reason (configuration,
4731 implementation) and the issue will not be immediately fixed. In such a case,
4732 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01004733 even if that does not make sense, by specifying this option. Similarly, the
4734 list of characters allowed to appear in a URI is well defined by RFC3986, and
4735 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
4736 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
4737 not allowed at all. Haproxy always blocks a number of them (0..32, 127). The
Willy Tarreau91852eb2015-05-01 13:26:00 +02004738 remaining ones are blocked by default unless this option is enabled. This
Willy Tarreau13317662015-05-01 13:47:08 +02004739 option also relaxes the test on the HTTP version, it allows HTTP/0.9 requests
4740 to pass through (no version specified) and multiple digits for both the major
4741 and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02004742
4743 This option should never be enabled by default as it hides application bugs
4744 and open security breaches. It should only be deployed after a problem has
4745 been confirmed.
4746
4747 When this option is enabled, erroneous header names will still be accepted in
4748 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01004749 analysis using the "show errors" request on the UNIX stats socket. Similarly,
4750 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02004751 also helps confirming that the issue has been solved.
4752
4753 If this option has been enabled in a "defaults" section, it can be disabled
4754 in a specific instance by prepending the "no" keyword before it.
4755
4756 See also : "option accept-invalid-http-response" and "show errors" on the
4757 stats socket.
4758
4759
4760option accept-invalid-http-response
4761no option accept-invalid-http-response
4762 Enable or disable relaxing of HTTP response parsing
4763 May be used in sections : defaults | frontend | listen | backend
4764 yes | no | yes | yes
4765 Arguments : none
4766
Willy Tarreau91852eb2015-05-01 13:26:00 +02004767 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02004768 means that invalid characters in header names are not permitted and cause an
4769 error to be returned to the client. This is the desired behaviour as such
4770 forbidden characters are essentially used to build attacks exploiting server
4771 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
4772 server will emit invalid header names for whatever reason (configuration,
4773 implementation) and the issue will not be immediately fixed. In such a case,
4774 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau91852eb2015-05-01 13:26:00 +02004775 even if that does not make sense, by specifying this option. This option also
4776 relaxes the test on the HTTP version format, it allows multiple digits for
4777 both the major and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02004778
4779 This option should never be enabled by default as it hides application bugs
4780 and open security breaches. It should only be deployed after a problem has
4781 been confirmed.
4782
4783 When this option is enabled, erroneous header names will still be accepted in
4784 responses, but the complete response will be captured in order to permit
4785 later analysis using the "show errors" request on the UNIX stats socket.
4786 Doing this also helps confirming that the issue has been solved.
4787
4788 If this option has been enabled in a "defaults" section, it can be disabled
4789 in a specific instance by prepending the "no" keyword before it.
4790
4791 See also : "option accept-invalid-http-request" and "show errors" on the
4792 stats socket.
4793
4794
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004795option allbackups
4796no option allbackups
4797 Use either all backup servers at a time or only the first one
4798 May be used in sections : defaults | frontend | listen | backend
4799 yes | no | yes | yes
4800 Arguments : none
4801
4802 By default, the first operational backup server gets all traffic when normal
4803 servers are all down. Sometimes, it may be preferred to use multiple backups
4804 at once, because one will not be enough. When "option allbackups" is enabled,
4805 the load balancing will be performed among all backup servers when all normal
4806 ones are unavailable. The same load balancing algorithm will be used and the
4807 servers' weights will be respected. Thus, there will not be any priority
4808 order between the backup servers anymore.
4809
4810 This option is mostly used with static server farms dedicated to return a
4811 "sorry" page when an application is completely offline.
4812
4813 If this option has been enabled in a "defaults" section, it can be disabled
4814 in a specific instance by prepending the "no" keyword before it.
4815
4816
4817option checkcache
4818no option checkcache
Godbach7056a352013-12-11 20:01:07 +08004819 Analyze all server responses and block responses with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004820 May be used in sections : defaults | frontend | listen | backend
4821 yes | no | yes | yes
4822 Arguments : none
4823
4824 Some high-level frameworks set application cookies everywhere and do not
4825 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004826 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004827 high risk of session crossing or stealing between users traversing the same
4828 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02004829 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004830
4831 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004832 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01004833 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004834 response to check if there's a risk of caching a cookie on a client-side
4835 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01004836 to the client are :
4837 - all those without "Set-Cookie" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004838 - all those with a return code other than 200, 203, 206, 300, 301, 410,
Willy Tarreau198a7442008-01-17 12:05:32 +01004839 provided that the server has not set a "Cache-control: public" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004840 - all those that come from a POST request, provided that the server has not
4841 set a 'Cache-Control: public' header ;
4842 - those with a 'Pragma: no-cache' header
4843 - those with a 'Cache-control: private' header
4844 - those with a 'Cache-control: no-store' header
4845 - those with a 'Cache-control: max-age=0' header
4846 - those with a 'Cache-control: s-maxage=0' header
4847 - those with a 'Cache-control: no-cache' header
4848 - those with a 'Cache-control: no-cache="set-cookie"' header
4849 - those with a 'Cache-control: no-cache="set-cookie,' header
4850 (allowing other fields after set-cookie)
4851
4852 If a response doesn't respect these requirements, then it will be blocked
Willy Tarreau198a7442008-01-17 12:05:32 +01004853 just as if it was from an "rspdeny" filter, with an "HTTP 502 bad gateway".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004854 The session state shows "PH--" meaning that the proxy blocked the response
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004855 during headers processing. Additionally, an alert will be sent in the logs so
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004856 that admins are informed that there's something to be fixed.
4857
4858 Due to the high impact on the application, the application should be tested
4859 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01004860 good practice to always activate it during tests, even if it is not used in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004861 production, as it will report potentially dangerous application behaviours.
4862
4863 If this option has been enabled in a "defaults" section, it can be disabled
4864 in a specific instance by prepending the "no" keyword before it.
4865
4866
4867option clitcpka
4868no option clitcpka
4869 Enable or disable the sending of TCP keepalive packets on the client side
4870 May be used in sections : defaults | frontend | listen | backend
4871 yes | yes | yes | no
4872 Arguments : none
4873
4874 When there is a firewall or any session-aware component between a client and
4875 a server, and when the protocol involves very long sessions with long idle
4876 periods (eg: remote desktops), there is a risk that one of the intermediate
4877 components decides to expire a session which has remained idle for too long.
4878
4879 Enabling socket-level TCP keep-alives makes the system regularly send packets
4880 to the other end of the connection, leaving it active. The delay between
4881 keep-alive probes is controlled by the system only and depends both on the
4882 operating system and its tuning parameters.
4883
4884 It is important to understand that keep-alive packets are neither emitted nor
4885 received at the application level. It is only the network stacks which sees
4886 them. For this reason, even if one side of the proxy already uses keep-alives
4887 to maintain its connection alive, those keep-alive packets will not be
4888 forwarded to the other side of the proxy.
4889
4890 Please note that this has nothing to do with HTTP keep-alive.
4891
4892 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
4893 client side of a connection, which should help when session expirations are
4894 noticed between HAProxy and a client.
4895
4896 If this option has been enabled in a "defaults" section, it can be disabled
4897 in a specific instance by prepending the "no" keyword before it.
4898
4899 See also : "option srvtcpka", "option tcpka"
4900
4901
Willy Tarreau0ba27502007-12-24 16:55:16 +01004902option contstats
4903 Enable continuous traffic statistics updates
4904 May be used in sections : defaults | frontend | listen | backend
4905 yes | yes | yes | no
4906 Arguments : none
4907
4908 By default, counters used for statistics calculation are incremented
4909 only when a session finishes. It works quite well when serving small
4910 objects, but with big ones (for example large images or archives) or
4911 with A/V streaming, a graph generated from haproxy counters looks like
4912 a hedgehog. With this option enabled counters get incremented continuously,
4913 during a whole session. Recounting touches a hotpath directly so
4914 it is not enabled by default, as it has small performance impact (~0.5%).
4915
4916
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02004917option dontlog-normal
4918no option dontlog-normal
4919 Enable or disable logging of normal, successful connections
4920 May be used in sections : defaults | frontend | listen | backend
4921 yes | yes | yes | no
4922 Arguments : none
4923
4924 There are large sites dealing with several thousand connections per second
4925 and for which logging is a major pain. Some of them are even forced to turn
4926 logs off and cannot debug production issues. Setting this option ensures that
4927 normal connections, those which experience no error, no timeout, no retry nor
4928 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
4929 mode, the response status code is checked and return codes 5xx will still be
4930 logged.
4931
4932 It is strongly discouraged to use this option as most of the time, the key to
4933 complex issues is in the normal logs which will not be logged here. If you
4934 need to separate logs, see the "log-separate-errors" option instead.
4935
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004936 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02004937 logging.
4938
4939
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004940option dontlognull
4941no option dontlognull
4942 Enable or disable logging of null connections
4943 May be used in sections : defaults | frontend | listen | backend
4944 yes | yes | yes | no
4945 Arguments : none
4946
4947 In certain environments, there are components which will regularly connect to
4948 various systems to ensure that they are still alive. It can be the case from
4949 another load balancer as well as from monitoring systems. By default, even a
4950 simple port probe or scan will produce a log. If those connections pollute
4951 the logs too much, it is possible to enable option "dontlognull" to indicate
4952 that a connection on which no data has been transferred will not be logged,
Willy Tarreau0f228a02015-05-01 15:37:53 +02004953 which typically corresponds to those probes. Note that errors will still be
4954 returned to the client and accounted for in the stats. If this is not what is
4955 desired, option http-ignore-probes can be used instead.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004956
4957 It is generally recommended not to use this option in uncontrolled
4958 environments (eg: internet), otherwise scans and other malicious activities
4959 would not be logged.
4960
4961 If this option has been enabled in a "defaults" section, it can be disabled
4962 in a specific instance by prepending the "no" keyword before it.
4963
Willy Tarreau0f228a02015-05-01 15:37:53 +02004964 See also : "log", "http-ignore-probes", "monitor-net", "monitor-uri", and
4965 section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004966
4967
4968option forceclose
4969no option forceclose
4970 Enable or disable active connection closing after response is transferred.
4971 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaua31e5df2009-12-30 01:10:35 +01004972 yes | yes | yes | yes
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004973 Arguments : none
4974
4975 Some HTTP servers do not necessarily close the connections when they receive
4976 the "Connection: close" set by "option httpclose", and if the client does not
4977 close either, then the connection remains open till the timeout expires. This
4978 causes high number of simultaneous connections on the servers and shows high
4979 global session times in the logs.
4980
4981 When this happens, it is possible to use "option forceclose". It will
Willy Tarreau82eeaf22009-12-29 12:09:05 +01004982 actively close the outgoing server channel as soon as the server has finished
Cyril Bonté653dcd62014-02-20 00:13:15 +01004983 to respond and release some resources earlier than with "option httpclose".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004984
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004985 This option may also be combined with "option http-pretend-keepalive", which
4986 will disable sending of the "Connection: close" header, but will still cause
4987 the connection to be closed once the whole response is received.
4988
Cyril Bonté653dcd62014-02-20 00:13:15 +01004989 This option disables and replaces any previous "option httpclose", "option
4990 http-server-close", "option http-keep-alive", or "option http-tunnel".
Willy Tarreau02bce8b2014-01-30 00:15:28 +01004991
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004992 If this option has been enabled in a "defaults" section, it can be disabled
4993 in a specific instance by prepending the "no" keyword before it.
4994
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004995 See also : "option httpclose" and "option http-pretend-keepalive"
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004996
4997
Willy Tarreau87cf5142011-08-19 22:57:24 +02004998option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01004999 Enable insertion of the X-Forwarded-For header to requests sent to servers
5000 May be used in sections : defaults | frontend | listen | backend
5001 yes | yes | yes | yes
5002 Arguments :
5003 <network> is an optional argument used to disable this option for sources
5004 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02005005 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01005006 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005007
5008 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
5009 their client address. This is sometimes annoying when the client's IP address
5010 is expected in server logs. To solve this problem, the well-known HTTP header
5011 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
5012 This header contains a value representing the client's IP address. Since this
5013 header is always appended at the end of the existing header list, the server
5014 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02005015 the server's manual to find how to enable use of this standard header. Note
5016 that only the last occurrence of the header must be used, since it is really
5017 possible that the client has already brought one.
5018
Willy Tarreaud72758d2010-01-12 10:42:19 +01005019 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02005020 the default "X-Forwarded-For". This can be useful where you might already
Willy Tarreaud72758d2010-01-12 10:42:19 +01005021 have a "X-Forwarded-For" header from a different application (eg: stunnel),
5022 and you need preserve it. Also if your backend server doesn't use the
Ross Westaf72a1d2008-08-03 10:51:45 +02005023 "X-Forwarded-For" header and requires different one (eg: Zeus Web Servers
5024 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01005025
5026 Sometimes, a same HAProxy instance may be shared between a direct client
5027 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
5028 used to decrypt HTTPS traffic). It is possible to disable the addition of the
5029 header for a known source address or network by adding the "except" keyword
5030 followed by the network address. In this case, any source IP matching the
5031 network will not cause an addition of this header. Most common uses are with
5032 private networks or 127.0.0.1.
5033
Willy Tarreau87cf5142011-08-19 22:57:24 +02005034 Alternatively, the keyword "if-none" states that the header will only be
5035 added if it is not present. This should only be used in perfectly trusted
5036 environment, as this might cause a security issue if headers reaching haproxy
5037 are under the control of the end-user.
5038
Willy Tarreauc27debf2008-01-06 08:57:02 +01005039 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02005040 least one of them uses it, the header will be added. Note that the backend's
5041 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02005042 both are defined. In the case of the "if-none" argument, if at least one of
5043 the frontend or the backend does not specify it, it wants the addition to be
5044 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005045
Ross Westaf72a1d2008-08-03 10:51:45 +02005046 Examples :
Willy Tarreauc27debf2008-01-06 08:57:02 +01005047 # Public HTTP address also used by stunnel on the same machine
5048 frontend www
5049 mode http
5050 option forwardfor except 127.0.0.1 # stunnel already adds the header
5051
Ross Westaf72a1d2008-08-03 10:51:45 +02005052 # Those servers want the IP Address in X-Client
5053 backend www
5054 mode http
5055 option forwardfor header X-Client
5056
Willy Tarreau87cf5142011-08-19 22:57:24 +02005057 See also : "option httpclose", "option http-server-close",
Willy Tarreau70dffda2014-01-30 03:07:23 +01005058 "option forceclose", "option http-keep-alive"
Willy Tarreauc27debf2008-01-06 08:57:02 +01005059
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005060
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005061option http-buffer-request
5062no option http-buffer-request
5063 Enable or disable waiting for whole HTTP request body before proceeding
5064 May be used in sections : defaults | frontend | listen | backend
5065 yes | yes | yes | yes
5066 Arguments : none
5067
5068 It is sometimes desirable to wait for the body of an HTTP request before
5069 taking a decision. This is what is being done by "balance url_param" for
5070 example. The first use case is to buffer requests from slow clients before
5071 connecting to the server. Another use case consists in taking the routing
5072 decision based on the request body's contents. This option placed in a
5073 frontend or backend forces the HTTP processing to wait until either the whole
5074 body is received, or the request buffer is full, or the first chunk is
5075 complete in case of chunked encoding. It can have undesired side effects with
5076 some applications abusing HTTP by expecting unbufferred transmissions between
5077 the frontend and the backend, so this should definitely not be used by
5078 default.
5079
Baptiste Assmanneccdf432015-10-28 13:49:01 +01005080 See also : "option http-no-delay", "timeout http-request"
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005081
5082
Willy Tarreau0f228a02015-05-01 15:37:53 +02005083option http-ignore-probes
5084no option http-ignore-probes
5085 Enable or disable logging of null connections and request timeouts
5086 May be used in sections : defaults | frontend | listen | backend
5087 yes | yes | yes | no
5088 Arguments : none
5089
5090 Recently some browsers started to implement a "pre-connect" feature
5091 consisting in speculatively connecting to some recently visited web sites
5092 just in case the user would like to visit them. This results in many
5093 connections being established to web sites, which end up in 408 Request
5094 Timeout if the timeout strikes first, or 400 Bad Request when the browser
5095 decides to close them first. These ones pollute the log and feed the error
5096 counters. There was already "option dontlognull" but it's insufficient in
5097 this case. Instead, this option does the following things :
5098 - prevent any 400/408 message from being sent to the client if nothing
5099 was received over a connection before it was closed ;
5100 - prevent any log from being emitted in this situation ;
5101 - prevent any error counter from being incremented
5102
5103 That way the empty connection is silently ignored. Note that it is better
5104 not to use this unless it is clear that it is needed, because it will hide
5105 real problems. The most common reason for not receiving a request and seeing
5106 a 408 is due to an MTU inconsistency between the client and an intermediary
5107 element such as a VPN, which blocks too large packets. These issues are
5108 generally seen with POST requests as well as GET with large cookies. The logs
5109 are often the only way to detect them.
5110
5111 If this option has been enabled in a "defaults" section, it can be disabled
5112 in a specific instance by prepending the "no" keyword before it.
5113
5114 See also : "log", "dontlognull", "errorfile", and section 8 about logging.
5115
5116
Willy Tarreau16bfb022010-01-16 19:48:41 +01005117option http-keep-alive
5118no option http-keep-alive
5119 Enable or disable HTTP keep-alive from client to server
5120 May be used in sections : defaults | frontend | listen | backend
5121 yes | yes | yes | yes
5122 Arguments : none
5123
Willy Tarreau70dffda2014-01-30 03:07:23 +01005124 By default HAProxy operates in keep-alive mode with regards to persistent
5125 connections: for each connection it processes each request and response, and
5126 leaves the connection idle on both sides between the end of a response and the
5127 start of a new request. This mode may be changed by several options such as
5128 "option http-server-close", "option forceclose", "option httpclose" or
5129 "option http-tunnel". This option allows to set back the keep-alive mode,
5130 which can be useful when another mode was used in a defaults section.
5131
5132 Setting "option http-keep-alive" enables HTTP keep-alive mode on the client-
5133 and server- sides. This provides the lowest latency on the client side (slow
Willy Tarreau16bfb022010-01-16 19:48:41 +01005134 network) and the fastest session reuse on the server side at the expense
5135 of maintaining idle connections to the servers. In general, it is possible
5136 with this option to achieve approximately twice the request rate that the
5137 "http-server-close" option achieves on small objects. There are mainly two
5138 situations where this option may be useful :
5139
5140 - when the server is non-HTTP compliant and authenticates the connection
5141 instead of requests (eg: NTLM authentication)
5142
5143 - when the cost of establishing the connection to the server is significant
5144 compared to the cost of retrieving the associated object from the server.
5145
5146 This last case can happen when the server is a fast static server of cache.
5147 In this case, the server will need to be properly tuned to support high enough
5148 connection counts because connections will last until the client sends another
5149 request.
5150
5151 If the client request has to go to another backend or another server due to
5152 content switching or the load balancing algorithm, the idle connection will
Willy Tarreau9420b122013-12-15 18:58:25 +01005153 immediately be closed and a new one re-opened. Option "prefer-last-server" is
5154 available to try optimize server selection so that if the server currently
5155 attached to an idle connection is usable, it will be used.
Willy Tarreau16bfb022010-01-16 19:48:41 +01005156
5157 In general it is preferred to use "option http-server-close" with application
5158 servers, and some static servers might benefit from "option http-keep-alive".
5159
5160 At the moment, logs will not indicate whether requests came from the same
5161 session or not. The accept date reported in the logs corresponds to the end
5162 of the previous request, and the request time corresponds to the time spent
5163 waiting for a new request. The keep-alive request time is still bound to the
5164 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
5165 not set.
5166
Cyril Bonté653dcd62014-02-20 00:13:15 +01005167 This option disables and replaces any previous "option httpclose", "option
5168 http-server-close", "option forceclose" or "option http-tunnel". When backend
Willy Tarreau70dffda2014-01-30 03:07:23 +01005169 and frontend options differ, all of these 4 options have precedence over
Cyril Bonté653dcd62014-02-20 00:13:15 +01005170 "option http-keep-alive".
Willy Tarreau16bfb022010-01-16 19:48:41 +01005171
5172 See also : "option forceclose", "option http-server-close",
Willy Tarreau9420b122013-12-15 18:58:25 +01005173 "option prefer-last-server", "option http-pretend-keepalive",
5174 "option httpclose", and "1.1. The HTTP transaction model".
Willy Tarreau16bfb022010-01-16 19:48:41 +01005175
5176
Willy Tarreau96e31212011-05-30 18:10:30 +02005177option http-no-delay
5178no option http-no-delay
5179 Instruct the system to favor low interactive delays over performance in HTTP
5180 May be used in sections : defaults | frontend | listen | backend
5181 yes | yes | yes | yes
5182 Arguments : none
5183
5184 In HTTP, each payload is unidirectional and has no notion of interactivity.
5185 Any agent is expected to queue data somewhat for a reasonably low delay.
5186 There are some very rare server-to-server applications that abuse the HTTP
5187 protocol and expect the payload phase to be highly interactive, with many
5188 interleaved data chunks in both directions within a single request. This is
5189 absolutely not supported by the HTTP specification and will not work across
5190 most proxies or servers. When such applications attempt to do this through
5191 haproxy, it works but they will experience high delays due to the network
5192 optimizations which favor performance by instructing the system to wait for
5193 enough data to be available in order to only send full packets. Typical
5194 delays are around 200 ms per round trip. Note that this only happens with
5195 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
5196 affected.
5197
5198 When "option http-no-delay" is present in either the frontend or the backend
5199 used by a connection, all such optimizations will be disabled in order to
5200 make the exchanges as fast as possible. Of course this offers no guarantee on
5201 the functionality, as it may break at any other place. But if it works via
5202 HAProxy, it will work as fast as possible. This option should never be used
5203 by default, and should never be used at all unless such a buggy application
5204 is discovered. The impact of using this option is an increase of bandwidth
5205 usage and CPU usage, which may significantly lower performance in high
5206 latency environments.
5207
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02005208 See also : "option http-buffer-request"
5209
Willy Tarreau96e31212011-05-30 18:10:30 +02005210
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005211option http-pretend-keepalive
5212no option http-pretend-keepalive
5213 Define whether haproxy will announce keepalive to the server or not
5214 May be used in sections : defaults | frontend | listen | backend
5215 yes | yes | yes | yes
5216 Arguments : none
5217
5218 When running with "option http-server-close" or "option forceclose", haproxy
5219 adds a "Connection: close" header to the request forwarded to the server.
5220 Unfortunately, when some servers see this header, they automatically refrain
5221 from using the chunked encoding for responses of unknown length, while this
5222 is totally unrelated. The immediate effect is that this prevents haproxy from
5223 maintaining the client connection alive. A second effect is that a client or
5224 a cache could receive an incomplete response without being aware of it, and
5225 consider the response complete.
5226
5227 By setting "option http-pretend-keepalive", haproxy will make the server
5228 believe it will keep the connection alive. The server will then not fall back
5229 to the abnormal undesired above. When haproxy gets the whole response, it
5230 will close the connection with the server just as it would do with the
5231 "forceclose" option. That way the client gets a normal response and the
5232 connection is correctly closed on the server side.
5233
5234 It is recommended not to enable this option by default, because most servers
5235 will more efficiently close the connection themselves after the last packet,
5236 and release its buffers slightly earlier. Also, the added packet on the
5237 network could slightly reduce the overall peak performance. However it is
5238 worth noting that when this option is enabled, haproxy will have slightly
5239 less work to do. So if haproxy is the bottleneck on the whole architecture,
5240 enabling this option might save a few CPU cycles.
5241
5242 This option may be set both in a frontend and in a backend. It is enabled if
5243 at least one of the frontend or backend holding a connection has it enabled.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005244 This option may be combined with "option httpclose", which will cause
Willy Tarreau22a95342010-09-29 14:31:41 +02005245 keepalive to be announced to the server and close to be announced to the
5246 client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005247
5248 If this option has been enabled in a "defaults" section, it can be disabled
5249 in a specific instance by prepending the "no" keyword before it.
5250
Willy Tarreau16bfb022010-01-16 19:48:41 +01005251 See also : "option forceclose", "option http-server-close", and
5252 "option http-keep-alive"
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02005253
Willy Tarreauc27debf2008-01-06 08:57:02 +01005254
Willy Tarreaub608feb2010-01-02 22:47:18 +01005255option http-server-close
5256no option http-server-close
5257 Enable or disable HTTP connection closing on the server side
5258 May be used in sections : defaults | frontend | listen | backend
5259 yes | yes | yes | yes
5260 Arguments : none
5261
Willy Tarreau70dffda2014-01-30 03:07:23 +01005262 By default HAProxy operates in keep-alive mode with regards to persistent
5263 connections: for each connection it processes each request and response, and
5264 leaves the connection idle on both sides between the end of a response and
5265 the start of a new request. This mode may be changed by several options such
5266 as "option http-server-close", "option forceclose", "option httpclose" or
5267 "option http-tunnel". Setting "option http-server-close" enables HTTP
5268 connection-close mode on the server side while keeping the ability to support
5269 HTTP keep-alive and pipelining on the client side. This provides the lowest
5270 latency on the client side (slow network) and the fastest session reuse on
5271 the server side to save server resources, similarly to "option forceclose".
5272 It also permits non-keepalive capable servers to be served in keep-alive mode
5273 to the clients if they conform to the requirements of RFC2616. Please note
5274 that some servers do not always conform to those requirements when they see
5275 "Connection: close" in the request. The effect will be that keep-alive will
5276 never be used. A workaround consists in enabling "option
5277 http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01005278
5279 At the moment, logs will not indicate whether requests came from the same
5280 session or not. The accept date reported in the logs corresponds to the end
5281 of the previous request, and the request time corresponds to the time spent
5282 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01005283 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
5284 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01005285
5286 This option may be set both in a frontend and in a backend. It is enabled if
5287 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01005288 It disables and replaces any previous "option httpclose", "option forceclose",
5289 "option http-tunnel" or "option http-keep-alive". Please check section 4
Willy Tarreau70dffda2014-01-30 03:07:23 +01005290 ("Proxies") to see how this option combines with others when frontend and
5291 backend options differ.
Willy Tarreaub608feb2010-01-02 22:47:18 +01005292
5293 If this option has been enabled in a "defaults" section, it can be disabled
5294 in a specific instance by prepending the "no" keyword before it.
5295
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02005296 See also : "option forceclose", "option http-pretend-keepalive",
Willy Tarreau16bfb022010-01-16 19:48:41 +01005297 "option httpclose", "option http-keep-alive", and
5298 "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01005299
5300
Willy Tarreau02bce8b2014-01-30 00:15:28 +01005301option http-tunnel
5302no option http-tunnel
5303 Disable or enable HTTP connection processing after first transaction
5304 May be used in sections : defaults | frontend | listen | backend
5305 yes | yes | yes | yes
5306 Arguments : none
5307
Willy Tarreau70dffda2014-01-30 03:07:23 +01005308 By default HAProxy operates in keep-alive mode with regards to persistent
5309 connections: for each connection it processes each request and response, and
5310 leaves the connection idle on both sides between the end of a response and
5311 the start of a new request. This mode may be changed by several options such
5312 as "option http-server-close", "option forceclose", "option httpclose" or
5313 "option http-tunnel".
5314
5315 Option "http-tunnel" disables any HTTP processing past the first request and
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005316 the first response. This is the mode which was used by default in versions
Willy Tarreau70dffda2014-01-30 03:07:23 +01005317 1.0 to 1.5-dev21. It is the mode with the lowest processing overhead, which
5318 is normally not needed anymore unless in very specific cases such as when
5319 using an in-house protocol that looks like HTTP but is not compatible, or
5320 just to log one request per client in order to reduce log size. Note that
5321 everything which works at the HTTP level, including header parsing/addition,
5322 cookie processing or content switching will only work for the first request
5323 and will be ignored after the first response.
Willy Tarreau02bce8b2014-01-30 00:15:28 +01005324
5325 If this option has been enabled in a "defaults" section, it can be disabled
5326 in a specific instance by prepending the "no" keyword before it.
5327
5328 See also : "option forceclose", "option http-server-close",
5329 "option httpclose", "option http-keep-alive", and
5330 "1.1. The HTTP transaction model".
5331
5332
Willy Tarreau88d349d2010-01-25 12:15:43 +01005333option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01005334no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01005335 Make use of non-standard Proxy-Connection header instead of Connection
5336 May be used in sections : defaults | frontend | listen | backend
5337 yes | yes | yes | no
5338 Arguments : none
5339
5340 While RFC2616 explicitly states that HTTP/1.1 agents must use the
5341 Connection header to indicate their wish of persistent or non-persistent
5342 connections, both browsers and proxies ignore this header for proxied
5343 connections and make use of the undocumented, non-standard Proxy-Connection
5344 header instead. The issue begins when trying to put a load balancer between
5345 browsers and such proxies, because there will be a difference between what
5346 haproxy understands and what the client and the proxy agree on.
5347
5348 By setting this option in a frontend, haproxy can automatically switch to use
5349 that non-standard header if it sees proxied requests. A proxied request is
5350 defined here as one where the URI begins with neither a '/' nor a '*'. The
5351 choice of header only affects requests passing through proxies making use of
5352 one of the "httpclose", "forceclose" and "http-server-close" options. Note
5353 that this option can only be specified in a frontend and will affect the
5354 request along its whole life.
5355
Willy Tarreau844a7e72010-01-31 21:46:18 +01005356 Also, when this option is set, a request which requires authentication will
5357 automatically switch to use proxy authentication headers if it is itself a
5358 proxied request. That makes it possible to check or enforce authentication in
5359 front of an existing proxy.
5360
Willy Tarreau88d349d2010-01-25 12:15:43 +01005361 This option should normally never be used, except in front of a proxy.
5362
5363 See also : "option httpclose", "option forceclose" and "option
5364 http-server-close".
5365
5366
Willy Tarreaud63335a2010-02-26 12:56:52 +01005367option httpchk
5368option httpchk <uri>
5369option httpchk <method> <uri>
5370option httpchk <method> <uri> <version>
5371 Enable HTTP protocol to check on the servers health
5372 May be used in sections : defaults | frontend | listen | backend
5373 yes | no | yes | yes
5374 Arguments :
5375 <method> is the optional HTTP method used with the requests. When not set,
5376 the "OPTIONS" method is used, as it generally requires low server
5377 processing and is easy to filter out from the logs. Any method
5378 may be used, though it is not recommended to invent non-standard
5379 ones.
5380
5381 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
5382 which is accessible by default on almost any server, but may be
5383 changed to any other URI. Query strings are permitted.
5384
5385 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
5386 but some servers might behave incorrectly in HTTP 1.0, so turning
5387 it to HTTP/1.1 may sometimes help. Note that the Host field is
5388 mandatory in HTTP/1.1, and as a trick, it is possible to pass it
5389 after "\r\n" following the version string.
5390
5391 By default, server health checks only consist in trying to establish a TCP
5392 connection. When "option httpchk" is specified, a complete HTTP request is
5393 sent once the TCP connection is established, and responses 2xx and 3xx are
5394 considered valid, while all other ones indicate a server failure, including
5395 the lack of any response.
5396
5397 The port and interval are specified in the server configuration.
5398
5399 This option does not necessarily require an HTTP backend, it also works with
5400 plain TCP backends. This is particularly useful to check simple scripts bound
5401 to some dedicated ports using the inetd daemon.
5402
5403 Examples :
5404 # Relay HTTPS traffic to Apache instance and check service availability
5405 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
5406 backend https_relay
5407 mode tcp
5408 option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www
5409 server apache1 192.168.1.1:443 check port 80
5410
Simon Hormanafc47ee2013-11-25 10:46:35 +09005411 See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",
5412 "option pgsql-check", "http-check" and the "check", "port" and
5413 "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01005414
5415
Willy Tarreauc27debf2008-01-06 08:57:02 +01005416option httpclose
5417no option httpclose
5418 Enable or disable passive HTTP connection closing
5419 May be used in sections : defaults | frontend | listen | backend
5420 yes | yes | yes | yes
5421 Arguments : none
5422
Willy Tarreau70dffda2014-01-30 03:07:23 +01005423 By default HAProxy operates in keep-alive mode with regards to persistent
5424 connections: for each connection it processes each request and response, and
5425 leaves the connection idle on both sides between the end of a response and
5426 the start of a new request. This mode may be changed by several options such
Cyril Bonté653dcd62014-02-20 00:13:15 +01005427 as "option http-server-close", "option forceclose", "option httpclose" or
Willy Tarreau70dffda2014-01-30 03:07:23 +01005428 "option http-tunnel".
5429
5430 If "option httpclose" is set, HAProxy will work in HTTP tunnel mode and check
5431 if a "Connection: close" header is already set in each direction, and will
5432 add one if missing. Each end should react to this by actively closing the TCP
5433 connection after each transfer, thus resulting in a switch to the HTTP close
5434 mode. Any "Connection" header different from "close" will also be removed.
5435 Note that this option is deprecated since what it does is very cheap but not
5436 reliable. Using "option http-server-close" or "option forceclose" is strongly
5437 recommended instead.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005438
5439 It seldom happens that some servers incorrectly ignore this header and do not
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005440 close the connection even though they reply "Connection: close". For this
Willy Tarreau0dfdf192010-01-05 11:33:11 +01005441 reason, they are not compatible with older HTTP 1.0 browsers. If this happens
5442 it is possible to use the "option forceclose" which actively closes the
5443 request connection once the server responds. Option "forceclose" also
5444 releases the server connection earlier because it does not have to wait for
5445 the client to acknowledge it.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005446
5447 This option may be set both in a frontend and in a backend. It is enabled if
5448 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01005449 It disables and replaces any previous "option http-server-close",
5450 "option forceclose", "option http-keep-alive" or "option http-tunnel". Please
Willy Tarreau70dffda2014-01-30 03:07:23 +01005451 check section 4 ("Proxies") to see how this option combines with others when
5452 frontend and backend options differ.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005453
5454 If this option has been enabled in a "defaults" section, it can be disabled
5455 in a specific instance by prepending the "no" keyword before it.
5456
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02005457 See also : "option forceclose", "option http-server-close" and
5458 "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01005459
5460
Emeric Brun3a058f32009-06-30 18:26:00 +02005461option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01005462 Enable logging of HTTP request, session state and timers
5463 May be used in sections : defaults | frontend | listen | backend
5464 yes | yes | yes | yes
Emeric Brun3a058f32009-06-30 18:26:00 +02005465 Arguments :
5466 clf if the "clf" argument is added, then the output format will be
5467 the CLF format instead of HAProxy's default HTTP format. You can
5468 use this when you need to feed HAProxy's logs through a specific
5469 log analyser which only support the CLF format and which is not
5470 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005471
5472 By default, the log output format is very poor, as it only contains the
5473 source and destination addresses, and the instance name. By specifying
5474 "option httplog", each log line turns into a much richer format including,
5475 but not limited to, the HTTP request, the connection timers, the session
5476 status, the connections numbers, the captured headers and cookies, the
5477 frontend, backend and server name, and of course the source address and
5478 ports.
5479
5480 This option may be set either in the frontend or the backend.
5481
PiBa-NLbd556bf2014-12-11 21:31:54 +01005482 Specifying only "option httplog" will automatically clear the 'clf' mode
5483 if it was set by default.
Emeric Brun3a058f32009-06-30 18:26:00 +02005484
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005485 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01005486
Willy Tarreau55165fe2009-05-10 12:02:55 +02005487
5488option http_proxy
5489no option http_proxy
5490 Enable or disable plain HTTP proxy mode
5491 May be used in sections : defaults | frontend | listen | backend
5492 yes | yes | yes | yes
5493 Arguments : none
5494
5495 It sometimes happens that people need a pure HTTP proxy which understands
5496 basic proxy requests without caching nor any fancy feature. In this case,
5497 it may be worth setting up an HAProxy instance with the "option http_proxy"
5498 set. In this mode, no server is declared, and the connection is forwarded to
5499 the IP address and port found in the URL after the "http://" scheme.
5500
5501 No host address resolution is performed, so this only works when pure IP
5502 addresses are passed. Since this option's usage perimeter is rather limited,
5503 it will probably be used only by experts who know they need exactly it. Last,
5504 if the clients are susceptible of sending keep-alive requests, it will be
Cyril Bonté2409e682010-12-14 22:47:51 +01005505 needed to add "option httpclose" to ensure that all requests will correctly
Willy Tarreau55165fe2009-05-10 12:02:55 +02005506 be analyzed.
5507
5508 If this option has been enabled in a "defaults" section, it can be disabled
5509 in a specific instance by prepending the "no" keyword before it.
5510
5511 Example :
5512 # this backend understands HTTP proxy requests and forwards them directly.
5513 backend direct_forward
5514 option httpclose
5515 option http_proxy
5516
5517 See also : "option httpclose"
5518
Willy Tarreau211ad242009-10-03 21:45:07 +02005519
Jamie Gloudon801a0a32012-08-25 00:18:33 -04005520option independent-streams
5521no option independent-streams
5522 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02005523 May be used in sections : defaults | frontend | listen | backend
5524 yes | yes | yes | yes
5525 Arguments : none
5526
5527 By default, when data is sent over a socket, both the write timeout and the
5528 read timeout for that socket are refreshed, because we consider that there is
5529 activity on that socket, and we have no other means of guessing if we should
5530 receive data or not.
5531
5532 While this default behaviour is desirable for almost all applications, there
5533 exists a situation where it is desirable to disable it, and only refresh the
5534 read timeout if there are incoming data. This happens on sessions with large
5535 timeouts and low amounts of exchanged data such as telnet session. If the
5536 server suddenly disappears, the output data accumulates in the system's
5537 socket buffers, both timeouts are correctly refreshed, and there is no way
5538 to know the server does not receive them, so we don't timeout. However, when
5539 the underlying protocol always echoes sent data, it would be enough by itself
5540 to detect the issue using the read timeout. Note that this problem does not
5541 happen with more verbose protocols because data won't accumulate long in the
5542 socket buffers.
5543
5544 When this option is set on the frontend, it will disable read timeout updates
5545 on data sent to the client. There probably is little use of this case. When
5546 the option is set on the backend, it will disable read timeout updates on
5547 data sent to the server. Doing so will typically break large HTTP posts from
5548 slow lines, so use it with caution.
5549
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005550 Note: older versions used to call this setting "option independent-streams"
Jamie Gloudon801a0a32012-08-25 00:18:33 -04005551 with a spelling mistake. This spelling is still supported but
5552 deprecated.
5553
Willy Tarreauce887fd2012-05-12 12:50:00 +02005554 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02005555
5556
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02005557option ldap-check
5558 Use LDAPv3 health checks for server testing
5559 May be used in sections : defaults | frontend | listen | backend
5560 yes | no | yes | yes
5561 Arguments : none
5562
5563 It is possible to test that the server correctly talks LDAPv3 instead of just
5564 testing that it accepts the TCP connection. When this option is set, an
5565 LDAPv3 anonymous simple bind message is sent to the server, and the response
5566 is analyzed to find an LDAPv3 bind response message.
5567
5568 The server is considered valid only when the LDAP response contains success
5569 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
5570
5571 Logging of bind requests is server dependent see your documentation how to
5572 configure it.
5573
5574 Example :
5575 option ldap-check
5576
5577 See also : "option httpchk"
5578
5579
Simon Horman98637e52014-06-20 12:30:16 +09005580option external-check
5581 Use external processes for server health checks
5582 May be used in sections : defaults | frontend | listen | backend
5583 yes | no | yes | yes
5584
5585 It is possible to test the health of a server using an external command.
5586 This is achieved by running the executable set using "external-check
5587 command".
5588
5589 Requires the "external-check" global to be set.
5590
5591 See also : "external-check", "external-check command", "external-check path"
5592
5593
Willy Tarreau211ad242009-10-03 21:45:07 +02005594option log-health-checks
5595no option log-health-checks
Willy Tarreaubef1b322014-05-13 21:01:39 +02005596 Enable or disable logging of health checks status updates
Willy Tarreau211ad242009-10-03 21:45:07 +02005597 May be used in sections : defaults | frontend | listen | backend
5598 yes | no | yes | yes
5599 Arguments : none
5600
Willy Tarreaubef1b322014-05-13 21:01:39 +02005601 By default, failed health check are logged if server is UP and successful
5602 health checks are logged if server is DOWN, so the amount of additional
5603 information is limited.
Willy Tarreau211ad242009-10-03 21:45:07 +02005604
Willy Tarreaubef1b322014-05-13 21:01:39 +02005605 When this option is enabled, any change of the health check status or to
5606 the server's health will be logged, so that it becomes possible to know
5607 that a server was failing occasional checks before crashing, or exactly when
5608 it failed to respond a valid HTTP status, then when the port started to
5609 reject connections, then when the server stopped responding at all.
5610
5611 Note that status changes not caused by health checks (eg: enable/disable on
5612 the CLI) are intentionally not logged by this option.
Willy Tarreau211ad242009-10-03 21:45:07 +02005613
Willy Tarreaubef1b322014-05-13 21:01:39 +02005614 See also: "option httpchk", "option ldap-check", "option mysql-check",
5615 "option pgsql-check", "option redis-check", "option smtpchk",
5616 "option tcp-check", "log" and section 8 about logging.
Willy Tarreau211ad242009-10-03 21:45:07 +02005617
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005618
5619option log-separate-errors
5620no option log-separate-errors
5621 Change log level for non-completely successful connections
5622 May be used in sections : defaults | frontend | listen | backend
5623 yes | yes | yes | no
5624 Arguments : none
5625
5626 Sometimes looking for errors in logs is not easy. This option makes haproxy
5627 raise the level of logs containing potentially interesting information such
5628 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
5629 level changes from "info" to "err". This makes it possible to log them
5630 separately to a different file with most syslog daemons. Be careful not to
5631 remove them from the original file, otherwise you would lose ordering which
5632 provides very important information.
5633
5634 Using this option, large sites dealing with several thousand connections per
5635 second may log normal traffic to a rotating buffer and only archive smaller
5636 error logs.
5637
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005638 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02005639 logging.
5640
Willy Tarreauc27debf2008-01-06 08:57:02 +01005641
5642option logasap
5643no option logasap
5644 Enable or disable early logging of HTTP requests
5645 May be used in sections : defaults | frontend | listen | backend
5646 yes | yes | yes | no
5647 Arguments : none
5648
5649 By default, HTTP requests are logged upon termination so that the total
5650 transfer time and the number of bytes appear in the logs. When large objects
5651 are being transferred, it may take a while before the request appears in the
5652 logs. Using "option logasap", the request gets logged as soon as the server
5653 sends the complete headers. The only missing information in the logs will be
5654 the total number of bytes which will indicate everything except the amount
5655 of data transferred, and the total time which will not take the transfer
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01005656 time into account. In such a situation, it's a good practice to capture the
Willy Tarreauc27debf2008-01-06 08:57:02 +01005657 "Content-Length" response header so that the logs at least indicate how many
5658 bytes are expected to be transferred.
5659
Willy Tarreaucc6c8912009-02-22 10:53:55 +01005660 Examples :
5661 listen http_proxy 0.0.0.0:80
5662 mode http
5663 option httplog
5664 option logasap
5665 log 192.168.2.200 local3
5666
5667 >>> Feb 6 12:14:14 localhost \
5668 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
5669 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
5670 "GET /image.iso HTTP/1.0"
5671
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005672 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01005673 logging.
5674
5675
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02005676option mysql-check [ user <username> [ post-41 ] ]
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005677 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005678 May be used in sections : defaults | frontend | listen | backend
5679 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005680 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005681 <username> This is the username which will be used when connecting to MySQL
5682 server.
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02005683 post-41 Send post v4.1 client compatible checks
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02005684
5685 If you specify a username, the check consists of sending two MySQL packet,
5686 one Client Authentication packet, and one QUIT packet, to correctly close
5687 MySQL session. We then parse the MySQL Handshake Initialisation packet and/or
5688 Error packet. It is a basic but useful test which does not produce error nor
5689 aborted connect on the server. However, it requires adding an authorization
5690 in the MySQL table, like this :
5691
5692 USE mysql;
5693 INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');
5694 FLUSH PRIVILEGES;
5695
5696 If you don't specify a username (it is deprecated and not recommended), the
5697 check only consists in parsing the Mysql Handshake Initialisation packet or
5698 Error packet, we don't send anything in this mode. It was reported that it
5699 can generate lockout if check is too frequent and/or if there is not enough
5700 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
5701 value as if a connection is established successfully within fewer than MySQL
5702 "max_connect_errors" attempts after a previous connection was interrupted,
5703 the error count for the host is cleared to zero. If HAProxy's server get
5704 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
5705
5706 Remember that this does not check database presence nor database consistency.
5707 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005708
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02005709 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005710
5711 Most often, an incoming MySQL server needs to see the client's IP address for
5712 various purposes, including IP privilege matching and connection logging.
5713 When possible, it is often wise to masquerade the client's IP address when
5714 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02005715 which requires the transparent proxy feature to be compiled in, and the MySQL
5716 server to route the client via the machine hosting haproxy.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01005717
5718 See also: "option httpchk"
5719
5720
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005721option nolinger
5722no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005723 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005724 May be used in sections: defaults | frontend | listen | backend
5725 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005726 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005727
5728 When clients or servers abort connections in a dirty way (eg: they are
5729 physically disconnected), the session timeouts triggers and the session is
5730 closed. But it will remain in FIN_WAIT1 state for some time in the system,
5731 using some resources and possibly limiting the ability to establish newer
5732 connections.
5733
5734 When this happens, it is possible to activate "option nolinger" which forces
5735 the system to immediately remove any socket's pending data on close. Thus,
5736 the session is instantly purged from the system's tables. This usually has
5737 side effects such as increased number of TCP resets due to old retransmits
5738 getting immediately rejected. Some firewalls may sometimes complain about
5739 this too.
5740
5741 For this reason, it is not recommended to use this option when not absolutely
5742 needed. You know that you need it when you have thousands of FIN_WAIT1
5743 sessions on your system (TIME_WAIT ones do not count).
5744
5745 This option may be used both on frontends and backends, depending on the side
5746 where it is required. Use it on the frontend for clients, and on the backend
5747 for servers.
5748
5749 If this option has been enabled in a "defaults" section, it can be disabled
5750 in a specific instance by prepending the "no" keyword before it.
5751
5752
Willy Tarreau55165fe2009-05-10 12:02:55 +02005753option originalto [ except <network> ] [ header <name> ]
5754 Enable insertion of the X-Original-To header to requests sent to servers
5755 May be used in sections : defaults | frontend | listen | backend
5756 yes | yes | yes | yes
5757 Arguments :
5758 <network> is an optional argument used to disable this option for sources
5759 matching <network>
5760 <name> an optional argument to specify a different "X-Original-To"
5761 header name.
5762
5763 Since HAProxy can work in transparent mode, every request from a client can
5764 be redirected to the proxy and HAProxy itself can proxy every request to a
5765 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
5766 be lost. This is annoying when you want access rules based on destination ip
5767 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
5768 added by HAProxy to all requests sent to the server. This header contains a
5769 value representing the original destination IP address. Since this must be
5770 configured to always use the last occurrence of this header only. Note that
5771 only the last occurrence of the header must be used, since it is really
5772 possible that the client has already brought one.
5773
5774 The keyword "header" may be used to supply a different header name to replace
5775 the default "X-Original-To". This can be useful where you might already
5776 have a "X-Original-To" header from a different application, and you need
5777 preserve it. Also if your backend server doesn't use the "X-Original-To"
5778 header and requires different one.
5779
5780 Sometimes, a same HAProxy instance may be shared between a direct client
5781 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
5782 used to decrypt HTTPS traffic). It is possible to disable the addition of the
5783 header for a known source address or network by adding the "except" keyword
5784 followed by the network address. In this case, any source IP matching the
5785 network will not cause an addition of this header. Most common uses are with
5786 private networks or 127.0.0.1.
5787
5788 This option may be specified either in the frontend or in the backend. If at
5789 least one of them uses it, the header will be added. Note that the backend's
5790 setting of the header subargument takes precedence over the frontend's if
5791 both are defined.
5792
Willy Tarreau55165fe2009-05-10 12:02:55 +02005793 Examples :
5794 # Original Destination address
5795 frontend www
5796 mode http
5797 option originalto except 127.0.0.1
5798
5799 # Those servers want the IP Address in X-Client-Dst
5800 backend www
5801 mode http
5802 option originalto header X-Client-Dst
5803
Willy Tarreau87cf5142011-08-19 22:57:24 +02005804 See also : "option httpclose", "option http-server-close",
5805 "option forceclose"
Willy Tarreau55165fe2009-05-10 12:02:55 +02005806
5807
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005808option persist
5809no option persist
5810 Enable or disable forced persistence on down servers
5811 May be used in sections: defaults | frontend | listen | backend
5812 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005813 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005814
5815 When an HTTP request reaches a backend with a cookie which references a dead
5816 server, by default it is redispatched to another server. It is possible to
5817 force the request to be sent to the dead server first using "option persist"
5818 if absolutely needed. A common use case is when servers are under extreme
5819 load and spend their time flapping. In this case, the users would still be
5820 directed to the server they opened the session on, in the hope they would be
5821 correctly served. It is recommended to use "option redispatch" in conjunction
5822 with this option so that in the event it would not be possible to connect to
5823 the server at all (server definitely dead), the client would finally be
5824 redirected to another valid server.
5825
5826 If this option has been enabled in a "defaults" section, it can be disabled
5827 in a specific instance by prepending the "no" keyword before it.
5828
Willy Tarreau4de91492010-01-22 19:10:05 +01005829 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005830
5831
Willy Tarreau0c122822013-12-15 18:49:01 +01005832option pgsql-check [ user <username> ]
5833 Use PostgreSQL health checks for server testing
5834 May be used in sections : defaults | frontend | listen | backend
5835 yes | no | yes | yes
5836 Arguments :
5837 <username> This is the username which will be used when connecting to
5838 PostgreSQL server.
5839
5840 The check sends a PostgreSQL StartupMessage and waits for either
5841 Authentication request or ErrorResponse message. It is a basic but useful
5842 test which does not produce error nor aborted connect on the server.
5843 This check is identical with the "mysql-check".
5844
5845 See also: "option httpchk"
5846
5847
Willy Tarreau9420b122013-12-15 18:58:25 +01005848option prefer-last-server
5849no option prefer-last-server
5850 Allow multiple load balanced requests to remain on the same server
5851 May be used in sections: defaults | frontend | listen | backend
5852 yes | no | yes | yes
5853 Arguments : none
5854
5855 When the load balancing algorithm in use is not deterministic, and a previous
5856 request was sent to a server to which haproxy still holds a connection, it is
5857 sometimes desirable that subsequent requests on a same session go to the same
5858 server as much as possible. Note that this is different from persistence, as
5859 we only indicate a preference which haproxy tries to apply without any form
5860 of warranty. The real use is for keep-alive connections sent to servers. When
5861 this option is used, haproxy will try to reuse the same connection that is
5862 attached to the server instead of rebalancing to another server, causing a
5863 close of the connection. This can make sense for static file servers. It does
Willy Tarreau068621e2013-12-23 15:11:25 +01005864 not make much sense to use this in combination with hashing algorithms. Note,
5865 haproxy already automatically tries to stick to a server which sends a 401 or
5866 to a proxy which sends a 407 (authentication required). This is mandatory for
5867 use with the broken NTLM authentication challenge, and significantly helps in
5868 troubleshooting some faulty applications. Option prefer-last-server might be
5869 desirable in these environments as well, to avoid redistributing the traffic
5870 after every other response.
Willy Tarreau9420b122013-12-15 18:58:25 +01005871
5872 If this option has been enabled in a "defaults" section, it can be disabled
5873 in a specific instance by prepending the "no" keyword before it.
5874
5875 See also: "option http-keep-alive"
5876
5877
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005878option redispatch
Joseph Lynch726ab712015-05-11 23:25:34 -07005879option redispatch <interval>
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005880no option redispatch
5881 Enable or disable session redistribution in case of connection failure
5882 May be used in sections: defaults | frontend | listen | backend
5883 yes | no | yes | yes
Joseph Lynch726ab712015-05-11 23:25:34 -07005884 Arguments :
5885 <interval> The optional integer value that controls how often redispatches
5886 occur when retrying connections. Positive value P indicates a
5887 redispatch is desired on every Pth retry, and negative value
5888 N indicate a redispath is desired on the Nth retry prior to the
5889 last retry. For example, the default of -1 preserves the
5890 historical behaviour of redispatching on the last retry, a
5891 positive value of 1 would indicate a redispatch on every retry,
5892 and a positive value of 3 would indicate a redispatch on every
5893 third retry. You can disable redispatches with a value of 0.
5894
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005895
5896 In HTTP mode, if a server designated by a cookie is down, clients may
5897 definitely stick to it because they cannot flush the cookie, so they will not
5898 be able to access the service anymore.
5899
5900 Specifying "option redispatch" will allow the proxy to break their
5901 persistence and redistribute them to a working server.
5902
Joseph Lynch726ab712015-05-11 23:25:34 -07005903 It also allows to retry connections to another server in case of multiple
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005904 connection failures. Of course, it requires having "retries" set to a nonzero
5905 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01005906
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005907 This form is the preferred form, which replaces both the "redispatch" and
5908 "redisp" keywords.
5909
5910 If this option has been enabled in a "defaults" section, it can be disabled
5911 in a specific instance by prepending the "no" keyword before it.
5912
Willy Tarreau4de91492010-01-22 19:10:05 +01005913 See also : "redispatch", "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005914
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005915
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02005916option redis-check
5917 Use redis health checks for server testing
5918 May be used in sections : defaults | frontend | listen | backend
5919 yes | no | yes | yes
5920 Arguments : none
5921
5922 It is possible to test that the server correctly talks REDIS protocol instead
5923 of just testing that it accepts the TCP connection. When this option is set,
5924 a PING redis command is sent to the server, and the response is analyzed to
5925 find the "+PONG" response message.
5926
5927 Example :
5928 option redis-check
5929
5930 See also : "option httpchk"
5931
5932
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005933option smtpchk
5934option smtpchk <hello> <domain>
5935 Use SMTP health checks for server testing
5936 May be used in sections : defaults | frontend | listen | backend
5937 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01005938 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005939 <hello> is an optional argument. It is the "hello" command to use. It can
5940 be either "HELO" (for SMTP) or "EHLO" (for ESTMP). All other
5941 values will be turned into the default command ("HELO").
5942
5943 <domain> is the domain name to present to the server. It may only be
5944 specified (and is mandatory) if the hello command has been
5945 specified. By default, "localhost" is used.
5946
5947 When "option smtpchk" is set, the health checks will consist in TCP
5948 connections followed by an SMTP command. By default, this command is
5949 "HELO localhost". The server's return code is analyzed and only return codes
5950 starting with a "2" will be considered as valid. All other responses,
5951 including a lack of response will constitute an error and will indicate a
5952 dead server.
5953
5954 This test is meant to be used with SMTP servers or relays. Depending on the
5955 request, it is possible that some servers do not log each connection attempt,
5956 so you may want to experiment to improve the behaviour. Using telnet on port
5957 25 is often easier than adjusting the configuration.
5958
5959 Most often, an incoming SMTP server needs to see the client's IP address for
5960 various purposes, including spam filtering, anti-spoofing and logging. When
5961 possible, it is often wise to masquerade the client's IP address when
5962 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02005963 which requires the transparent proxy feature to be compiled in.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005964
5965 Example :
5966 option smtpchk HELO mydomain.org
5967
5968 See also : "option httpchk", "source"
5969
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005970
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02005971option socket-stats
5972no option socket-stats
5973
5974 Enable or disable collecting & providing separate statistics for each socket.
5975 May be used in sections : defaults | frontend | listen | backend
5976 yes | yes | yes | no
5977
5978 Arguments : none
5979
5980
Willy Tarreauff4f82d2009-02-06 11:28:13 +01005981option splice-auto
5982no option splice-auto
5983 Enable or disable automatic kernel acceleration on sockets in both directions
5984 May be used in sections : defaults | frontend | listen | backend
5985 yes | yes | yes | yes
5986 Arguments : none
5987
5988 When this option is enabled either on a frontend or on a backend, haproxy
5989 will automatically evaluate the opportunity to use kernel tcp splicing to
5990 forward data between the client and the server, in either direction. Haproxy
5991 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005992 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01005993 are not much aggressive in order to limit excessive use of splicing. This
5994 option requires splicing to be enabled at compile time, and may be globally
5995 disabled with the global option "nosplice". Since splice uses pipes, using it
5996 requires that there are enough spare pipes.
5997
5998 Important note: kernel-based TCP splicing is a Linux-specific feature which
5999 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
6000 transfer data between sockets without copying these data to user-space, thus
6001 providing noticeable performance gains and CPU cycles savings. Since many
6002 early implementations are buggy, corrupt data and/or are inefficient, this
6003 feature is not enabled by default, and it should be used with extreme care.
6004 While it is not possible to detect the correctness of an implementation,
6005 2.6.29 is the first version offering a properly working implementation. In
6006 case of doubt, splicing may be globally disabled using the global "nosplice"
6007 keyword.
6008
6009 Example :
6010 option splice-auto
6011
6012 If this option has been enabled in a "defaults" section, it can be disabled
6013 in a specific instance by prepending the "no" keyword before it.
6014
6015 See also : "option splice-request", "option splice-response", and global
6016 options "nosplice" and "maxpipes"
6017
6018
6019option splice-request
6020no option splice-request
6021 Enable or disable automatic kernel acceleration on sockets for requests
6022 May be used in sections : defaults | frontend | listen | backend
6023 yes | yes | yes | yes
6024 Arguments : none
6025
6026 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04006027 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006028 the client to the server. It might still use the recv/send scheme if there
6029 are no spare pipes left. This option requires splicing to be enabled at
6030 compile time, and may be globally disabled with the global option "nosplice".
6031 Since splice uses pipes, using it requires that there are enough spare pipes.
6032
6033 Important note: see "option splice-auto" for usage limitations.
6034
6035 Example :
6036 option splice-request
6037
6038 If this option has been enabled in a "defaults" section, it can be disabled
6039 in a specific instance by prepending the "no" keyword before it.
6040
6041 See also : "option splice-auto", "option splice-response", and global options
6042 "nosplice" and "maxpipes"
6043
6044
6045option splice-response
6046no option splice-response
6047 Enable or disable automatic kernel acceleration on sockets for responses
6048 May be used in sections : defaults | frontend | listen | backend
6049 yes | yes | yes | yes
6050 Arguments : none
6051
6052 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04006053 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01006054 the server to the client. It might still use the recv/send scheme if there
6055 are no spare pipes left. This option requires splicing to be enabled at
6056 compile time, and may be globally disabled with the global option "nosplice".
6057 Since splice uses pipes, using it requires that there are enough spare pipes.
6058
6059 Important note: see "option splice-auto" for usage limitations.
6060
6061 Example :
6062 option splice-response
6063
6064 If this option has been enabled in a "defaults" section, it can be disabled
6065 in a specific instance by prepending the "no" keyword before it.
6066
6067 See also : "option splice-auto", "option splice-request", and global options
6068 "nosplice" and "maxpipes"
6069
6070
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006071option srvtcpka
6072no option srvtcpka
6073 Enable or disable the sending of TCP keepalive packets on the server side
6074 May be used in sections : defaults | frontend | listen | backend
6075 yes | no | yes | yes
6076 Arguments : none
6077
6078 When there is a firewall or any session-aware component between a client and
6079 a server, and when the protocol involves very long sessions with long idle
6080 periods (eg: remote desktops), there is a risk that one of the intermediate
6081 components decides to expire a session which has remained idle for too long.
6082
6083 Enabling socket-level TCP keep-alives makes the system regularly send packets
6084 to the other end of the connection, leaving it active. The delay between
6085 keep-alive probes is controlled by the system only and depends both on the
6086 operating system and its tuning parameters.
6087
6088 It is important to understand that keep-alive packets are neither emitted nor
6089 received at the application level. It is only the network stacks which sees
6090 them. For this reason, even if one side of the proxy already uses keep-alives
6091 to maintain its connection alive, those keep-alive packets will not be
6092 forwarded to the other side of the proxy.
6093
6094 Please note that this has nothing to do with HTTP keep-alive.
6095
6096 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
6097 server side of a connection, which should help when session expirations are
6098 noticed between HAProxy and a server.
6099
6100 If this option has been enabled in a "defaults" section, it can be disabled
6101 in a specific instance by prepending the "no" keyword before it.
6102
6103 See also : "option clitcpka", "option tcpka"
6104
6105
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006106option ssl-hello-chk
6107 Use SSLv3 client hello health checks for server testing
6108 May be used in sections : defaults | frontend | listen | backend
6109 yes | no | yes | yes
6110 Arguments : none
6111
6112 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
6113 possible to test that the server correctly talks SSL instead of just testing
6114 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
6115 SSLv3 client hello messages are sent once the connection is established to
6116 the server, and the response is analyzed to find an SSL server hello message.
6117 The server is considered valid only when the response contains this server
6118 hello message.
6119
6120 All servers tested till there correctly reply to SSLv3 client hello messages,
6121 and most servers tested do not even log the requests containing only hello
6122 messages, which is appreciable.
6123
Willy Tarreau763a95b2012-10-04 23:15:39 +02006124 Note that this check works even when SSL support was not built into haproxy
6125 because it forges the SSL message. When SSL support is available, it is best
6126 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006127
Willy Tarreau763a95b2012-10-04 23:15:39 +02006128 See also: "option httpchk", "check-ssl"
6129
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006130
Willy Tarreaued179852013-12-16 01:07:00 +01006131option tcp-check
6132 Perform health checks using tcp-check send/expect sequences
6133 May be used in sections: defaults | frontend | listen | backend
6134 yes | no | yes | yes
6135
6136 This health check method is intended to be combined with "tcp-check" command
6137 lists in order to support send/expect types of health check sequences.
6138
6139 TCP checks currently support 4 modes of operations :
6140 - no "tcp-check" directive : the health check only consists in a connection
6141 attempt, which remains the default mode.
6142
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006143 - "tcp-check send" or "tcp-check send-binary" only is mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01006144 used to send a string along with a connection opening. With some
6145 protocols, it helps sending a "QUIT" message for example that prevents
6146 the server from logging a connection error for each health check. The
6147 check result will still be based on the ability to open the connection
6148 only.
6149
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006150 - "tcp-check expect" only is mentioned : this is used to test a banner.
Willy Tarreaued179852013-12-16 01:07:00 +01006151 The connection is opened and haproxy waits for the server to present some
6152 contents which must validate some rules. The check result will be based
6153 on the matching between the contents and the rules. This is suited for
6154 POP, IMAP, SMTP, FTP, SSH, TELNET.
6155
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006156 - both "tcp-check send" and "tcp-check expect" are mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01006157 used to test a hello-type protocol. Haproxy sends a message, the server
6158 responds and its response is analysed. the check result will be based on
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006159 the matching between the response contents and the rules. This is often
Willy Tarreaued179852013-12-16 01:07:00 +01006160 suited for protocols which require a binding or a request/response model.
6161 LDAP, MySQL, Redis and SSL are example of such protocols, though they
6162 already all have their dedicated checks with a deeper understanding of
6163 the respective protocols.
6164 In this mode, many questions may be sent and many answers may be
6165 analysed.
6166
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006167 A fifth mode can be used to insert comments in different steps of the
6168 script.
6169
6170 For each tcp-check rule you create, you can add a "comment" directive,
6171 followed by a string. This string will be reported in the log and stderr
6172 in debug mode. It is useful to make user-friendly error reporting.
6173 The "comment" is of course optional.
6174
6175
Willy Tarreaued179852013-12-16 01:07:00 +01006176 Examples :
6177 # perform a POP check (analyse only server's banner)
6178 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006179 tcp-check expect string +OK\ POP3\ ready comment POP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01006180
6181 # perform an IMAP check (analyse only server's banner)
6182 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006183 tcp-check expect string *\ OK\ IMAP4\ ready comment IMAP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01006184
6185 # look for the redis master server after ensuring it speaks well
6186 # redis protocol, then it exits properly.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006187 # (send a command then analyse the response 3 times)
Willy Tarreaued179852013-12-16 01:07:00 +01006188 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006189 tcp-check comment PING\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01006190 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02006191 tcp-check expect string +PONG
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006192 tcp-check comment role\ check
Willy Tarreaued179852013-12-16 01:07:00 +01006193 tcp-check send info\ replication\r\n
6194 tcp-check expect string role:master
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006195 tcp-check comment QUIT\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01006196 tcp-check send QUIT\r\n
6197 tcp-check expect string +OK
6198
6199 forge a HTTP request, then analyse the response
6200 (send many headers before analyzing)
6201 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006202 tcp-check comment forge\ and\ send\ HTTP\ request
Willy Tarreaued179852013-12-16 01:07:00 +01006203 tcp-check send HEAD\ /\ HTTP/1.1\r\n
6204 tcp-check send Host:\ www.mydomain.com\r\n
6205 tcp-check send User-Agent:\ HAProxy\ tcpcheck\r\n
6206 tcp-check send \r\n
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02006207 tcp-check expect rstring HTTP/1\..\ (2..|3..) comment check\ HTTP\ response
Willy Tarreaued179852013-12-16 01:07:00 +01006208
6209
6210 See also : "tcp-check expect", "tcp-check send"
6211
6212
Willy Tarreau9ea05a72009-06-14 12:07:01 +02006213option tcp-smart-accept
6214no option tcp-smart-accept
6215 Enable or disable the saving of one ACK packet during the accept sequence
6216 May be used in sections : defaults | frontend | listen | backend
6217 yes | yes | yes | no
6218 Arguments : none
6219
6220 When an HTTP connection request comes in, the system acknowledges it on
6221 behalf of HAProxy, then the client immediately sends its request, and the
6222 system acknowledges it too while it is notifying HAProxy about the new
6223 connection. HAProxy then reads the request and responds. This means that we
6224 have one TCP ACK sent by the system for nothing, because the request could
6225 very well be acknowledged by HAProxy when it sends its response.
6226
6227 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
6228 sending this useless ACK on platforms which support it (currently at least
6229 Linux). It must not cause any problem, because the system will send it anyway
6230 after 40 ms if the response takes more time than expected to come.
6231
6232 During complex network debugging sessions, it may be desirable to disable
6233 this optimization because delayed ACKs can make troubleshooting more complex
6234 when trying to identify where packets are delayed. It is then possible to
6235 fall back to normal behaviour by specifying "no option tcp-smart-accept".
6236
6237 It is also possible to force it for non-HTTP proxies by simply specifying
6238 "option tcp-smart-accept". For instance, it can make sense with some services
6239 such as SMTP where the server speaks first.
6240
6241 It is recommended to avoid forcing this option in a defaults section. In case
6242 of doubt, consider setting it back to automatic values by prepending the
6243 "default" keyword before it, or disabling it using the "no" keyword.
6244
Willy Tarreaud88edf22009-06-14 15:48:17 +02006245 See also : "option tcp-smart-connect"
6246
6247
6248option tcp-smart-connect
6249no option tcp-smart-connect
6250 Enable or disable the saving of one ACK packet during the connect sequence
6251 May be used in sections : defaults | frontend | listen | backend
6252 yes | no | yes | yes
6253 Arguments : none
6254
6255 On certain systems (at least Linux), HAProxy can ask the kernel not to
6256 immediately send an empty ACK upon a connection request, but to directly
6257 send the buffer request instead. This saves one packet on the network and
6258 thus boosts performance. It can also be useful for some servers, because they
6259 immediately get the request along with the incoming connection.
6260
6261 This feature is enabled when "option tcp-smart-connect" is set in a backend.
6262 It is not enabled by default because it makes network troubleshooting more
6263 complex.
6264
6265 It only makes sense to enable it with protocols where the client speaks first
6266 such as HTTP. In other situations, if there is no data to send in place of
6267 the ACK, a normal ACK is sent.
6268
6269 If this option has been enabled in a "defaults" section, it can be disabled
6270 in a specific instance by prepending the "no" keyword before it.
6271
6272 See also : "option tcp-smart-accept"
6273
Willy Tarreau9ea05a72009-06-14 12:07:01 +02006274
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006275option tcpka
6276 Enable or disable the sending of TCP keepalive packets on both sides
6277 May be used in sections : defaults | frontend | listen | backend
6278 yes | yes | yes | yes
6279 Arguments : none
6280
6281 When there is a firewall or any session-aware component between a client and
6282 a server, and when the protocol involves very long sessions with long idle
6283 periods (eg: remote desktops), there is a risk that one of the intermediate
6284 components decides to expire a session which has remained idle for too long.
6285
6286 Enabling socket-level TCP keep-alives makes the system regularly send packets
6287 to the other end of the connection, leaving it active. The delay between
6288 keep-alive probes is controlled by the system only and depends both on the
6289 operating system and its tuning parameters.
6290
6291 It is important to understand that keep-alive packets are neither emitted nor
6292 received at the application level. It is only the network stacks which sees
6293 them. For this reason, even if one side of the proxy already uses keep-alives
6294 to maintain its connection alive, those keep-alive packets will not be
6295 forwarded to the other side of the proxy.
6296
6297 Please note that this has nothing to do with HTTP keep-alive.
6298
6299 Using option "tcpka" enables the emission of TCP keep-alive probes on both
6300 the client and server sides of a connection. Note that this is meaningful
6301 only in "defaults" or "listen" sections. If this option is used in a
6302 frontend, only the client side will get keep-alives, and if this option is
6303 used in a backend, only the server side will get keep-alives. For this
6304 reason, it is strongly recommended to explicitly use "option clitcpka" and
6305 "option srvtcpka" when the configuration is split between frontends and
6306 backends.
6307
6308 See also : "option clitcpka", "option srvtcpka"
6309
Willy Tarreau844e3c52008-01-11 16:28:18 +01006310
6311option tcplog
6312 Enable advanced logging of TCP connections with session state and timers
6313 May be used in sections : defaults | frontend | listen | backend
6314 yes | yes | yes | yes
6315 Arguments : none
6316
6317 By default, the log output format is very poor, as it only contains the
6318 source and destination addresses, and the instance name. By specifying
6319 "option tcplog", each log line turns into a much richer format including, but
6320 not limited to, the connection timers, the session status, the connections
6321 numbers, the frontend, backend and server name, and of course the source
6322 address and ports. This option is useful for pure TCP proxies in order to
6323 find which of the client or server disconnects or times out. For normal HTTP
6324 proxies, it's better to use "option httplog" which is even more complete.
6325
6326 This option may be set either in the frontend or the backend.
6327
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006328 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01006329
6330
Willy Tarreau844e3c52008-01-11 16:28:18 +01006331option transparent
6332no option transparent
6333 Enable client-side transparent proxying
6334 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01006335 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01006336 Arguments : none
6337
6338 This option was introduced in order to provide layer 7 persistence to layer 3
6339 load balancers. The idea is to use the OS's ability to redirect an incoming
6340 connection for a remote address to a local process (here HAProxy), and let
6341 this process know what address was initially requested. When this option is
6342 used, sessions without cookies will be forwarded to the original destination
6343 IP address of the incoming request (which should match that of another
6344 equipment), while requests with cookies will still be forwarded to the
6345 appropriate server.
6346
6347 Note that contrary to a common belief, this option does NOT make HAProxy
6348 present the client's IP to the server when establishing the connection.
6349
Willy Tarreaua1146052011-03-01 09:51:54 +01006350 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006351 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01006352
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006353
Simon Horman98637e52014-06-20 12:30:16 +09006354external-check command <command>
6355 Executable to run when performing an external-check
6356 May be used in sections : defaults | frontend | listen | backend
6357 yes | no | yes | yes
6358
6359 Arguments :
6360 <command> is the external command to run
6361
Simon Horman98637e52014-06-20 12:30:16 +09006362 The arguments passed to the to the command are:
6363
Cyril Bonté777be862014-12-02 21:21:35 +01006364 <proxy_address> <proxy_port> <server_address> <server_port>
Simon Horman98637e52014-06-20 12:30:16 +09006365
Cyril Bonté777be862014-12-02 21:21:35 +01006366 The <proxy_address> and <proxy_port> are derived from the first listener
6367 that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
6368 listener the proxy_address will be the path of the socket and the
6369 <proxy_port> will be the string "NOT_USED". In a backend section, it's not
6370 possible to determine a listener, and both <proxy_address> and <proxy_port>
6371 will have the string value "NOT_USED".
Simon Horman98637e52014-06-20 12:30:16 +09006372
Cyril Bonté72cda2a2014-12-27 22:28:39 +01006373 Some values are also provided through environment variables.
6374
6375 Environment variables :
6376 HAPROXY_PROXY_ADDR The first bind address if available (or empty if not
6377 applicable, for example in a "backend" section).
6378
6379 HAPROXY_PROXY_ID The backend id.
6380
6381 HAPROXY_PROXY_NAME The backend name.
6382
6383 HAPROXY_PROXY_PORT The first bind port if available (or empty if not
6384 applicable, for example in a "backend" section or
6385 for a UNIX socket).
6386
6387 HAPROXY_SERVER_ADDR The server address.
6388
6389 HAPROXY_SERVER_CURCONN The current number of connections on the server.
6390
6391 HAPROXY_SERVER_ID The server id.
6392
6393 HAPROXY_SERVER_MAXCONN The server max connections.
6394
6395 HAPROXY_SERVER_NAME The server name.
6396
6397 HAPROXY_SERVER_PORT The server port if available (or empty for a UNIX
6398 socket).
6399
6400 PATH The PATH environment variable used when executing
6401 the command may be set using "external-check path".
6402
Simon Horman98637e52014-06-20 12:30:16 +09006403 If the command executed and exits with a zero status then the check is
6404 considered to have passed, otherwise the check is considered to have
6405 failed.
6406
6407 Example :
6408 external-check command /bin/true
6409
6410 See also : "external-check", "option external-check", "external-check path"
6411
6412
6413external-check path <path>
6414 The value of the PATH environment variable used when running an external-check
6415 May be used in sections : defaults | frontend | listen | backend
6416 yes | no | yes | yes
6417
6418 Arguments :
6419 <path> is the path used when executing external command to run
6420
6421 The default path is "".
6422
6423 Example :
6424 external-check path "/usr/bin:/bin"
6425
6426 See also : "external-check", "option external-check",
6427 "external-check command"
6428
6429
Emeric Brun647caf12009-06-30 17:57:00 +02006430persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02006431persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02006432 Enable RDP cookie-based persistence
6433 May be used in sections : defaults | frontend | listen | backend
6434 yes | no | yes | yes
6435 Arguments :
6436 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02006437 default cookie name "msts" will be used. There currently is no
6438 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02006439
6440 This statement enables persistence based on an RDP cookie. The RDP cookie
6441 contains all information required to find the server in the list of known
6442 servers. So when this option is set in the backend, the request is analysed
6443 and if an RDP cookie is found, it is decoded. If it matches a known server
6444 which is still UP (or if "option persist" is set), then the connection is
6445 forwarded to this server.
6446
6447 Note that this only makes sense in a TCP backend, but for this to work, the
6448 frontend must have waited long enough to ensure that an RDP cookie is present
6449 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006450 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02006451 a single "listen" section.
6452
Willy Tarreau61e28f22010-05-16 22:31:05 +02006453 Also, it is important to understand that the terminal server will emit this
6454 RDP cookie only if it is configured for "token redirection mode", which means
6455 that the "IP address redirection" option is disabled.
6456
Emeric Brun647caf12009-06-30 17:57:00 +02006457 Example :
6458 listen tse-farm
6459 bind :3389
6460 # wait up to 5s for an RDP cookie in the request
6461 tcp-request inspect-delay 5s
6462 tcp-request content accept if RDP_COOKIE
6463 # apply RDP cookie persistence
6464 persist rdp-cookie
6465 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02006466 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02006467 balance rdp-cookie
6468 server srv1 1.1.1.1:3389
6469 server srv2 1.1.1.2:3389
6470
Simon Hormanab814e02011-06-24 14:50:20 +09006471 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
6472 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02006473
6474
Willy Tarreau3a7d2072009-03-05 23:48:25 +01006475rate-limit sessions <rate>
6476 Set a limit on the number of new sessions accepted per second on a frontend
6477 May be used in sections : defaults | frontend | listen | backend
6478 yes | yes | yes | no
6479 Arguments :
6480 <rate> The <rate> parameter is an integer designating the maximum number
6481 of new sessions per second to accept on the frontend.
6482
6483 When the frontend reaches the specified number of new sessions per second, it
6484 stops accepting new connections until the rate drops below the limit again.
6485 During this time, the pending sessions will be kept in the socket's backlog
6486 (in system buffers) and haproxy will not even be aware that sessions are
6487 pending. When applying very low limit on a highly loaded service, it may make
6488 sense to increase the socket's backlog using the "backlog" keyword.
6489
6490 This feature is particularly efficient at blocking connection-based attacks
6491 or service abuse on fragile servers. Since the session rate is measured every
6492 millisecond, it is extremely accurate. Also, the limit applies immediately,
6493 no delay is needed at all to detect the threshold.
6494
6495 Example : limit the connection rate on SMTP to 10 per second max
6496 listen smtp
6497 mode tcp
6498 bind :25
6499 rate-limit sessions 10
6500 server 127.0.0.1:1025
6501
Willy Tarreaua17c2d92011-07-25 08:16:20 +02006502 Note : when the maximum rate is reached, the frontend's status is not changed
6503 but its sockets appear as "WAITING" in the statistics if the
6504 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +01006505
6506 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
6507
6508
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006509redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
6510redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
6511redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006512 Return an HTTP redirection if/unless a condition is matched
6513 May be used in sections : defaults | frontend | listen | backend
6514 no | yes | yes | yes
6515
6516 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +01006517 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006518
Willy Tarreau0140f252008-11-19 21:07:09 +01006519 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006520 <loc> With "redirect location", the exact value in <loc> is placed into
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006521 the HTTP "Location" header. When used in an "http-request" rule,
6522 <loc> value follows the log-format rules and can include some
6523 dynamic values (see Custom Log Format in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006524
6525 <pfx> With "redirect prefix", the "Location" header is built from the
6526 concatenation of <pfx> and the complete URI path, including the
6527 query string, unless the "drop-query" option is specified (see
6528 below). As a special case, if <pfx> equals exactly "/", then
6529 nothing is inserted before the original URI. It allows one to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006530 redirect to the same URL (for instance, to insert a cookie). When
6531 used in an "http-request" rule, <pfx> value follows the log-format
6532 rules and can include some dynamic values (see Custom Log Format
6533 in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006534
6535 <sch> With "redirect scheme", then the "Location" header is built by
6536 concatenating <sch> with "://" then the first occurrence of the
6537 "Host" header, and then the URI path, including the query string
6538 unless the "drop-query" option is specified (see below). If no
6539 path is found or if the path is "*", then "/" is used instead. If
6540 no "Host" header is found, then an empty host component will be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006541 returned, which most recent browsers interpret as redirecting to
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006542 the same host. This directive is mostly used to redirect HTTP to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006543 HTTPS. When used in an "http-request" rule, <sch> value follows
6544 the log-format rules and can include some dynamic values (see
6545 Custom Log Format in section 8.2.4).
Willy Tarreau0140f252008-11-19 21:07:09 +01006546
6547 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +01006548 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
6549 with 302 used by default if no code is specified. 301 means
6550 "Moved permanently", and a browser may cache the Location. 302
Baptiste Assmannea849c02015-08-03 11:42:50 +02006551 means "Moved temporarily" and means that the browser should not
Willy Tarreaub67fdc42013-03-29 19:28:11 +01006552 cache the redirection. 303 is equivalent to 302 except that the
6553 browser will fetch the location with a GET method. 307 is just
6554 like 302 but makes it clear that the same method must be reused.
6555 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +01006556
6557 <option> There are several options which can be specified to adjust the
6558 expected behaviour of a redirection :
6559
6560 - "drop-query"
6561 When this keyword is used in a prefix-based redirection, then the
6562 location will be set without any possible query-string, which is useful
6563 for directing users to a non-secure page for instance. It has no effect
6564 with a location-type redirect.
6565
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01006566 - "append-slash"
6567 This keyword may be used in conjunction with "drop-query" to redirect
6568 users who use a URL not ending with a '/' to the same one with the '/'.
6569 It can be useful to ensure that search engines will only see one URL.
6570 For this, a return code 301 is preferred.
6571
Willy Tarreau0140f252008-11-19 21:07:09 +01006572 - "set-cookie NAME[=value]"
6573 A "Set-Cookie" header will be added with NAME (and optionally "=value")
6574 to the response. This is sometimes used to indicate that a user has
6575 been seen, for instance to protect against some types of DoS. No other
6576 cookie option is added, so the cookie will be a session cookie. Note
6577 that for a browser, a sole cookie name without an equal sign is
6578 different from a cookie with an equal sign.
6579
6580 - "clear-cookie NAME[=]"
6581 A "Set-Cookie" header will be added with NAME (and optionally "="), but
6582 with the "Max-Age" attribute set to zero. This will tell the browser to
6583 delete this cookie. It is useful for instance on logout pages. It is
6584 important to note that clearing the cookie "NAME" will not remove a
6585 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
6586 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006587
6588 Example: move the login URL only to HTTPS.
6589 acl clear dst_port 80
6590 acl secure dst_port 8080
6591 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +01006592 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +01006593 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +01006594 acl cookie_set hdr_sub(cookie) SEEN=1
6595
6596 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +01006597 redirect prefix https://mysite.com if login_page !secure
6598 redirect prefix http://mysite.com drop-query if login_page !uid_given
6599 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +01006600 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006601
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01006602 Example: send redirects for request for articles without a '/'.
6603 acl missing_slash path_reg ^/article/[^/]*$
6604 redirect code 301 prefix / drop-query append-slash if missing_slash
6605
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006606 Example: redirect all HTTP traffic to HTTPS when SSL is handled by haproxy.
David BERARDe7153042012-11-03 00:11:31 +01006607 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +02006608
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01006609 Example: append 'www.' prefix in front of all hosts not having it
6610 http-request redirect code 301 location www.%[hdr(host)]%[req.uri] \
6611 unless { hdr_beg(host) -i www }
6612
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006613 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02006614
6615
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006616redisp (deprecated)
6617redispatch (deprecated)
6618 Enable or disable session redistribution in case of connection failure
6619 May be used in sections: defaults | frontend | listen | backend
6620 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006621 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006622
6623 In HTTP mode, if a server designated by a cookie is down, clients may
6624 definitely stick to it because they cannot flush the cookie, so they will not
6625 be able to access the service anymore.
6626
6627 Specifying "redispatch" will allow the proxy to break their persistence and
6628 redistribute them to a working server.
6629
6630 It also allows to retry last connection to another server in case of multiple
6631 connection failures. Of course, it requires having "retries" set to a nonzero
6632 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01006633
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006634 This form is deprecated, do not use it in any new configuration, use the new
6635 "option redispatch" instead.
6636
6637 See also : "option redispatch"
6638
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006639
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006640reqadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01006641 Add a header at the end of the HTTP request
6642 May be used in sections : defaults | frontend | listen | backend
6643 no | yes | yes | yes
6644 Arguments :
6645 <string> is the complete line to be added. Any space or known delimiter
6646 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006647 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006648
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006649 <cond> is an optional matching condition built from ACLs. It makes it
6650 possible to ignore this rule when other conditions are not met.
6651
Willy Tarreau303c0352008-01-17 19:01:39 +01006652 A new line consisting in <string> followed by a line feed will be added after
6653 the last header of an HTTP request.
6654
6655 Header transformations only apply to traffic which passes through HAProxy,
6656 and not to traffic generated by HAProxy, such as health-checks or error
6657 responses.
6658
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01006659 Example : add "X-Proto: SSL" to requests coming via port 81
6660 acl is-ssl dst_port 81
6661 reqadd X-Proto:\ SSL if is-ssl
6662
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006663 See also: "rspadd", "http-request", section 6 about HTTP header manipulation,
6664 and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006665
6666
Willy Tarreau5321c422010-01-28 20:35:13 +01006667reqallow <search> [{if | unless} <cond>]
6668reqiallow <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006669 Definitely allow an HTTP request if a line matches a regular expression
6670 May be used in sections : defaults | frontend | listen | backend
6671 no | yes | yes | yes
6672 Arguments :
6673 <search> is the regular expression applied to HTTP headers and to the
6674 request line. This is an extended regular expression. Parenthesis
6675 grouping is supported and no preliminary backslash is required.
6676 Any space or known delimiter must be escaped using a backslash
6677 ('\'). The pattern applies to a full line at a time. The
6678 "reqallow" keyword strictly matches case while "reqiallow"
6679 ignores case.
6680
Willy Tarreau5321c422010-01-28 20:35:13 +01006681 <cond> is an optional matching condition built from ACLs. It makes it
6682 possible to ignore this rule when other conditions are not met.
6683
Willy Tarreau303c0352008-01-17 19:01:39 +01006684 A request containing any line which matches extended regular expression
6685 <search> will mark the request as allowed, even if any later test would
6686 result in a deny. The test applies both to the request line and to request
6687 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01006688 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01006689
6690 It is easier, faster and more powerful to use ACLs to write access policies.
6691 Reqdeny, reqallow and reqpass should be avoided in new designs.
6692
6693 Example :
6694 # allow www.* but refuse *.local
6695 reqiallow ^Host:\ www\.
6696 reqideny ^Host:\ .*\.local
6697
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006698 See also: "reqdeny", "block", "http-request", section 6 about HTTP header
6699 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006700
6701
Willy Tarreau5321c422010-01-28 20:35:13 +01006702reqdel <search> [{if | unless} <cond>]
6703reqidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006704 Delete all headers matching a regular expression in an HTTP request
6705 May be used in sections : defaults | frontend | listen | backend
6706 no | yes | yes | yes
6707 Arguments :
6708 <search> is the regular expression applied to HTTP headers and to the
6709 request line. This is an extended regular expression. Parenthesis
6710 grouping is supported and no preliminary backslash is required.
6711 Any space or known delimiter must be escaped using a backslash
6712 ('\'). The pattern applies to a full line at a time. The "reqdel"
6713 keyword strictly matches case while "reqidel" ignores case.
6714
Willy Tarreau5321c422010-01-28 20:35:13 +01006715 <cond> is an optional matching condition built from ACLs. It makes it
6716 possible to ignore this rule when other conditions are not met.
6717
Willy Tarreau303c0352008-01-17 19:01:39 +01006718 Any header line matching extended regular expression <search> in the request
6719 will be completely deleted. Most common use of this is to remove unwanted
6720 and/or dangerous headers or cookies from a request before passing it to the
6721 next servers.
6722
6723 Header transformations only apply to traffic which passes through HAProxy,
6724 and not to traffic generated by HAProxy, such as health-checks or error
6725 responses. Keep in mind that header names are not case-sensitive.
6726
6727 Example :
6728 # remove X-Forwarded-For header and SERVER cookie
6729 reqidel ^X-Forwarded-For:.*
6730 reqidel ^Cookie:.*SERVER=
6731
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006732 See also: "reqadd", "reqrep", "rspdel", "http-request", section 6 about
6733 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006734
6735
Willy Tarreau5321c422010-01-28 20:35:13 +01006736reqdeny <search> [{if | unless} <cond>]
6737reqideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006738 Deny an HTTP request if a line matches a regular expression
6739 May be used in sections : defaults | frontend | listen | backend
6740 no | yes | yes | yes
6741 Arguments :
6742 <search> is the regular expression applied to HTTP headers and to the
6743 request line. This is an extended regular expression. Parenthesis
6744 grouping is supported and no preliminary backslash is required.
6745 Any space or known delimiter must be escaped using a backslash
6746 ('\'). The pattern applies to a full line at a time. The
6747 "reqdeny" keyword strictly matches case while "reqideny" ignores
6748 case.
6749
Willy Tarreau5321c422010-01-28 20:35:13 +01006750 <cond> is an optional matching condition built from ACLs. It makes it
6751 possible to ignore this rule when other conditions are not met.
6752
Willy Tarreau303c0352008-01-17 19:01:39 +01006753 A request containing any line which matches extended regular expression
6754 <search> will mark the request as denied, even if any later test would
6755 result in an allow. The test applies both to the request line and to request
6756 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01006757 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01006758
Willy Tarreauced27012008-01-17 20:35:34 +01006759 A denied request will generate an "HTTP 403 forbidden" response once the
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006760 complete request has been parsed. This is consistent with what is practiced
Willy Tarreaud72758d2010-01-12 10:42:19 +01006761 using ACLs.
Willy Tarreauced27012008-01-17 20:35:34 +01006762
Willy Tarreau303c0352008-01-17 19:01:39 +01006763 It is easier, faster and more powerful to use ACLs to write access policies.
6764 Reqdeny, reqallow and reqpass should be avoided in new designs.
6765
6766 Example :
6767 # refuse *.local, then allow www.*
6768 reqideny ^Host:\ .*\.local
6769 reqiallow ^Host:\ www\.
6770
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006771 See also: "reqallow", "rspdeny", "block", "http-request", section 6 about
6772 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006773
6774
Willy Tarreau5321c422010-01-28 20:35:13 +01006775reqpass <search> [{if | unless} <cond>]
6776reqipass <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006777 Ignore any HTTP request line matching a regular expression in next rules
6778 May be used in sections : defaults | frontend | listen | backend
6779 no | yes | yes | yes
6780 Arguments :
6781 <search> is the regular expression applied to HTTP headers and to the
6782 request line. This is an extended regular expression. Parenthesis
6783 grouping is supported and no preliminary backslash is required.
6784 Any space or known delimiter must be escaped using a backslash
6785 ('\'). The pattern applies to a full line at a time. The
6786 "reqpass" keyword strictly matches case while "reqipass" ignores
6787 case.
6788
Willy Tarreau5321c422010-01-28 20:35:13 +01006789 <cond> is an optional matching condition built from ACLs. It makes it
6790 possible to ignore this rule when other conditions are not met.
6791
Willy Tarreau303c0352008-01-17 19:01:39 +01006792 A request containing any line which matches extended regular expression
6793 <search> will skip next rules, without assigning any deny or allow verdict.
6794 The test applies both to the request line and to request headers. Keep in
6795 mind that URLs in request line are case-sensitive while header names are not.
6796
6797 It is easier, faster and more powerful to use ACLs to write access policies.
6798 Reqdeny, reqallow and reqpass should be avoided in new designs.
6799
6800 Example :
6801 # refuse *.local, then allow www.*, but ignore "www.private.local"
6802 reqipass ^Host:\ www.private\.local
6803 reqideny ^Host:\ .*\.local
6804 reqiallow ^Host:\ www\.
6805
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006806 See also: "reqallow", "reqdeny", "block", "http-request", section 6 about
6807 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006808
6809
Willy Tarreau5321c422010-01-28 20:35:13 +01006810reqrep <search> <string> [{if | unless} <cond>]
6811reqirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006812 Replace a regular expression with a string in an HTTP request line
6813 May be used in sections : defaults | frontend | listen | backend
6814 no | yes | yes | yes
6815 Arguments :
6816 <search> is the regular expression applied to HTTP headers and to the
6817 request line. This is an extended regular expression. Parenthesis
6818 grouping is supported and no preliminary backslash is required.
6819 Any space or known delimiter must be escaped using a backslash
6820 ('\'). The pattern applies to a full line at a time. The "reqrep"
6821 keyword strictly matches case while "reqirep" ignores case.
6822
6823 <string> is the complete line to be added. Any space or known delimiter
6824 must be escaped using a backslash ('\'). References to matched
6825 pattern groups are possible using the common \N form, with N
6826 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006827 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006828
Willy Tarreau5321c422010-01-28 20:35:13 +01006829 <cond> is an optional matching condition built from ACLs. It makes it
6830 possible to ignore this rule when other conditions are not met.
6831
Willy Tarreau303c0352008-01-17 19:01:39 +01006832 Any line matching extended regular expression <search> in the request (both
6833 the request line and header lines) will be completely replaced with <string>.
6834 Most common use of this is to rewrite URLs or domain names in "Host" headers.
6835
6836 Header transformations only apply to traffic which passes through HAProxy,
6837 and not to traffic generated by HAProxy, such as health-checks or error
6838 responses. Note that for increased readability, it is suggested to add enough
6839 spaces between the request and the response. Keep in mind that URLs in
6840 request line are case-sensitive while header names are not.
6841
6842 Example :
6843 # replace "/static/" with "/" at the beginning of any request path.
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04006844 reqrep ^([^\ :]*)\ /static/(.*) \1\ /\2
Willy Tarreau303c0352008-01-17 19:01:39 +01006845 # replace "www.mydomain.com" with "www" in the host name.
6846 reqirep ^Host:\ www.mydomain.com Host:\ www
6847
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006848 See also: "reqadd", "reqdel", "rsprep", "tune.bufsize", "http-request",
6849 section 6 about HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006850
6851
Willy Tarreau5321c422010-01-28 20:35:13 +01006852reqtarpit <search> [{if | unless} <cond>]
6853reqitarpit <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006854 Tarpit an HTTP request containing a line matching a regular expression
6855 May be used in sections : defaults | frontend | listen | backend
6856 no | yes | yes | yes
6857 Arguments :
6858 <search> is the regular expression applied to HTTP headers and to the
6859 request line. This is an extended regular expression. Parenthesis
6860 grouping is supported and no preliminary backslash is required.
6861 Any space or known delimiter must be escaped using a backslash
6862 ('\'). The pattern applies to a full line at a time. The
6863 "reqtarpit" keyword strictly matches case while "reqitarpit"
6864 ignores case.
6865
Willy Tarreau5321c422010-01-28 20:35:13 +01006866 <cond> is an optional matching condition built from ACLs. It makes it
6867 possible to ignore this rule when other conditions are not met.
6868
Willy Tarreau303c0352008-01-17 19:01:39 +01006869 A request containing any line which matches extended regular expression
6870 <search> will be tarpitted, which means that it will connect to nowhere, will
Willy Tarreauced27012008-01-17 20:35:34 +01006871 be kept open for a pre-defined time, then will return an HTTP error 500 so
6872 that the attacker does not suspect it has been tarpitted. The status 500 will
6873 be reported in the logs, but the completion flags will indicate "PT". The
Willy Tarreau303c0352008-01-17 19:01:39 +01006874 delay is defined by "timeout tarpit", or "timeout connect" if the former is
6875 not set.
6876
6877 The goal of the tarpit is to slow down robots attacking servers with
6878 identifiable requests. Many robots limit their outgoing number of connections
6879 and stay connected waiting for a reply which can take several minutes to
6880 come. Depending on the environment and attack, it may be particularly
6881 efficient at reducing the load on the network and firewalls.
6882
Willy Tarreau5321c422010-01-28 20:35:13 +01006883 Examples :
Willy Tarreau303c0352008-01-17 19:01:39 +01006884 # ignore user-agents reporting any flavour of "Mozilla" or "MSIE", but
6885 # block all others.
6886 reqipass ^User-Agent:\.*(Mozilla|MSIE)
6887 reqitarpit ^User-Agent:
6888
Willy Tarreau5321c422010-01-28 20:35:13 +01006889 # block bad guys
6890 acl badguys src 10.1.0.3 172.16.13.20/28
6891 reqitarpit . if badguys
6892
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006893 See also: "reqallow", "reqdeny", "reqpass", "http-request", section 6
6894 about HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006895
6896
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02006897retries <value>
6898 Set the number of retries to perform on a server after a connection failure
6899 May be used in sections: defaults | frontend | listen | backend
6900 yes | no | yes | yes
6901 Arguments :
6902 <value> is the number of times a connection attempt should be retried on
6903 a server when a connection either is refused or times out. The
6904 default value is 3.
6905
6906 It is important to understand that this value applies to the number of
6907 connection attempts, not full requests. When a connection has effectively
6908 been established to a server, there will be no more retry.
6909
6910 In order to avoid immediate reconnections to a server which is restarting,
Joseph Lynch726ab712015-05-11 23:25:34 -07006911 a turn-around timer of min("timeout connect", one second) is applied before
6912 a retry occurs.
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02006913
6914 When "option redispatch" is set, the last retry may be performed on another
6915 server even if a cookie references a different server.
6916
6917 See also : "option redispatch"
6918
6919
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006920rspadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01006921 Add a header at the end of the HTTP response
6922 May be used in sections : defaults | frontend | listen | backend
6923 no | yes | yes | yes
6924 Arguments :
6925 <string> is the complete line to be added. Any space or known delimiter
6926 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006927 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006928
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006929 <cond> is an optional matching condition built from ACLs. It makes it
6930 possible to ignore this rule when other conditions are not met.
6931
Willy Tarreau303c0352008-01-17 19:01:39 +01006932 A new line consisting in <string> followed by a line feed will be added after
6933 the last header of an HTTP response.
6934
6935 Header transformations only apply to traffic which passes through HAProxy,
6936 and not to traffic generated by HAProxy, such as health-checks or error
6937 responses.
6938
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006939 See also: "rspdel" "reqadd", "http-response", section 6 about HTTP header
6940 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006941
6942
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006943rspdel <search> [{if | unless} <cond>]
6944rspidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006945 Delete all headers matching a regular expression in an HTTP response
6946 May be used in sections : defaults | frontend | listen | backend
6947 no | yes | yes | yes
6948 Arguments :
6949 <search> is the regular expression applied to HTTP headers and to the
6950 response line. This is an extended regular expression, so
6951 parenthesis grouping is supported and no preliminary backslash
6952 is required. Any space or known delimiter must be escaped using
6953 a backslash ('\'). The pattern applies to a full line at a time.
6954 The "rspdel" keyword strictly matches case while "rspidel"
6955 ignores case.
6956
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006957 <cond> is an optional matching condition built from ACLs. It makes it
6958 possible to ignore this rule when other conditions are not met.
6959
Willy Tarreau303c0352008-01-17 19:01:39 +01006960 Any header line matching extended regular expression <search> in the response
6961 will be completely deleted. Most common use of this is to remove unwanted
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02006962 and/or sensitive headers or cookies from a response before passing it to the
Willy Tarreau303c0352008-01-17 19:01:39 +01006963 client.
6964
6965 Header transformations only apply to traffic which passes through HAProxy,
6966 and not to traffic generated by HAProxy, such as health-checks or error
6967 responses. Keep in mind that header names are not case-sensitive.
6968
6969 Example :
6970 # remove the Server header from responses
Willy Tarreau5e80e022013-05-25 08:31:25 +02006971 rspidel ^Server:.*
Willy Tarreau303c0352008-01-17 19:01:39 +01006972
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006973 See also: "rspadd", "rsprep", "reqdel", "http-response", section 6 about
6974 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006975
6976
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006977rspdeny <search> [{if | unless} <cond>]
6978rspideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006979 Block an HTTP response if a line matches a regular expression
6980 May be used in sections : defaults | frontend | listen | backend
6981 no | yes | yes | yes
6982 Arguments :
6983 <search> is the regular expression applied to HTTP headers and to the
6984 response line. This is an extended regular expression, so
6985 parenthesis grouping is supported and no preliminary backslash
6986 is required. Any space or known delimiter must be escaped using
6987 a backslash ('\'). The pattern applies to a full line at a time.
6988 The "rspdeny" keyword strictly matches case while "rspideny"
6989 ignores case.
6990
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006991 <cond> is an optional matching condition built from ACLs. It makes it
6992 possible to ignore this rule when other conditions are not met.
6993
Willy Tarreau303c0352008-01-17 19:01:39 +01006994 A response containing any line which matches extended regular expression
6995 <search> will mark the request as denied. The test applies both to the
6996 response line and to response headers. Keep in mind that header names are not
6997 case-sensitive.
6998
6999 Main use of this keyword is to prevent sensitive information leak and to
Willy Tarreauced27012008-01-17 20:35:34 +01007000 block the response before it reaches the client. If a response is denied, it
7001 will be replaced with an HTTP 502 error so that the client never retrieves
7002 any sensitive data.
Willy Tarreau303c0352008-01-17 19:01:39 +01007003
7004 It is easier, faster and more powerful to use ACLs to write access policies.
7005 Rspdeny should be avoided in new designs.
7006
7007 Example :
7008 # Ensure that no content type matching ms-word will leak
7009 rspideny ^Content-type:\.*/ms-word
7010
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007011 See also: "reqdeny", "acl", "block", "http-response", section 6 about
7012 HTTP header manipulation and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007013
7014
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007015rsprep <search> <string> [{if | unless} <cond>]
7016rspirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01007017 Replace a regular expression with a string in an HTTP response line
7018 May be used in sections : defaults | frontend | listen | backend
7019 no | yes | yes | yes
7020 Arguments :
7021 <search> is the regular expression applied to HTTP headers and to the
7022 response line. This is an extended regular expression, so
7023 parenthesis grouping is supported and no preliminary backslash
7024 is required. Any space or known delimiter must be escaped using
7025 a backslash ('\'). The pattern applies to a full line at a time.
7026 The "rsprep" keyword strictly matches case while "rspirep"
7027 ignores case.
7028
7029 <string> is the complete line to be added. Any space or known delimiter
7030 must be escaped using a backslash ('\'). References to matched
7031 pattern groups are possible using the common \N form, with N
7032 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007033 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01007034
Willy Tarreaufdb563c2010-01-31 15:43:27 +01007035 <cond> is an optional matching condition built from ACLs. It makes it
7036 possible to ignore this rule when other conditions are not met.
7037
Willy Tarreau303c0352008-01-17 19:01:39 +01007038 Any line matching extended regular expression <search> in the response (both
7039 the response line and header lines) will be completely replaced with
7040 <string>. Most common use of this is to rewrite Location headers.
7041
7042 Header transformations only apply to traffic which passes through HAProxy,
7043 and not to traffic generated by HAProxy, such as health-checks or error
7044 responses. Note that for increased readability, it is suggested to add enough
7045 spaces between the request and the response. Keep in mind that header names
7046 are not case-sensitive.
7047
7048 Example :
7049 # replace "Location: 127.0.0.1:8080" with "Location: www.mydomain.com"
7050 rspirep ^Location:\ 127.0.0.1:8080 Location:\ www.mydomain.com
7051
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007052 See also: "rspadd", "rspdel", "reqrep", "http-response", section 6 about
7053 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007054
7055
David du Colombier486df472011-03-17 10:40:26 +01007056server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007057 Declare a server in a backend
7058 May be used in sections : defaults | frontend | listen | backend
7059 no | no | yes | yes
7060 Arguments :
7061 <name> is the internal name assigned to this server. This name will
Cyril Bonté941a0c62012-10-15 19:44:24 +02007062 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -05007063 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007064
David du Colombier486df472011-03-17 10:40:26 +01007065 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
7066 resolvable hostname is supported, but this name will be resolved
7067 during start-up. Address "0.0.0.0" or "*" has a special meaning.
7068 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +02007069 address as the one from the client connection. This is useful in
7070 transparent proxy architectures where the client's connection is
7071 intercepted and haproxy must forward to the original destination
7072 address. This is more or less what the "transparent" keyword does
7073 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +01007074 to report statistics. Optionally, an address family prefix may be
7075 used before the address to force the family regardless of the
7076 address format, which can be useful to specify a path to a unix
7077 socket with no slash ('/'). Currently supported prefixes are :
7078 - 'ipv4@' -> address is always IPv4
7079 - 'ipv6@' -> address is always IPv6
7080 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02007081 - 'abns@' -> address is in abstract namespace (Linux only)
William Lallemandb2f07452015-05-12 14:27:13 +02007082 You may want to reference some environment variables in the
7083 address parameter, see section 2.3 about environment
7084 variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007085
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007086 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007087 be sent to this port. If unset, the same port the client
7088 connected to will be used. The port may also be prefixed by a "+"
7089 or a "-". In this case, the server's port will be determined by
7090 adding this value to the client's port.
7091
7092 <param*> is a list of parameters for this server. The "server" keywords
7093 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007094 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007095
7096 Examples :
7097 server first 10.1.1.1:1080 cookie first check inter 1000
7098 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +01007099 server transp ipv4@
William Lallemandb2f07452015-05-12 14:27:13 +02007100 server backup "${SRV_BACKUP}:1080" backup
7101 server www1_dc1 "${LAN_DC1}.101:80"
7102 server www1_dc2 "${LAN_DC2}.101:80"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007103
Willy Tarreau55dcaf62015-09-27 15:03:15 +02007104 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
7105 sun_path length is used for the address length. Some other programs
7106 such as socat use the string length only by default. Pass the option
7107 ",unix-tightsocklen=0" to any abstract socket definition in socat to
7108 make it compatible with HAProxy's.
7109
Mark Lamourinec2247f02012-01-04 13:02:01 -05007110 See also: "default-server", "http-send-name-header" and section 5 about
7111 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007112
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007113server-state-file-name [<file>]
7114 Set the server state file to read, load and apply to servers available in
7115 this backend. It only applies when the directive "load-server-state-from-file"
7116 is set to "local". When <file> is not provided or if this directive is not
7117 set, then backend name is used. If <file> starts with a slash '/', then it is
7118 considered as an absolute path. Otherwise, <file> is concatenated to the
7119 global directive "server-state-file-base".
7120
7121 Example: the minimal configuration below would make HAProxy look for the
7122 state server file '/etc/haproxy/states/bk':
7123
7124 global
7125 server-state-file-base /etc/haproxy/states
7126
7127 backend bk
7128 load-server-state-from-file
7129
7130 See also: "server-state-file-base", "load-server-state-from-file", and
7131 "show servers state"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007132
7133source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02007134source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +01007135source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007136 Set the source address for outgoing connections
7137 May be used in sections : defaults | frontend | listen | backend
7138 yes | no | yes | yes
7139 Arguments :
7140 <addr> is the IPv4 address HAProxy will bind to before connecting to a
7141 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +01007142
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007143 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +01007144 the most appropriate address to reach its destination. Optionally
7145 an address family prefix may be used before the address to force
7146 the family regardless of the address format, which can be useful
7147 to specify a path to a unix socket with no slash ('/'). Currently
7148 supported prefixes are :
7149 - 'ipv4@' -> address is always IPv4
7150 - 'ipv6@' -> address is always IPv6
7151 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02007152 - 'abns@' -> address is in abstract namespace (Linux only)
Cyril Bonté307ee1e2015-09-28 23:16:06 +02007153 You may want to reference some environment variables in the
7154 address parameter, see section 2.3 about environment variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007155
7156 <port> is an optional port. It is normally not needed but may be useful
7157 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02007158 the system will select a free port. Note that port ranges are not
7159 supported in the backend. If you want to force port ranges, you
7160 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007161
7162 <addr2> is the IP address to present to the server when connections are
7163 forwarded in full transparent proxy mode. This is currently only
7164 supported on some patched Linux kernels. When this address is
7165 specified, clients connecting to the server will be presented
7166 with this address, while health checks will still use the address
7167 <addr>.
7168
7169 <port2> is the optional port to present to the server when connections
7170 are forwarded in full transparent proxy mode (see <addr2> above).
7171 The default value of zero means the system will select a free
7172 port.
7173
Willy Tarreaubce70882009-09-07 11:51:47 +02007174 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
7175 This is the name of a comma-separated header list which can
7176 contain multiple IP addresses. By default, the last occurrence is
7177 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +01007178 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +02007179 by previous proxy, typically Stunnel. In order to use another
7180 occurrence from the last one, please see the <occ> parameter
7181 below. When the header (or occurrence) is not found, no binding
7182 is performed so that the proxy's default IP address is used. Also
7183 keep in mind that the header name is case insensitive, as for any
7184 HTTP header.
7185
7186 <occ> is the occurrence number of a value to be used in a multi-value
7187 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007188 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +02007189 address. Positive values indicate a position from the first
7190 occurrence, 1 being the first one. Negative values indicate
7191 positions relative to the last one, -1 being the last one. This
7192 is helpful for situations where an X-Forwarded-For header is set
7193 at the entry point of an infrastructure and must be used several
7194 proxy layers away. When this value is not specified, -1 is
7195 assumed. Passing a zero here disables the feature.
7196
Willy Tarreaud53f96b2009-02-04 18:46:54 +01007197 <name> is an optional interface name to which to bind to for outgoing
7198 traffic. On systems supporting this features (currently, only
7199 Linux), this allows one to bind all traffic to the server to
7200 this interface even if it is not the one the system would select
7201 based on routing tables. This should be used with extreme care.
7202 Note that using this option requires root privileges.
7203
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007204 The "source" keyword is useful in complex environments where a specific
7205 address only is allowed to connect to the servers. It may be needed when a
7206 private address must be used through a public gateway for instance, and it is
7207 known that the system cannot determine the adequate source address by itself.
7208
7209 An extension which is available on certain patched Linux kernels may be used
7210 through the "usesrc" optional keyword. It makes it possible to connect to the
7211 servers with an IP address which does not belong to the system itself. This
7212 is called "full transparent proxy mode". For this to work, the destination
7213 servers have to route their traffic back to this address through the machine
7214 running HAProxy, and IP forwarding must generally be enabled on this machine.
7215
7216 In this "full transparent proxy" mode, it is possible to force a specific IP
7217 address to be presented to the servers. This is not much used in fact. A more
7218 common use is to tell HAProxy to present the client's IP address. For this,
7219 there are two methods :
7220
7221 - present the client's IP and port addresses. This is the most transparent
7222 mode, but it can cause problems when IP connection tracking is enabled on
7223 the machine, because a same connection may be seen twice with different
7224 states. However, this solution presents the huge advantage of not
7225 limiting the system to the 64k outgoing address+port couples, because all
7226 of the client ranges may be used.
7227
7228 - present only the client's IP address and select a spare port. This
7229 solution is still quite elegant but slightly less transparent (downstream
7230 firewalls logs will not match upstream's). It also presents the downside
7231 of limiting the number of concurrent connections to the usual 64k ports.
7232 However, since the upstream and downstream ports are different, local IP
7233 connection tracking on the machine will not be upset by the reuse of the
7234 same session.
7235
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007236 This option sets the default source for all servers in the backend. It may
7237 also be specified in a "defaults" section. Finer source address specification
7238 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007239 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007240
Baptiste Assmann91bd3372015-07-17 21:59:42 +02007241 In order to work, "usesrc" requires root privileges.
7242
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007243 Examples :
7244 backend private
7245 # Connect to the servers using our 192.168.1.200 source address
7246 source 192.168.1.200
7247
7248 backend transparent_ssl1
7249 # Connect to the SSL farm from the client's source address
7250 source 192.168.1.200 usesrc clientip
7251
7252 backend transparent_ssl2
7253 # Connect to the SSL farm from the client's source address and port
7254 # not recommended if IP conntrack is present on the local machine.
7255 source 192.168.1.200 usesrc client
7256
7257 backend transparent_ssl3
7258 # Connect to the SSL farm from the client's source address. It
7259 # is more conntrack-friendly.
7260 source 192.168.1.200 usesrc clientip
7261
7262 backend transparent_smtp
7263 # Connect to the SMTP farm from the client's source address/port
7264 # with Tproxy version 4.
7265 source 0.0.0.0 usesrc clientip
7266
Willy Tarreaubce70882009-09-07 11:51:47 +02007267 backend transparent_http
7268 # Connect to the servers using the client's IP as seen by previous
7269 # proxy.
7270 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
7271
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007272 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007273 the Linux kernel on www.balabit.com, the "bind" keyword.
7274
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007275
Willy Tarreau844e3c52008-01-11 16:28:18 +01007276srvtimeout <timeout> (deprecated)
7277 Set the maximum inactivity time on the server side.
7278 May be used in sections : defaults | frontend | listen | backend
7279 yes | no | yes | yes
7280 Arguments :
7281 <timeout> is the timeout value specified in milliseconds by default, but
7282 can be in any other unit if the number is suffixed by the unit,
7283 as explained at the top of this document.
7284
7285 The inactivity timeout applies when the server is expected to acknowledge or
7286 send data. In HTTP mode, this timeout is particularly important to consider
7287 during the first phase of the server's response, when it has to send the
7288 headers, as it directly represents the server's processing time for the
7289 request. To find out what value to put there, it's often good to start with
7290 what would be considered as unacceptable response times, then check the logs
7291 to observe the response time distribution, and adjust the value accordingly.
7292
7293 The value is specified in milliseconds by default, but can be in any other
7294 unit if the number is suffixed by the unit, as specified at the top of this
7295 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
7296 recommended that the client timeout remains equal to the server timeout in
7297 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007298 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01007299 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreaud72758d2010-01-12 10:42:19 +01007300 seconds (eg: 4 or 5 seconds minimum).
Willy Tarreau844e3c52008-01-11 16:28:18 +01007301
7302 This parameter is specific to backends, but can be specified once for all in
7303 "defaults" sections. This is in fact one of the easiest solutions not to
7304 forget about it. An unspecified timeout results in an infinite timeout, which
7305 is not recommended. Such a usage is accepted and works but reports a warning
7306 during startup because it may results in accumulation of expired sessions in
7307 the system if the system's timeouts are not configured either.
7308
7309 This parameter is provided for compatibility but is currently deprecated.
7310 Please use "timeout server" instead.
7311
Willy Tarreauce887fd2012-05-12 12:50:00 +02007312 See also : "timeout server", "timeout tunnel", "timeout client" and
7313 "clitimeout".
Willy Tarreau844e3c52008-01-11 16:28:18 +01007314
7315
Cyril Bonté66c327d2010-10-12 00:14:37 +02007316stats admin { if | unless } <cond>
7317 Enable statistics admin level if/unless a condition is matched
7318 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007319 no | yes | yes | yes
Cyril Bonté66c327d2010-10-12 00:14:37 +02007320
7321 This statement enables the statistics admin level if/unless a condition is
7322 matched.
7323
7324 The admin level allows to enable/disable servers from the web interface. By
7325 default, statistics page is read-only for security reasons.
7326
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007327 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7328 unless you know what you do : memory is not shared between the
7329 processes, which can result in random behaviours.
7330
Cyril Bonté23b39d92011-02-10 22:54:44 +01007331 Currently, the POST request is limited to the buffer size minus the reserved
7332 buffer space, which means that if the list of servers is too long, the
7333 request won't be processed. It is recommended to alter few servers at a
7334 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +02007335
7336 Example :
7337 # statistics admin level only for localhost
7338 backend stats_localhost
7339 stats enable
7340 stats admin if LOCALHOST
7341
7342 Example :
7343 # statistics admin level always enabled because of the authentication
7344 backend stats_auth
7345 stats enable
7346 stats auth admin:AdMiN123
7347 stats admin if TRUE
7348
7349 Example :
7350 # statistics admin level depends on the authenticated user
7351 userlist stats-auth
7352 group admin users admin
7353 user admin insecure-password AdMiN123
7354 group readonly users haproxy
7355 user haproxy insecure-password haproxy
7356
7357 backend stats_auth
7358 stats enable
7359 acl AUTH http_auth(stats-auth)
7360 acl AUTH_ADMIN http_auth_group(stats-auth) admin
7361 stats http-request auth unless AUTH
7362 stats admin if AUTH_ADMIN
7363
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007364 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
7365 "bind-process", section 3.4 about userlists and section 7 about
7366 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +02007367
7368
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007369stats auth <user>:<passwd>
7370 Enable statistics with authentication and grant access to an account
7371 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007372 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007373 Arguments :
7374 <user> is a user name to grant access to
7375
7376 <passwd> is the cleartext password associated to this user
7377
7378 This statement enables statistics with default settings, and restricts access
7379 to declared users only. It may be repeated as many times as necessary to
7380 allow as many users as desired. When a user tries to access the statistics
7381 without a valid account, a "401 Forbidden" response will be returned so that
7382 the browser asks the user to provide a valid user and password. The real
7383 which will be returned to the browser is configurable using "stats realm".
7384
7385 Since the authentication method is HTTP Basic Authentication, the passwords
7386 circulate in cleartext on the network. Thus, it was decided that the
7387 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02007388 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007389
7390 It is also possible to reduce the scope of the proxies which appear in the
7391 report using "stats scope".
7392
7393 Though this statement alone is enough to enable statistics reporting, it is
7394 recommended to set all other settings in order to avoid relying on default
7395 unobvious parameters.
7396
7397 Example :
7398 # public access (limited to this backend only)
7399 backend public_www
7400 server srv1 192.168.0.1:80
7401 stats enable
7402 stats hide-version
7403 stats scope .
7404 stats uri /admin?stats
7405 stats realm Haproxy\ Statistics
7406 stats auth admin1:AdMiN123
7407 stats auth admin2:AdMiN321
7408
7409 # internal monitoring access (unlimited)
7410 backend private_monitoring
7411 stats enable
7412 stats uri /admin?stats
7413 stats refresh 5s
7414
7415 See also : "stats enable", "stats realm", "stats scope", "stats uri"
7416
7417
7418stats enable
7419 Enable statistics reporting with default settings
7420 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007421 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007422 Arguments : none
7423
7424 This statement enables statistics reporting with default settings defined
7425 at build time. Unless stated otherwise, these settings are used :
7426 - stats uri : /haproxy?stats
7427 - stats realm : "HAProxy Statistics"
7428 - stats auth : no authentication
7429 - stats scope : no restriction
7430
7431 Though this statement alone is enough to enable statistics reporting, it is
7432 recommended to set all other settings in order to avoid relying on default
7433 unobvious parameters.
7434
7435 Example :
7436 # public access (limited to this backend only)
7437 backend public_www
7438 server srv1 192.168.0.1:80
7439 stats enable
7440 stats hide-version
7441 stats scope .
7442 stats uri /admin?stats
7443 stats realm Haproxy\ Statistics
7444 stats auth admin1:AdMiN123
7445 stats auth admin2:AdMiN321
7446
7447 # internal monitoring access (unlimited)
7448 backend private_monitoring
7449 stats enable
7450 stats uri /admin?stats
7451 stats refresh 5s
7452
7453 See also : "stats auth", "stats realm", "stats uri"
7454
7455
Willy Tarreaud63335a2010-02-26 12:56:52 +01007456stats hide-version
7457 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007458 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007459 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01007460 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007461
Willy Tarreaud63335a2010-02-26 12:56:52 +01007462 By default, the stats page reports some useful status information along with
7463 the statistics. Among them is HAProxy's version. However, it is generally
7464 considered dangerous to report precise version to anyone, as it can help them
7465 target known weaknesses with specific attacks. The "stats hide-version"
7466 statement removes the version from the statistics report. This is recommended
7467 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007468
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02007469 Though this statement alone is enough to enable statistics reporting, it is
7470 recommended to set all other settings in order to avoid relying on default
7471 unobvious parameters.
7472
Willy Tarreaud63335a2010-02-26 12:56:52 +01007473 Example :
7474 # public access (limited to this backend only)
7475 backend public_www
7476 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02007477 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01007478 stats hide-version
7479 stats scope .
7480 stats uri /admin?stats
7481 stats realm Haproxy\ Statistics
7482 stats auth admin1:AdMiN123
7483 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007484
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007485 # internal monitoring access (unlimited)
7486 backend private_monitoring
7487 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01007488 stats uri /admin?stats
7489 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +01007490
Willy Tarreaud63335a2010-02-26 12:56:52 +01007491 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02007492
Willy Tarreau983e01e2010-01-11 18:42:06 +01007493
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02007494stats http-request { allow | deny | auth [realm <realm>] }
7495 [ { if | unless } <condition> ]
7496 Access control for statistics
7497
7498 May be used in sections: defaults | frontend | listen | backend
7499 no | no | yes | yes
7500
7501 As "http-request", these set of options allow to fine control access to
7502 statistics. Each option may be followed by if/unless and acl.
7503 First option with matched condition (or option without condition) is final.
7504 For "deny" a 403 error will be returned, for "allow" normal processing is
7505 performed, for "auth" a 401/407 error code is returned so the client
7506 should be asked to enter a username and password.
7507
7508 There is no fixed limit to the number of http-request statements per
7509 instance.
7510
7511 See also : "http-request", section 3.4 about userlists and section 7
7512 about ACL usage.
7513
7514
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007515stats realm <realm>
7516 Enable statistics and set authentication realm
7517 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007518 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007519 Arguments :
7520 <realm> is the name of the HTTP Basic Authentication realm reported to
7521 the browser. The browser uses it to display it in the pop-up
7522 inviting the user to enter a valid username and password.
7523
7524 The realm is read as a single word, so any spaces in it should be escaped
7525 using a backslash ('\').
7526
7527 This statement is useful only in conjunction with "stats auth" since it is
7528 only related to authentication.
7529
7530 Though this statement alone is enough to enable statistics reporting, it is
7531 recommended to set all other settings in order to avoid relying on default
7532 unobvious parameters.
7533
7534 Example :
7535 # public access (limited to this backend only)
7536 backend public_www
7537 server srv1 192.168.0.1:80
7538 stats enable
7539 stats hide-version
7540 stats scope .
7541 stats uri /admin?stats
7542 stats realm Haproxy\ Statistics
7543 stats auth admin1:AdMiN123
7544 stats auth admin2:AdMiN321
7545
7546 # internal monitoring access (unlimited)
7547 backend private_monitoring
7548 stats enable
7549 stats uri /admin?stats
7550 stats refresh 5s
7551
7552 See also : "stats auth", "stats enable", "stats uri"
7553
7554
7555stats refresh <delay>
7556 Enable statistics with automatic refresh
7557 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007558 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007559 Arguments :
7560 <delay> is the suggested refresh delay, specified in seconds, which will
7561 be returned to the browser consulting the report page. While the
7562 browser is free to apply any delay, it will generally respect it
7563 and refresh the page this every seconds. The refresh interval may
7564 be specified in any other non-default time unit, by suffixing the
7565 unit after the value, as explained at the top of this document.
7566
7567 This statement is useful on monitoring displays with a permanent page
7568 reporting the load balancer's activity. When set, the HTML report page will
7569 include a link "refresh"/"stop refresh" so that the user can select whether
7570 he wants automatic refresh of the page or not.
7571
7572 Though this statement alone is enough to enable statistics reporting, it is
7573 recommended to set all other settings in order to avoid relying on default
7574 unobvious parameters.
7575
7576 Example :
7577 # public access (limited to this backend only)
7578 backend public_www
7579 server srv1 192.168.0.1:80
7580 stats enable
7581 stats hide-version
7582 stats scope .
7583 stats uri /admin?stats
7584 stats realm Haproxy\ Statistics
7585 stats auth admin1:AdMiN123
7586 stats auth admin2:AdMiN321
7587
7588 # internal monitoring access (unlimited)
7589 backend private_monitoring
7590 stats enable
7591 stats uri /admin?stats
7592 stats refresh 5s
7593
7594 See also : "stats auth", "stats enable", "stats realm", "stats uri"
7595
7596
7597stats scope { <name> | "." }
7598 Enable statistics and limit access scope
7599 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007600 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007601 Arguments :
7602 <name> is the name of a listen, frontend or backend section to be
7603 reported. The special name "." (a single dot) designates the
7604 section in which the statement appears.
7605
7606 When this statement is specified, only the sections enumerated with this
7607 statement will appear in the report. All other ones will be hidden. This
7608 statement may appear as many times as needed if multiple sections need to be
7609 reported. Please note that the name checking is performed as simple string
7610 comparisons, and that it is never checked that a give section name really
7611 exists.
7612
7613 Though this statement alone is enough to enable statistics reporting, it is
7614 recommended to set all other settings in order to avoid relying on default
7615 unobvious parameters.
7616
7617 Example :
7618 # public access (limited to this backend only)
7619 backend public_www
7620 server srv1 192.168.0.1:80
7621 stats enable
7622 stats hide-version
7623 stats scope .
7624 stats uri /admin?stats
7625 stats realm Haproxy\ Statistics
7626 stats auth admin1:AdMiN123
7627 stats auth admin2:AdMiN321
7628
7629 # internal monitoring access (unlimited)
7630 backend private_monitoring
7631 stats enable
7632 stats uri /admin?stats
7633 stats refresh 5s
7634
7635 See also : "stats auth", "stats enable", "stats realm", "stats uri"
7636
Willy Tarreaud63335a2010-02-26 12:56:52 +01007637
Willy Tarreauc9705a12010-07-27 20:05:50 +02007638stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +01007639 Enable reporting of a description on the statistics page.
7640 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007641 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01007642
Willy Tarreauc9705a12010-07-27 20:05:50 +02007643 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +01007644 description from global section is automatically used instead.
7645
7646 This statement is useful for users that offer shared services to their
7647 customers, where node or description should be different for each customer.
7648
7649 Though this statement alone is enough to enable statistics reporting, it is
7650 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007651 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +01007652
7653 Example :
7654 # internal monitoring access (unlimited)
7655 backend private_monitoring
7656 stats enable
7657 stats show-desc Master node for Europe, Asia, Africa
7658 stats uri /admin?stats
7659 stats refresh 5s
7660
7661 See also: "show-node", "stats enable", "stats uri" and "description" in
7662 global section.
7663
7664
7665stats show-legends
Willy Tarreaued2119c2014-04-24 22:10:39 +02007666 Enable reporting additional information on the statistics page
7667 May be used in sections : defaults | frontend | listen | backend
7668 yes | yes | yes | yes
7669 Arguments : none
7670
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007671 Enable reporting additional information on the statistics page :
Willy Tarreaud63335a2010-02-26 12:56:52 +01007672 - cap: capabilities (proxy)
7673 - mode: one of tcp, http or health (proxy)
7674 - id: SNMP ID (proxy, socket, server)
7675 - IP (socket, server)
7676 - cookie (backend, server)
7677
7678 Though this statement alone is enough to enable statistics reporting, it is
7679 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007680 unobvious parameters. Default behaviour is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +01007681
7682 See also: "stats enable", "stats uri".
7683
7684
7685stats show-node [ <name> ]
7686 Enable reporting of a host name on the statistics page.
7687 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007688 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01007689 Arguments:
7690 <name> is an optional name to be reported. If unspecified, the
7691 node name from global section is automatically used instead.
7692
7693 This statement is useful for users that offer shared services to their
7694 customers, where node or description might be different on a stats page
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007695 provided for each customer. Default behaviour is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +01007696
7697 Though this statement alone is enough to enable statistics reporting, it is
7698 recommended to set all other settings in order to avoid relying on default
7699 unobvious parameters.
7700
7701 Example:
7702 # internal monitoring access (unlimited)
7703 backend private_monitoring
7704 stats enable
7705 stats show-node Europe-1
7706 stats uri /admin?stats
7707 stats refresh 5s
7708
7709 See also: "show-desc", "stats enable", "stats uri", and "node" in global
7710 section.
7711
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007712
7713stats uri <prefix>
7714 Enable statistics and define the URI prefix to access them
7715 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02007716 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007717 Arguments :
7718 <prefix> is the prefix of any URI which will be redirected to stats. This
7719 prefix may contain a question mark ('?') to indicate part of a
7720 query string.
7721
7722 The statistics URI is intercepted on the relayed traffic, so it appears as a
7723 page within the normal application. It is strongly advised to ensure that the
7724 selected URI will never appear in the application, otherwise it will never be
7725 possible to reach it in the application.
7726
7727 The default URI compiled in haproxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007728 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007729 It is generally a good idea to include a question mark in the URI so that
7730 intermediate proxies refrain from caching the results. Also, since any string
7731 beginning with the prefix will be accepted as a stats request, the question
7732 mark helps ensuring that no valid URI will begin with the same words.
7733
7734 It is sometimes very convenient to use "/" as the URI prefix, and put that
7735 statement in a "listen" instance of its own. That makes it easy to dedicate
7736 an address or a port to statistics only.
7737
7738 Though this statement alone is enough to enable statistics reporting, it is
7739 recommended to set all other settings in order to avoid relying on default
7740 unobvious parameters.
7741
7742 Example :
7743 # public access (limited to this backend only)
7744 backend public_www
7745 server srv1 192.168.0.1:80
7746 stats enable
7747 stats hide-version
7748 stats scope .
7749 stats uri /admin?stats
7750 stats realm Haproxy\ Statistics
7751 stats auth admin1:AdMiN123
7752 stats auth admin2:AdMiN321
7753
7754 # internal monitoring access (unlimited)
7755 backend private_monitoring
7756 stats enable
7757 stats uri /admin?stats
7758 stats refresh 5s
7759
7760 See also : "stats auth", "stats enable", "stats realm"
7761
7762
Willy Tarreaud63335a2010-02-26 12:56:52 +01007763stick match <pattern> [table <table>] [{if | unless} <cond>]
7764 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007765 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +01007766 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007767
7768 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02007769 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007770 describes what elements of the incoming request or connection
7771 will be analysed in the hope to find a matching entry in a
7772 stickiness table. This rule is mandatory.
7773
7774 <table> is an optional stickiness table name. If unspecified, the same
7775 backend's table is used. A stickiness table is declared using
7776 the "stick-table" statement.
7777
7778 <cond> is an optional matching condition. It makes it possible to match
7779 on a certain criterion only when other conditions are met (or
7780 not met). For instance, it could be used to match on a source IP
7781 address except when a request passes through a known proxy, in
7782 which case we'd match on a header containing that IP address.
7783
7784 Some protocols or applications require complex stickiness rules and cannot
7785 always simply rely on cookies nor hashing. The "stick match" statement
7786 describes a rule to extract the stickiness criterion from an incoming request
7787 or connection. See section 7 for a complete list of possible patterns and
7788 transformation rules.
7789
7790 The table has to be declared using the "stick-table" statement. It must be of
7791 a type compatible with the pattern. By default it is the one which is present
7792 in the same backend. It is possible to share a table with other backends by
7793 referencing it using the "table" keyword. If another table is referenced,
7794 the server's ID inside the backends are used. By default, all server IDs
7795 start at 1 in each backend, so the server ordering is enough. But in case of
7796 doubt, it is highly recommended to force server IDs using their "id" setting.
7797
7798 It is possible to restrict the conditions where a "stick match" statement
7799 will apply, using "if" or "unless" followed by a condition. See section 7 for
7800 ACL based conditions.
7801
7802 There is no limit on the number of "stick match" statements. The first that
7803 applies and matches will cause the request to be directed to the same server
7804 as was used for the request which created the entry. That way, multiple
7805 matches can be used as fallbacks.
7806
7807 The stick rules are checked after the persistence cookies, so they will not
7808 affect stickiness if a cookie has already been used to select a server. That
7809 way, it becomes very easy to insert cookies and match on IP addresses in
7810 order to maintain stickiness between HTTP and HTTPS.
7811
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007812 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7813 unless you know what you do : memory is not shared between the
7814 processes, which can result in random behaviours.
7815
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007816 Example :
7817 # forward SMTP users to the same server they just used for POP in the
7818 # last 30 minutes
7819 backend pop
7820 mode tcp
7821 balance roundrobin
7822 stick store-request src
7823 stick-table type ip size 200k expire 30m
7824 server s1 192.168.1.1:110
7825 server s2 192.168.1.1:110
7826
7827 backend smtp
7828 mode tcp
7829 balance roundrobin
7830 stick match src table pop
7831 server s1 192.168.1.1:25
7832 server s2 192.168.1.1:25
7833
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007834 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02007835 about ACLs and samples fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007836
7837
7838stick on <pattern> [table <table>] [{if | unless} <condition>]
7839 Define a request pattern to associate a user to a server
7840 May be used in sections : defaults | frontend | listen | backend
7841 no | no | yes | yes
7842
7843 Note : This form is exactly equivalent to "stick match" followed by
7844 "stick store-request", all with the same arguments. Please refer
7845 to both keywords for details. It is only provided as a convenience
7846 for writing more maintainable configurations.
7847
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007848 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7849 unless you know what you do : memory is not shared between the
7850 processes, which can result in random behaviours.
7851
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007852 Examples :
7853 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +01007854 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007855
7856 # ...is strictly equivalent to this one :
7857 stick match src table pop if !localhost
7858 stick store-request src table pop if !localhost
7859
7860
7861 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
7862 # well as HTTP without cookie. Share the same table between both accesses.
7863 backend http
7864 mode http
7865 balance roundrobin
7866 stick on src table https
7867 cookie SRV insert indirect nocache
7868 server s1 192.168.1.1:80 cookie s1
7869 server s2 192.168.1.1:80 cookie s2
7870
7871 backend https
7872 mode tcp
7873 balance roundrobin
7874 stick-table type ip size 200k expire 30m
7875 stick on src
7876 server s1 192.168.1.1:443
7877 server s2 192.168.1.1:443
7878
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007879 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007880
7881
7882stick store-request <pattern> [table <table>] [{if | unless} <condition>]
7883 Define a request pattern used to create an entry in a stickiness table
7884 May be used in sections : defaults | frontend | listen | backend
7885 no | no | yes | yes
7886
7887 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02007888 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007889 describes what elements of the incoming request or connection
7890 will be analysed, extracted and stored in the table once a
7891 server is selected.
7892
7893 <table> is an optional stickiness table name. If unspecified, the same
7894 backend's table is used. A stickiness table is declared using
7895 the "stick-table" statement.
7896
7897 <cond> is an optional storage condition. It makes it possible to store
7898 certain criteria only when some conditions are met (or not met).
7899 For instance, it could be used to store the source IP address
7900 except when the request passes through a known proxy, in which
7901 case we'd store a converted form of a header containing that IP
7902 address.
7903
7904 Some protocols or applications require complex stickiness rules and cannot
7905 always simply rely on cookies nor hashing. The "stick store-request" statement
7906 describes a rule to decide what to extract from the request and when to do
7907 it, in order to store it into a stickiness table for further requests to
7908 match it using the "stick match" statement. Obviously the extracted part must
7909 make sense and have a chance to be matched in a further request. Storing a
7910 client's IP address for instance often makes sense. Storing an ID found in a
7911 URL parameter also makes sense. Storing a source port will almost never make
7912 any sense because it will be randomly matched. See section 7 for a complete
7913 list of possible patterns and transformation rules.
7914
7915 The table has to be declared using the "stick-table" statement. It must be of
7916 a type compatible with the pattern. By default it is the one which is present
7917 in the same backend. It is possible to share a table with other backends by
7918 referencing it using the "table" keyword. If another table is referenced,
7919 the server's ID inside the backends are used. By default, all server IDs
7920 start at 1 in each backend, so the server ordering is enough. But in case of
7921 doubt, it is highly recommended to force server IDs using their "id" setting.
7922
7923 It is possible to restrict the conditions where a "stick store-request"
7924 statement will apply, using "if" or "unless" followed by a condition. This
7925 condition will be evaluated while parsing the request, so any criteria can be
7926 used. See section 7 for ACL based conditions.
7927
7928 There is no limit on the number of "stick store-request" statements, but
7929 there is a limit of 8 simultaneous stores per request or response. This
7930 makes it possible to store up to 8 criteria, all extracted from either the
7931 request or the response, regardless of the number of rules. Only the 8 first
7932 ones which match will be kept. Using this, it is possible to feed multiple
7933 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01007934 another protocol or access method. Using multiple store-request rules with
7935 the same table is possible and may be used to find the best criterion to rely
7936 on, by arranging the rules by decreasing preference order. Only the first
7937 extracted criterion for a given table will be stored. All subsequent store-
7938 request rules referencing the same table will be skipped and their ACLs will
7939 not be evaluated.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007940
7941 The "store-request" rules are evaluated once the server connection has been
7942 established, so that the table will contain the real server that processed
7943 the request.
7944
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007945 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7946 unless you know what you do : memory is not shared between the
7947 processes, which can result in random behaviours.
7948
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007949 Example :
7950 # forward SMTP users to the same server they just used for POP in the
7951 # last 30 minutes
7952 backend pop
7953 mode tcp
7954 balance roundrobin
7955 stick store-request src
7956 stick-table type ip size 200k expire 30m
7957 server s1 192.168.1.1:110
7958 server s2 192.168.1.1:110
7959
7960 backend smtp
7961 mode tcp
7962 balance roundrobin
7963 stick match src table pop
7964 server s1 192.168.1.1:25
7965 server s2 192.168.1.1:25
7966
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007967 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02007968 about ACLs and sample fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007969
7970
Emeric Brun7c6b82e2010-09-24 16:34:28 +02007971stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Emeric Brunf099e792010-09-27 12:05:28 +02007972 size <size> [expire <expire>] [nopurge] [peers <peersect>]
7973 [store <data_type>]*
Godbach64cef792013-12-04 16:08:22 +08007974 Configure the stickiness table for the current section
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007975 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +02007976 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007977
7978 Arguments :
7979 ip a table declared with "type ip" will only store IPv4 addresses.
7980 This form is very compact (about 50 bytes per entry) and allows
7981 very fast entry lookup and stores with almost no overhead. This
7982 is mainly used to store client source IP addresses.
7983
David du Colombier9a6d3c92011-03-17 10:40:24 +01007984 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
7985 This form is very compact (about 60 bytes per entry) and allows
7986 very fast entry lookup and stores with almost no overhead. This
7987 is mainly used to store client source IP addresses.
7988
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007989 integer a table declared with "type integer" will store 32bit integers
7990 which can represent a client identifier found in a request for
7991 instance.
7992
7993 string a table declared with "type string" will store substrings of up
7994 to <len> characters. If the string provided by the pattern
7995 extractor is larger than <len>, it will be truncated before
7996 being stored. During matching, at most <len> characters will be
7997 compared between the string in the table and the extracted
7998 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +02007999 to 32 characters.
8000
8001 binary a table declared with "type binary" will store binary blocks
8002 of <len> bytes. If the block provided by the pattern
8003 extractor is larger than <len>, it will be truncated before
Willy Tarreaube722a22014-06-13 16:31:59 +02008004 being stored. If the block provided by the sample expression
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008005 is shorter than <len>, it will be padded by 0. When not
8006 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008007
8008 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008009 "string" type table (See type "string" above). Or the number
8010 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008011 changing this parameter as memory usage will proportionally
8012 increase.
8013
8014 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +01008015 value directly impacts memory usage. Count approximately
8016 50 bytes per entry, plus the size of a string if any. The size
8017 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008018
8019 [nopurge] indicates that we refuse to purge older entries when the table
8020 is full. When not specified and the table is full when haproxy
8021 wants to store an entry in it, it will flush a few of the oldest
8022 entries in order to release some space for the new ones. This is
8023 most often the desired behaviour. In some specific cases, it
8024 be desirable to refuse new entries instead of purging the older
8025 ones. That may be the case when the amount of data to store is
8026 far above the hardware limits and we prefer not to offer access
8027 to new clients than to reject the ones already connected. When
8028 using this parameter, be sure to properly set the "expire"
8029 parameter (see below).
8030
Emeric Brunf099e792010-09-27 12:05:28 +02008031 <peersect> is the name of the peers section to use for replication. Entries
8032 which associate keys to server IDs are kept synchronized with
8033 the remote peers declared in this section. All entries are also
8034 automatically learned from the local peer (old process) during a
8035 soft restart.
8036
Willy Tarreau1abc6732015-05-01 19:21:02 +02008037 NOTE : each peers section may be referenced only by tables
8038 belonging to the same unique process.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008039
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008040 <expire> defines the maximum duration of an entry in the table since it
8041 was last created, refreshed or matched. The expiration delay is
8042 defined using the standard time format, similarly as the various
8043 timeouts. The maximum duration is slightly above 24 days. See
8044 section 2.2 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +02008045 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008046 be removed once full. Be sure not to use the "nopurge" parameter
8047 if not expiration delay is specified.
8048
Willy Tarreau08d5f982010-06-06 13:34:54 +02008049 <data_type> is used to store additional information in the stick-table. This
8050 may be used by ACLs in order to control various criteria related
8051 to the activity of the client matching the stick-table. For each
8052 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +02008053 that the additional data can fit. Several data types may be
8054 stored with an entry. Multiple data types may be specified after
8055 the "store" keyword, as a comma-separated list. Alternatively,
8056 it is possible to repeat the "store" keyword followed by one or
8057 several data types. Except for the "server_id" type which is
8058 automatically detected and enabled, all data types must be
8059 explicitly declared to be stored. If an ACL references a data
8060 type which is not stored, the ACL will simply not match. Some
8061 data types require an argument which must be passed just after
8062 the type between parenthesis. See below for the supported data
8063 types and their arguments.
8064
8065 The data types that can be stored with an entry are the following :
8066 - server_id : this is an integer which holds the numeric ID of the server a
8067 request was assigned to. It is used by the "stick match", "stick store",
8068 and "stick on" rules. It is automatically enabled when referenced.
8069
8070 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
8071 integer which may be used for anything. Most of the time it will be used
8072 to put a special tag on some entries, for instance to note that a
8073 specific behaviour was detected and must be known for future matches.
8074
Willy Tarreauba2ffd12013-05-29 15:54:14 +02008075 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
8076 over a period. It is a positive 32-bit integer integer which may be used
8077 for anything. Just like <gpc0>, it counts events, but instead of keeping
8078 a cumulative count, it maintains the rate at which the counter is
8079 incremented. Most of the time it will be used to measure the frequency of
8080 occurrence of certain events (eg: requests to a specific URL).
8081
Willy Tarreauc9705a12010-07-27 20:05:50 +02008082 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
8083 the absolute number of connections received from clients which matched
8084 this entry. It does not mean the connections were accepted, just that
8085 they were received.
8086
8087 - conn_cur : Current Connections. It is a positive 32-bit integer which
8088 stores the concurrent connection counts for the entry. It is incremented
8089 once an incoming connection matches the entry, and decremented once the
8090 connection leaves. That way it is possible to know at any time the exact
8091 number of concurrent connections for an entry.
8092
8093 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8094 integer parameter <period> which indicates in milliseconds the length
8095 of the period over which the average is measured. It reports the average
8096 incoming connection rate over that period, in connections per period. The
8097 result is an integer which can be matched using ACLs.
8098
8099 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
8100 the absolute number of sessions received from clients which matched this
8101 entry. A session is a connection that was accepted by the layer 4 rules.
8102
8103 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8104 integer parameter <period> which indicates in milliseconds the length
8105 of the period over which the average is measured. It reports the average
8106 incoming session rate over that period, in sessions per period. The
8107 result is an integer which can be matched using ACLs.
8108
8109 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
8110 counts the absolute number of HTTP requests received from clients which
8111 matched this entry. It does not matter whether they are valid requests or
8112 not. Note that this is different from sessions when keep-alive is used on
8113 the client side.
8114
8115 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8116 integer parameter <period> which indicates in milliseconds the length
8117 of the period over which the average is measured. It reports the average
8118 HTTP request rate over that period, in requests per period. The result is
8119 an integer which can be matched using ACLs. It does not matter whether
8120 they are valid requests or not. Note that this is different from sessions
8121 when keep-alive is used on the client side.
8122
8123 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
8124 counts the absolute number of HTTP requests errors induced by clients
8125 which matched this entry. Errors are counted on invalid and truncated
8126 requests, as well as on denied or tarpitted requests, and on failed
8127 authentications. If the server responds with 4xx, then the request is
8128 also counted as an error since it's an error triggered by the client
8129 (eg: vulnerability scan).
8130
8131 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8132 integer parameter <period> which indicates in milliseconds the length
8133 of the period over which the average is measured. It reports the average
8134 HTTP request error rate over that period, in requests per period (see
8135 http_err_cnt above for what is accounted as an error). The result is an
8136 integer which can be matched using ACLs.
8137
8138 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
8139 integer which counts the cumulated amount of bytes received from clients
8140 which matched this entry. Headers are included in the count. This may be
8141 used to limit abuse of upload features on photo or video servers.
8142
8143 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8144 integer parameter <period> which indicates in milliseconds the length
8145 of the period over which the average is measured. It reports the average
8146 incoming bytes rate over that period, in bytes per period. It may be used
8147 to detect users which upload too much and too fast. Warning: with large
8148 uploads, it is possible that the amount of uploaded data will be counted
8149 once upon termination, thus causing spikes in the average transfer speed
8150 instead of having a smooth one. This may partially be smoothed with
8151 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
8152 recommended for better fairness.
8153
8154 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
8155 integer which counts the cumulated amount of bytes sent to clients which
8156 matched this entry. Headers are included in the count. This may be used
8157 to limit abuse of bots sucking the whole site.
8158
8159 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
8160 an integer parameter <period> which indicates in milliseconds the length
8161 of the period over which the average is measured. It reports the average
8162 outgoing bytes rate over that period, in bytes per period. It may be used
8163 to detect users which download too much and too fast. Warning: with large
8164 transfers, it is possible that the amount of transferred data will be
8165 counted once upon termination, thus causing spikes in the average
8166 transfer speed instead of having a smooth one. This may partially be
8167 smoothed with "option contstats" though this is not perfect yet. Use of
8168 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +02008169
Willy Tarreauc00cdc22010-06-06 16:48:26 +02008170 There is only one stick-table per proxy. At the moment of writing this doc,
8171 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008172 to be required, simply create a dummy backend with a stick-table in it and
8173 reference it.
8174
8175 It is important to understand that stickiness based on learning information
8176 has some limitations, including the fact that all learned associations are
8177 lost upon restart. In general it can be good as a complement but not always
8178 as an exclusive stickiness.
8179
Willy Tarreauc9705a12010-07-27 20:05:50 +02008180 Last, memory requirements may be important when storing many data types.
8181 Indeed, storing all indicators above at once in each entry requires 116 bytes
8182 per entry, or 116 MB for a 1-million entries table. This is definitely not
8183 something that can be ignored.
8184
8185 Example:
8186 # Keep track of counters of up to 1 million IP addresses over 5 minutes
8187 # and store a general purpose counter and the average connection rate
8188 # computed over a sliding window of 30 seconds.
8189 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
8190
8191 See also : "stick match", "stick on", "stick store-request", section 2.2
David du Colombiera13d1b92011-03-17 10:40:22 +01008192 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008193
8194
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008195stick store-response <pattern> [table <table>] [{if | unless} <condition>]
8196 Define a request pattern used to create an entry in a stickiness table
8197 May be used in sections : defaults | frontend | listen | backend
8198 no | no | yes | yes
8199
8200 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008201 <pattern> is a sample expression rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008202 describes what elements of the response or connection will
8203 be analysed, extracted and stored in the table once a
8204 server is selected.
8205
8206 <table> is an optional stickiness table name. If unspecified, the same
8207 backend's table is used. A stickiness table is declared using
8208 the "stick-table" statement.
8209
8210 <cond> is an optional storage condition. It makes it possible to store
8211 certain criteria only when some conditions are met (or not met).
8212 For instance, it could be used to store the SSL session ID only
8213 when the response is a SSL server hello.
8214
8215 Some protocols or applications require complex stickiness rules and cannot
8216 always simply rely on cookies nor hashing. The "stick store-response"
8217 statement describes a rule to decide what to extract from the response and
8218 when to do it, in order to store it into a stickiness table for further
8219 requests to match it using the "stick match" statement. Obviously the
8220 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008221 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008222 See section 7 for a complete list of possible patterns and transformation
8223 rules.
8224
8225 The table has to be declared using the "stick-table" statement. It must be of
8226 a type compatible with the pattern. By default it is the one which is present
8227 in the same backend. It is possible to share a table with other backends by
8228 referencing it using the "table" keyword. If another table is referenced,
8229 the server's ID inside the backends are used. By default, all server IDs
8230 start at 1 in each backend, so the server ordering is enough. But in case of
8231 doubt, it is highly recommended to force server IDs using their "id" setting.
8232
8233 It is possible to restrict the conditions where a "stick store-response"
8234 statement will apply, using "if" or "unless" followed by a condition. This
8235 condition will be evaluated while parsing the response, so any criteria can
8236 be used. See section 7 for ACL based conditions.
8237
8238 There is no limit on the number of "stick store-response" statements, but
8239 there is a limit of 8 simultaneous stores per request or response. This
8240 makes it possible to store up to 8 criteria, all extracted from either the
8241 request or the response, regardless of the number of rules. Only the 8 first
8242 ones which match will be kept. Using this, it is possible to feed multiple
8243 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01008244 another protocol or access method. Using multiple store-response rules with
8245 the same table is possible and may be used to find the best criterion to rely
8246 on, by arranging the rules by decreasing preference order. Only the first
8247 extracted criterion for a given table will be stored. All subsequent store-
8248 response rules referencing the same table will be skipped and their ACLs will
8249 not be evaluated. However, even if a store-request rule references a table, a
8250 store-response rule may also use the same table. This means that each table
8251 may learn exactly one element from the request and one element from the
8252 response at once.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008253
8254 The table will contain the real server that processed the request.
8255
8256 Example :
8257 # Learn SSL session ID from both request and response and create affinity.
8258 backend https
8259 mode tcp
8260 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +02008261 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008262 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008263
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008264 acl clienthello req_ssl_hello_type 1
8265 acl serverhello rep_ssl_hello_type 2
8266
8267 # use tcp content accepts to detects ssl client and server hello.
8268 tcp-request inspect-delay 5s
8269 tcp-request content accept if clienthello
8270
8271 # no timeout on response inspect delay by default.
8272 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008273
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008274 # SSL session ID (SSLID) may be present on a client or server hello.
8275 # Its length is coded on 1 byte at offset 43 and its value starts
8276 # at offset 44.
8277
8278 # Match and learn on request if client hello.
8279 stick on payload_lv(43,1) if clienthello
8280
8281 # Learn on response if server hello.
8282 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +02008283
Emeric Brun6a1cefa2010-09-24 18:15:17 +02008284 server s1 192.168.1.1:443
8285 server s2 192.168.1.1:443
8286
8287 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
8288 extraction.
8289
8290
Willy Tarreau938c7fe2014-04-25 14:21:39 +02008291tcp-check connect [params*]
8292 Opens a new connection
8293 May be used in sections: defaults | frontend | listen | backend
8294 no | no | yes | yes
8295
8296 When an application lies on more than a single TCP port or when HAProxy
8297 load-balance many services in a single backend, it makes sense to probe all
8298 the services individually before considering a server as operational.
8299
8300 When there are no TCP port configured on the server line neither server port
8301 directive, then the 'tcp-check connect port <port>' must be the first step
8302 of the sequence.
8303
8304 In a tcp-check ruleset a 'connect' is required, it is also mandatory to start
8305 the ruleset with a 'connect' rule. Purpose is to ensure admin know what they
8306 do.
8307
8308 Parameters :
8309 They are optional and can be used to describe how HAProxy should open and
8310 use the TCP connection.
8311
8312 port if not set, check port or server port is used.
8313 It tells HAProxy where to open the connection to.
8314 <port> must be a valid TCP port source integer, from 1 to 65535.
8315
8316 send-proxy send a PROXY protocol string
8317
8318 ssl opens a ciphered connection
8319
8320 Examples:
8321 # check HTTP and HTTPs services on a server.
8322 # first open port 80 thanks to server line port directive, then
8323 # tcp-check opens port 443, ciphered and run a request on it:
8324 option tcp-check
8325 tcp-check connect
8326 tcp-check send GET\ /\ HTTP/1.0\r\n
8327 tcp-check send Host:\ haproxy.1wt.eu\r\n
8328 tcp-check send \r\n
8329 tcp-check expect rstring (2..|3..)
8330 tcp-check connect port 443 ssl
8331 tcp-check send GET\ /\ HTTP/1.0\r\n
8332 tcp-check send Host:\ haproxy.1wt.eu\r\n
8333 tcp-check send \r\n
8334 tcp-check expect rstring (2..|3..)
8335 server www 10.0.0.1 check port 80
8336
8337 # check both POP and IMAP from a single server:
8338 option tcp-check
8339 tcp-check connect port 110
8340 tcp-check expect string +OK\ POP3\ ready
8341 tcp-check connect port 143
8342 tcp-check expect string *\ OK\ IMAP4\ ready
8343 server mail 10.0.0.1 check
8344
8345 See also : "option tcp-check", "tcp-check send", "tcp-check expect"
8346
8347
8348tcp-check expect [!] <match> <pattern>
8349 Specify data to be collected and analysed during a generic health check
8350 May be used in sections: defaults | frontend | listen | backend
8351 no | no | yes | yes
8352
8353 Arguments :
8354 <match> is a keyword indicating how to look for a specific pattern in the
8355 response. The keyword may be one of "string", "rstring" or
8356 binary.
8357 The keyword may be preceded by an exclamation mark ("!") to negate
8358 the match. Spaces are allowed between the exclamation mark and the
8359 keyword. See below for more details on the supported keywords.
8360
8361 <pattern> is the pattern to look for. It may be a string or a regular
8362 expression. If the pattern contains spaces, they must be escaped
8363 with the usual backslash ('\').
8364 If the match is set to binary, then the pattern must be passed as
8365 a serie of hexadecimal digits in an even number. Each sequence of
8366 two digits will represent a byte. The hexadecimal digits may be
8367 used upper or lower case.
8368
8369
8370 The available matches are intentionally similar to their http-check cousins :
8371
8372 string <string> : test the exact string matches in the response buffer.
8373 A health check response will be considered valid if the
8374 response's buffer contains this exact string. If the
8375 "string" keyword is prefixed with "!", then the response
8376 will be considered invalid if the body contains this
8377 string. This can be used to look for a mandatory pattern
8378 in a protocol response, or to detect a failure when a
8379 specific error appears in a protocol banner.
8380
8381 rstring <regex> : test a regular expression on the response buffer.
8382 A health check response will be considered valid if the
8383 response's buffer matches this expression. If the
8384 "rstring" keyword is prefixed with "!", then the response
8385 will be considered invalid if the body matches the
8386 expression.
8387
8388 binary <hexstring> : test the exact string in its hexadecimal form matches
8389 in the response buffer. A health check response will
8390 be considered valid if the response's buffer contains
8391 this exact hexadecimal string.
8392 Purpose is to match data on binary protocols.
8393
8394 It is important to note that the responses will be limited to a certain size
8395 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
8396 Thus, too large responses may not contain the mandatory pattern when using
8397 "string", "rstring" or binary. If a large response is absolutely required, it
8398 is possible to change the default max size by setting the global variable.
8399 However, it is worth keeping in mind that parsing very large responses can
8400 waste some CPU cycles, especially when regular expressions are used, and that
8401 it is always better to focus the checks on smaller resources. Also, in its
8402 current state, the check will not find any string nor regex past a null
8403 character in the response. Similarly it is not possible to request matching
8404 the null character.
8405
8406 Examples :
8407 # perform a POP check
8408 option tcp-check
8409 tcp-check expect string +OK\ POP3\ ready
8410
8411 # perform an IMAP check
8412 option tcp-check
8413 tcp-check expect string *\ OK\ IMAP4\ ready
8414
8415 # look for the redis master server
8416 option tcp-check
8417 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02008418 tcp-check expect string +PONG
Willy Tarreau938c7fe2014-04-25 14:21:39 +02008419 tcp-check send info\ replication\r\n
8420 tcp-check expect string role:master
8421 tcp-check send QUIT\r\n
8422 tcp-check expect string +OK
8423
8424
8425 See also : "option tcp-check", "tcp-check connect", "tcp-check send",
8426 "tcp-check send-binary", "http-check expect", tune.chksize
8427
8428
8429tcp-check send <data>
8430 Specify a string to be sent as a question during a generic health check
8431 May be used in sections: defaults | frontend | listen | backend
8432 no | no | yes | yes
8433
8434 <data> : the data to be sent as a question during a generic health check
8435 session. For now, <data> must be a string.
8436
8437 Examples :
8438 # look for the redis master server
8439 option tcp-check
8440 tcp-check send info\ replication\r\n
8441 tcp-check expect string role:master
8442
8443 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
8444 "tcp-check send-binary", tune.chksize
8445
8446
8447tcp-check send-binary <hexastring>
8448 Specify an hexa digits string to be sent as a binary question during a raw
8449 tcp health check
8450 May be used in sections: defaults | frontend | listen | backend
8451 no | no | yes | yes
8452
8453 <data> : the data to be sent as a question during a generic health check
8454 session. For now, <data> must be a string.
8455 <hexastring> : test the exact string in its hexadecimal form matches in the
8456 response buffer. A health check response will be considered
8457 valid if the response's buffer contains this exact
8458 hexadecimal string.
8459 Purpose is to send binary data to ask on binary protocols.
8460
8461 Examples :
8462 # redis check in binary
8463 option tcp-check
8464 tcp-check send-binary 50494e470d0a # PING\r\n
8465 tcp-check expect binary 2b504F4e47 # +PONG
8466
8467
8468 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
8469 "tcp-check send", tune.chksize
8470
8471
Willy Tarreaue9656522010-08-17 15:40:09 +02008472tcp-request connection <action> [{if | unless} <condition>]
8473 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +02008474 May be used in sections : defaults | frontend | listen | backend
8475 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +02008476 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02008477 <action> defines the action to perform if the condition applies. See
8478 below.
Willy Tarreau1a687942010-05-23 22:40:30 +02008479
Willy Tarreaue9656522010-08-17 15:40:09 +02008480 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008481
8482 Immediately after acceptance of a new incoming connection, it is possible to
8483 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +02008484 or dropped or have its counters tracked. Those conditions cannot make use of
8485 any data contents because the connection has not been read from yet, and the
8486 buffers are not yet allocated. This is used to selectively and very quickly
8487 accept or drop connections from various sources with a very low overhead. If
8488 some contents need to be inspected in order to take the decision, the
8489 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008490
Willy Tarreaue9656522010-08-17 15:40:09 +02008491 The "tcp-request connection" rules are evaluated in their exact declaration
8492 order. If no rule matches or if there is no rule, the default action is to
8493 accept the incoming connection. There is no specific limit to the number of
8494 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008495
Willy Tarreaua9083d02015-05-08 15:27:59 +02008496 Four types of actions are supported :
Willy Tarreaue9656522010-08-17 15:40:09 +02008497 - accept :
8498 accepts the connection if the condition is true (when used with "if")
8499 or false (when used with "unless"). The first such rule executed ends
8500 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008501
Willy Tarreaue9656522010-08-17 15:40:09 +02008502 - reject :
8503 rejects the connection if the condition is true (when used with "if")
8504 or false (when used with "unless"). The first such rule executed ends
8505 the rules evaluation. Rejected connections do not even become a
8506 session, which is why they are accounted separately for in the stats,
8507 as "denied connections". They are not considered for the session
8508 rate-limit and are not logged either. The reason is that these rules
8509 should only be used to filter extremely high connection rates such as
8510 the ones encountered during a massive DDoS attack. Under these extreme
8511 conditions, the simple action of logging each event would make the
8512 system collapse and would considerably lower the filtering capacity. If
8513 logging is absolutely desired, then "tcp-request content" rules should
8514 be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008515
Willy Tarreau4f0d9192013-06-11 20:40:55 +02008516 - expect-proxy layer4 :
8517 configures the client-facing connection to receive a PROXY protocol
8518 header before any byte is read from the socket. This is equivalent to
8519 having the "accept-proxy" keyword on the "bind" line, except that using
8520 the TCP rule allows the PROXY protocol to be accepted only for certain
8521 IP address ranges using an ACL. This is convenient when multiple layers
8522 of load balancers are passed through by traffic coming from public
8523 hosts.
8524
Willy Tarreau18bf01e2014-06-13 16:18:52 +02008525 - capture <sample> len <length> :
8526 This only applies to "tcp-request content" rules. It captures sample
8527 expression <sample> from the request buffer, and converts it to a
8528 string of at most <len> characters. The resulting string is stored into
8529 the next request "capture" slot, so it will possibly appear next to
8530 some captured HTTP headers. It will then automatically appear in the
8531 logs, and it will be possible to extract it using sample fetch rules to
8532 feed it into headers or anything. The length should be limited given
8533 that this size will be allocated for each capture during the whole
Willy Tarreaua9083d02015-05-08 15:27:59 +02008534 session life. Please check section 7.3 (Fetching samples) and "capture
8535 request header" for more information.
Willy Tarreau18bf01e2014-06-13 16:18:52 +02008536
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008537 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +02008538 enables tracking of sticky counters from current connection. These
Willy Tarreau09448f72014-06-25 18:12:15 +02008539 rules do not stop evaluation and do not change default action. 3 sets
Willy Tarreaue9656522010-08-17 15:40:09 +02008540 of counters may be simultaneously tracked by the same connection. The
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008541 first "track-sc0" rule executed enables tracking of the counters of the
8542 specified table as the first set. The first "track-sc1" rule executed
Willy Tarreaue9656522010-08-17 15:40:09 +02008543 enables tracking of the counters of the specified table as the second
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008544 set. The first "track-sc2" rule executed enables tracking of the
8545 counters of the specified table as the third set. It is a recommended
8546 practice to use the first set of counters for the per-frontend counters
8547 and the second set for the per-backend ones. But this is just a
8548 guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008549
Willy Tarreaue9656522010-08-17 15:40:09 +02008550 These actions take one or two arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008551 <key> is mandatory, and is a sample expression rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +02008552 in section 7.3. It describes what elements of the incoming
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008553 request or connection will be analysed, extracted, combined,
8554 and used to select which table entry to update the counters.
8555 Note that "tcp-request connection" cannot use content-based
8556 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008557
Willy Tarreaue9656522010-08-17 15:40:09 +02008558 <table> is an optional table to be used instead of the default one,
8559 which is the stick-table declared in the current proxy. All
8560 the counters for the matches and updates for the key will
8561 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008562
Willy Tarreaue9656522010-08-17 15:40:09 +02008563 Once a "track-sc*" rule is executed, the key is looked up in the table
8564 and if it is not found, an entry is allocated for it. Then a pointer to
8565 that entry is kept during all the session's life, and this entry's
8566 counters are updated as often as possible, every time the session's
8567 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008568 Counters are only updated for events that happen after the tracking has
8569 been started. For example, connection counters will not be updated when
8570 tracking layer 7 information, since the connection event happens before
8571 layer7 information is extracted.
8572
Willy Tarreaue9656522010-08-17 15:40:09 +02008573 If the entry tracks concurrent connection counters, one connection is
8574 counted for as long as the entry is tracked, and the entry will not
8575 expire during that time. Tracking counters also provides a performance
8576 advantage over just checking the keys, because only one table lookup is
8577 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008578
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02008579 - sc-inc-gpc0(<sc-id>):
8580 The "sc-inc-gpc0" increments the GPC0 counter according to the sticky
8581 counter designated by <sc-id>. If an error occurs, this action silently
8582 fails and the actions evaluation continues.
8583
Thierry FOURNIER236657b2015-08-19 08:25:14 +02008584 - sc-set-gpt0(<sc-id>) <int>:
8585 This action sets the GPT0 tag according to the sticky counter designated
8586 by <sc-id> and the value of <int>. The expected result is a boolean. If
8587 an error occurs, this action silently fails and the actions evaluation
8588 continues.
8589
Willy Tarreau2d392c22015-08-24 01:43:45 +02008590 - "silent-drop" :
8591 This stops the evaluation of the rules and makes the client-facing
8592 connection suddenly disappear using a system-dependant way that tries
8593 to prevent the client from being notified. The effect it then that the
8594 client still sees an established connection while there's none on
8595 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
8596 except that it doesn't use any local resource at all on the machine
8597 running HAProxy. It can resist much higher loads than "tarpit", and
8598 slow down stronger attackers. It is important to undestand the impact
8599 of using this mechanism. All stateful equipments placed between the
8600 client and HAProxy (firewalls, proxies, load balancers) will also keep
8601 the established connection for a long time and may suffer from this
8602 action. On modern Linux systems running with enough privileges, the
8603 TCP_REPAIR socket option is used to block the emission of a TCP
8604 reset. On other systems, the socket's TTL is reduced to 1 so that the
8605 TCP reset doesn't pass the first router, though it's still delivered to
8606 local networks. Do not use it unless you fully understand how it works.
8607
Willy Tarreaue9656522010-08-17 15:40:09 +02008608 Note that the "if/unless" condition is optional. If no condition is set on
8609 the action, it is simply performed unconditionally. That can be useful for
8610 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008611
Willy Tarreaue9656522010-08-17 15:40:09 +02008612 Example: accept all connections from white-listed hosts, reject too fast
8613 connection without counting them, and track accepted connections.
8614 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008615
Willy Tarreaue9656522010-08-17 15:40:09 +02008616 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008617 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008618 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008619
Willy Tarreaue9656522010-08-17 15:40:09 +02008620 Example: accept all connections from white-listed hosts, count all other
8621 connections and reject too fast ones. This results in abusive ones
8622 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008623
Willy Tarreaue9656522010-08-17 15:40:09 +02008624 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008625 tcp-request connection track-sc0 src
8626 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008627
Willy Tarreau4f0d9192013-06-11 20:40:55 +02008628 Example: enable the PROXY protocol for traffic coming from all known proxies.
8629
8630 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
8631
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008632 See section 7 about ACL usage.
8633
Willy Tarreaue9656522010-08-17 15:40:09 +02008634 See also : "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008635
8636
Willy Tarreaue9656522010-08-17 15:40:09 +02008637tcp-request content <action> [{if | unless} <condition>]
8638 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +02008639 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02008640 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +02008641 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02008642 <action> defines the action to perform if the condition applies. See
8643 below.
Willy Tarreau62644772008-07-16 18:36:06 +02008644
Willy Tarreaue9656522010-08-17 15:40:09 +02008645 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +02008646
Willy Tarreaue9656522010-08-17 15:40:09 +02008647 A request's contents can be analysed at an early stage of request processing
8648 called "TCP content inspection". During this stage, ACL-based rules are
8649 evaluated every time the request contents are updated, until either an
8650 "accept" or a "reject" rule matches, or the TCP request inspection delay
8651 expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +02008652
Willy Tarreaue9656522010-08-17 15:40:09 +02008653 The first difference between these rules and "tcp-request connection" rules
8654 is that "tcp-request content" rules can make use of contents to take a
8655 decision. Most often, these decisions will consider a protocol recognition or
8656 validity. The second difference is that content-based rules can be used in
Willy Tarreauf3338342014-01-28 21:40:28 +01008657 both frontends and backends. In case of HTTP keep-alive with the client, all
8658 tcp-request content rules are evaluated again, so haproxy keeps a record of
8659 what sticky counters were assigned by a "tcp-request connection" versus a
8660 "tcp-request content" rule, and flushes all the content-related ones after
8661 processing an HTTP request, so that they may be evaluated again by the rules
8662 being evaluated again for the next request. This is of particular importance
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008663 when the rule tracks some L7 information or when it is conditioned by an
Willy Tarreauf3338342014-01-28 21:40:28 +01008664 L7-based ACL, since tracking may change between requests.
Willy Tarreau62644772008-07-16 18:36:06 +02008665
Willy Tarreaue9656522010-08-17 15:40:09 +02008666 Content-based rules are evaluated in their exact declaration order. If no
8667 rule matches or if there is no rule, the default action is to accept the
8668 contents. There is no specific limit to the number of rules which may be
8669 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +02008670
Thierry FOURNIER236657b2015-08-19 08:25:14 +02008671 Several types of actions are supported :
Willy Tarreau18bf01e2014-06-13 16:18:52 +02008672 - accept : the request is accepted
8673 - reject : the request is rejected and the connection is closed
8674 - capture : the specified sample expression is captured
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008675 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02008676 - sc-inc-gpc0(<sc-id>)
Thierry FOURNIER236657b2015-08-19 08:25:14 +02008677 - set-gpt0(<sc-id>) <int>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02008678 - set-var(<var-name>) <expr>
Willy Tarreau2d392c22015-08-24 01:43:45 +02008679 - silent-drop
Willy Tarreau62644772008-07-16 18:36:06 +02008680
Willy Tarreaue9656522010-08-17 15:40:09 +02008681 They have the same meaning as their counter-parts in "tcp-request connection"
8682 so please refer to that section for a complete description.
Willy Tarreau62644772008-07-16 18:36:06 +02008683
Willy Tarreauf3338342014-01-28 21:40:28 +01008684 While there is nothing mandatory about it, it is recommended to use the
8685 track-sc0 in "tcp-request connection" rules, track-sc1 for "tcp-request
8686 content" rules in the frontend, and track-sc2 for "tcp-request content"
8687 rules in the backend, because that makes the configuration more readable
8688 and easier to troubleshoot, but this is just a guideline and all counters
8689 may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +02008690
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008691 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +02008692 the action, it is simply performed unconditionally. That can be useful for
8693 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +02008694
Willy Tarreaue9656522010-08-17 15:40:09 +02008695 It is perfectly possible to match layer 7 contents with "tcp-request content"
Willy Tarreauc0239e02012-04-16 14:42:55 +02008696 rules, since HTTP-specific ACL matches are able to preliminarily parse the
8697 contents of a buffer before extracting the required data. If the buffered
8698 contents do not parse as a valid HTTP message, then the ACL does not match.
8699 The parser which is involved there is exactly the same as for all other HTTP
Willy Tarreauf3338342014-01-28 21:40:28 +01008700 processing, so there is no risk of parsing something differently. In an HTTP
8701 backend connected to from an HTTP frontend, it is guaranteed that HTTP
8702 contents will always be immediately present when the rule is evaluated first.
Willy Tarreau62644772008-07-16 18:36:06 +02008703
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008704 Tracking layer7 information is also possible provided that the information
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02008705 are present when the rule is processed. The rule processing engine is able to
8706 wait until the inspect delay expires when the data to be tracked is not yet
8707 available.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008708
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02008709 The "set-var" is used to set the content of a variable. The variable is
8710 declared inline.
8711
8712 <var-name> The name of the variable starts by an indication about its scope.
8713 The allowed scopes are:
8714 "sess" : the variable is shared with all the session,
8715 "txn" : the variable is shared with all the transaction
8716 (request and response)
8717 "req" : the variable is shared only during the request
8718 processing
8719 "res" : the variable is shared only during the response
8720 processing.
8721 This prefix is followed by a name. The separator is a '.'.
8722 The name may only contain characters 'a-z', 'A-Z', '0-9' and '_'.
8723
8724 <expr> Is a standard HAProxy expression formed by a sample-fetch
8725 followed by some converters.
8726
8727 Example:
8728
8729 tcp-request content set-var(sess.my_var) src
8730
Willy Tarreau62644772008-07-16 18:36:06 +02008731 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +02008732 # Accept HTTP requests containing a Host header saying "example.com"
8733 # and reject everything else.
8734 acl is_host_com hdr(Host) -i example.com
8735 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +02008736 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +02008737 tcp-request content reject
8738
8739 Example:
Willy Tarreau62644772008-07-16 18:36:06 +02008740 # reject SMTP connection if client speaks first
8741 tcp-request inspect-delay 30s
8742 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008743 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +02008744
8745 # Forward HTTPS connection only if client speaks
8746 tcp-request inspect-delay 30s
8747 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02008748 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +02008749 tcp-request content reject
8750
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008751 Example:
8752 # Track the last IP from X-Forwarded-For
8753 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02008754 tcp-request content track-sc0 hdr(x-forwarded-for,-1)
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008755
8756 Example:
8757 # track request counts per "base" (concatenation of Host+URL)
8758 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02008759 tcp-request content track-sc0 base table req-rate
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01008760
Willy Tarreaue9656522010-08-17 15:40:09 +02008761 Example: track per-frontend and per-backend counters, block abusers at the
8762 frontend when the backend detects abuse.
8763
8764 frontend http
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008765 # Use General Purpose Couter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +02008766 # protecting all our sites
8767 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008768 tcp-request connection track-sc0 src
8769 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +02008770 ...
8771 use_backend http_dynamic if { path_end .php }
8772
8773 backend http_dynamic
8774 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008775 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +02008776 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02008777 acl click_too_fast sc1_http_req_rate gt 10
8778 acl mark_as_abuser sc0_inc_gpc0 gt 0
8779 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +02008780 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +02008781
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008782 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +02008783
Willy Tarreaue9656522010-08-17 15:40:09 +02008784 See also : "tcp-request connection", "tcp-request inspect-delay"
Willy Tarreau62644772008-07-16 18:36:06 +02008785
8786
8787tcp-request inspect-delay <timeout>
8788 Set the maximum allowed time to wait for data during content inspection
8789 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02008790 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +02008791 Arguments :
8792 <timeout> is the timeout value specified in milliseconds by default, but
8793 can be in any other unit if the number is suffixed by the unit,
8794 as explained at the top of this document.
8795
8796 People using haproxy primarily as a TCP relay are often worried about the
8797 risk of passing any type of protocol to a server without any analysis. In
8798 order to be able to analyze the request contents, we must first withhold
8799 the data then analyze them. This statement simply enables withholding of
8800 data for at most the specified amount of time.
8801
Willy Tarreaufb356202010-08-03 14:02:05 +02008802 TCP content inspection applies very early when a connection reaches a
8803 frontend, then very early when the connection is forwarded to a backend. This
8804 means that a connection may experience a first delay in the frontend and a
8805 second delay in the backend if both have tcp-request rules.
8806
Willy Tarreau62644772008-07-16 18:36:06 +02008807 Note that when performing content inspection, haproxy will evaluate the whole
8808 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008809 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +02008810 a last check is performed upon expiration, this time considering that the
Willy Tarreaud869b242009-03-15 14:43:58 +01008811 contents are definitive. If no delay is set, haproxy will not wait at all
8812 and will immediately apply a verdict based on the available information.
8813 Obviously this is unlikely to be very useful and might even be racy, so such
8814 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +02008815
8816 As soon as a rule matches, the request is released and continues as usual. If
8817 the timeout is reached and no rule matches, the default policy will be to let
8818 it pass through unaffected.
8819
8820 For most protocols, it is enough to set it to a few seconds, as most clients
8821 send the full request immediately upon connection. Add 3 or more seconds to
8822 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +01008823 to use large values, for instance to ensure that the client never talks
Willy Tarreau62644772008-07-16 18:36:06 +02008824 before the server (eg: SMTP), or to wait for a client to talk before passing
8825 data to the server (eg: SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +02008826 least the inspection delay, otherwise it will expire first. If the client
8827 closes the connection or if the buffer is full, the delay immediately expires
8828 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +02008829
Willy Tarreau55165fe2009-05-10 12:02:55 +02008830 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +02008831 "timeout client".
8832
8833
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008834tcp-response content <action> [{if | unless} <condition>]
8835 Perform an action on a session response depending on a layer 4-7 condition
8836 May be used in sections : defaults | frontend | listen | backend
8837 no | no | yes | yes
8838 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02008839 <action> defines the action to perform if the condition applies. See
8840 below.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008841
8842 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
8843
8844 Response contents can be analysed at an early stage of response processing
8845 called "TCP content inspection". During this stage, ACL-based rules are
8846 evaluated every time the response contents are updated, until either an
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02008847 "accept", "close" or a "reject" rule matches, or a TCP response inspection
8848 delay is set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008849
8850 Most often, these decisions will consider a protocol recognition or validity.
8851
8852 Content-based rules are evaluated in their exact declaration order. If no
8853 rule matches or if there is no rule, the default action is to accept the
8854 contents. There is no specific limit to the number of rules which may be
8855 inserted.
8856
Thierry FOURNIER236657b2015-08-19 08:25:14 +02008857 Several types of actions are supported :
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008858 - accept :
8859 accepts the response if the condition is true (when used with "if")
8860 or false (when used with "unless"). The first such rule executed ends
8861 the rules evaluation.
8862
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02008863 - close :
8864 immediately closes the connection with the server if the condition is
8865 true (when used with "if"), or false (when used with "unless"). The
8866 first such rule executed ends the rules evaluation. The main purpose of
8867 this action is to force a connection to be finished between a client
8868 and a server after an exchange when the application protocol expects
8869 some long time outs to elapse first. The goal is to eliminate idle
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008870 connections which take significant resources on servers with certain
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02008871 protocols.
8872
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008873 - reject :
8874 rejects the response if the condition is true (when used with "if")
8875 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -04008876 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008877
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02008878 - set-var(<var-name>) <expr>
8879 Sets a variable.
8880
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02008881 - sc-inc-gpc0(<sc-id>):
8882 This action increments the GPC0 counter according to the sticky
8883 counter designated by <sc-id>. If an error occurs, this action fails
8884 silently and the actions evaluation continues.
8885
Thierry FOURNIER236657b2015-08-19 08:25:14 +02008886 - sc-set-gpt0(<sc-id>) <int> :
8887 This action sets the GPT0 tag according to the sticky counter designated
8888 by <sc-id> and the value of <int>. The expected result is a boolean. If
8889 an error occurs, this action silently fails and the actions evaluation
8890 continues.
8891
Willy Tarreau2d392c22015-08-24 01:43:45 +02008892 - "silent-drop" :
8893 This stops the evaluation of the rules and makes the client-facing
8894 connection suddenly disappear using a system-dependant way that tries
8895 to prevent the client from being notified. The effect it then that the
8896 client still sees an established connection while there's none on
8897 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
8898 except that it doesn't use any local resource at all on the machine
8899 running HAProxy. It can resist much higher loads than "tarpit", and
8900 slow down stronger attackers. It is important to undestand the impact
8901 of using this mechanism. All stateful equipments placed between the
8902 client and HAProxy (firewalls, proxies, load balancers) will also keep
8903 the established connection for a long time and may suffer from this
8904 action. On modern Linux systems running with enough privileges, the
8905 TCP_REPAIR socket option is used to block the emission of a TCP
8906 reset. On other systems, the socket's TTL is reduced to 1 so that the
8907 TCP reset doesn't pass the first router, though it's still delivered to
8908 local networks. Do not use it unless you fully understand how it works.
8909
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008910 Note that the "if/unless" condition is optional. If no condition is set on
8911 the action, it is simply performed unconditionally. That can be useful for
8912 for changing the default action to a reject.
8913
Jamie Gloudonaaa21002012-08-25 00:18:33 -04008914 It is perfectly possible to match layer 7 contents with "tcp-response
8915 content" rules, but then it is important to ensure that a full response has
8916 been buffered, otherwise no contents will match. In order to achieve this,
8917 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008918 period.
8919
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02008920 The "set-var" is used to set the content of a variable. The variable is
8921 declared inline.
8922
8923 <var-name> The name of the variable starts by an indication about its scope.
8924 The allowed scopes are:
8925 "sess" : the variable is shared with all the session,
8926 "txn" : the variable is shared with all the transaction
8927 (request and response)
8928 "req" : the variable is shared only during the request
8929 processing
8930 "res" : the variable is shared only during the response
8931 processing.
8932 This prefix is followed by a name. The separator is a '.'.
8933 The name may only contain characters 'a-z', 'A-Z', '0-9' and '_'.
8934
8935 <expr> Is a standard HAProxy expression formed by a sample-fetch
8936 followed by some converters.
8937
8938 Example:
8939
8940 tcp-request content set-var(sess.my_var) src
8941
Emeric Brun0a3b67f2010-09-24 15:34:53 +02008942 See section 7 about ACL usage.
8943
8944 See also : "tcp-request content", "tcp-response inspect-delay"
8945
8946
8947tcp-response inspect-delay <timeout>
8948 Set the maximum allowed time to wait for a response during content inspection
8949 May be used in sections : defaults | frontend | listen | backend
8950 no | no | yes | yes
8951 Arguments :
8952 <timeout> is the timeout value specified in milliseconds by default, but
8953 can be in any other unit if the number is suffixed by the unit,
8954 as explained at the top of this document.
8955
8956 See also : "tcp-response content", "tcp-request inspect-delay".
8957
8958
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01008959timeout check <timeout>
8960 Set additional check timeout, but only after a connection has been already
8961 established.
8962
8963 May be used in sections: defaults | frontend | listen | backend
8964 yes | no | yes | yes
8965 Arguments:
8966 <timeout> is the timeout value specified in milliseconds by default, but
8967 can be in any other unit if the number is suffixed by the unit,
8968 as explained at the top of this document.
8969
8970 If set, haproxy uses min("timeout connect", "inter") as a connect timeout
8971 for check and "timeout check" as an additional read timeout. The "min" is
8972 used so that people running with *very* long "timeout connect" (eg. those
8973 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +01008974 (Please also note that there is no valid reason to have such long connect
8975 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
8976 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01008977
8978 If "timeout check" is not set haproxy uses "inter" for complete check
8979 timeout (connect + read) exactly like all <1.3.15 version.
8980
8981 In most cases check request is much simpler and faster to handle than normal
8982 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +01008983 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01008984
8985 This parameter is specific to backends, but can be specified once for all in
8986 "defaults" sections. This is in fact one of the easiest solutions not to
8987 forget about it.
8988
Willy Tarreau41a340d2008-01-22 12:25:31 +01008989 See also: "timeout connect", "timeout queue", "timeout server",
8990 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01008991
8992
Willy Tarreau0ba27502007-12-24 16:55:16 +01008993timeout client <timeout>
8994timeout clitimeout <timeout> (deprecated)
8995 Set the maximum inactivity time on the client side.
8996 May be used in sections : defaults | frontend | listen | backend
8997 yes | yes | yes | no
8998 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01008999 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01009000 can be in any other unit if the number is suffixed by the unit,
9001 as explained at the top of this document.
9002
9003 The inactivity timeout applies when the client is expected to acknowledge or
9004 send data. In HTTP mode, this timeout is particularly important to consider
9005 during the first phase, when the client sends the request, and during the
9006 response while it is reading data sent by the server. The value is specified
9007 in milliseconds by default, but can be in any other unit if the number is
9008 suffixed by the unit, as specified at the top of this document. In TCP mode
9009 (and to a lesser extent, in HTTP mode), it is highly recommended that the
9010 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01009011 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01009012 losses by specifying timeouts that are slightly above multiples of 3 seconds
Willy Tarreauce887fd2012-05-12 12:50:00 +02009013 (eg: 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
9014 sessions (eg: WebSocket and HTTP), it's worth considering "timeout tunnel",
Willy Tarreau05cdd962014-05-10 14:30:07 +02009015 which overrides "timeout client" and "timeout server" for tunnels, as well as
9016 "timeout client-fin" for half-closed connections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01009017
9018 This parameter is specific to frontends, but can be specified once for all in
9019 "defaults" sections. This is in fact one of the easiest solutions not to
9020 forget about it. An unspecified timeout results in an infinite timeout, which
9021 is not recommended. Such a usage is accepted and works but reports a warning
9022 during startup because it may results in accumulation of expired sessions in
9023 the system if the system's timeouts are not configured either.
9024
9025 This parameter replaces the old, deprecated "clitimeout". It is recommended
9026 to use it to write new configurations. The form "timeout clitimeout" is
9027 provided only by backwards compatibility but its use is strongly discouraged.
9028
Willy Tarreauce887fd2012-05-12 12:50:00 +02009029 See also : "clitimeout", "timeout server", "timeout tunnel".
Willy Tarreau0ba27502007-12-24 16:55:16 +01009030
9031
Willy Tarreau05cdd962014-05-10 14:30:07 +02009032timeout client-fin <timeout>
9033 Set the inactivity timeout on the client side for half-closed connections.
9034 May be used in sections : defaults | frontend | listen | backend
9035 yes | yes | yes | no
9036 Arguments :
9037 <timeout> is the timeout value specified in milliseconds by default, but
9038 can be in any other unit if the number is suffixed by the unit,
9039 as explained at the top of this document.
9040
9041 The inactivity timeout applies when the client is expected to acknowledge or
9042 send data while one direction is already shut down. This timeout is different
9043 from "timeout client" in that it only applies to connections which are closed
9044 in one direction. This is particularly useful to avoid keeping connections in
9045 FIN_WAIT state for too long when clients do not disconnect cleanly. This
9046 problem is particularly common long connections such as RDP or WebSocket.
9047 Note that this timeout can override "timeout tunnel" when a connection shuts
9048 down in one direction.
9049
9050 This parameter is specific to frontends, but can be specified once for all in
9051 "defaults" sections. By default it is not set, so half-closed connections
9052 will use the other timeouts (timeout.client or timeout.tunnel).
9053
9054 See also : "timeout client", "timeout server-fin", and "timeout tunnel".
9055
9056
Willy Tarreau0ba27502007-12-24 16:55:16 +01009057timeout connect <timeout>
9058timeout contimeout <timeout> (deprecated)
9059 Set the maximum time to wait for a connection attempt to a server to succeed.
9060 May be used in sections : defaults | frontend | listen | backend
9061 yes | no | yes | yes
9062 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01009063 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01009064 can be in any other unit if the number is suffixed by the unit,
9065 as explained at the top of this document.
9066
9067 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01009068 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01009069 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01009070 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01009071 connect timeout also presets both queue and tarpit timeouts to the same value
9072 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +01009073
9074 This parameter is specific to backends, but can be specified once for all in
9075 "defaults" sections. This is in fact one of the easiest solutions not to
9076 forget about it. An unspecified timeout results in an infinite timeout, which
9077 is not recommended. Such a usage is accepted and works but reports a warning
9078 during startup because it may results in accumulation of failed sessions in
9079 the system if the system's timeouts are not configured either.
9080
9081 This parameter replaces the old, deprecated "contimeout". It is recommended
9082 to use it to write new configurations. The form "timeout contimeout" is
9083 provided only by backwards compatibility but its use is strongly discouraged.
9084
Willy Tarreau41a340d2008-01-22 12:25:31 +01009085 See also: "timeout check", "timeout queue", "timeout server", "contimeout",
9086 "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +01009087
9088
Willy Tarreaub16a5742010-01-10 14:46:16 +01009089timeout http-keep-alive <timeout>
9090 Set the maximum allowed time to wait for a new HTTP request to appear
9091 May be used in sections : defaults | frontend | listen | backend
9092 yes | yes | yes | yes
9093 Arguments :
9094 <timeout> is the timeout value specified in milliseconds by default, but
9095 can be in any other unit if the number is suffixed by the unit,
9096 as explained at the top of this document.
9097
9098 By default, the time to wait for a new request in case of keep-alive is set
9099 by "timeout http-request". However this is not always convenient because some
9100 people want very short keep-alive timeouts in order to release connections
9101 faster, and others prefer to have larger ones but still have short timeouts
9102 once the request has started to present itself.
9103
9104 The "http-keep-alive" timeout covers these needs. It will define how long to
9105 wait for a new HTTP request to start coming after a response was sent. Once
9106 the first byte of request has been seen, the "http-request" timeout is used
9107 to wait for the complete request to come. Note that empty lines prior to a
9108 new request do not refresh the timeout and are not counted as a new request.
9109
9110 There is also another difference between the two timeouts : when a connection
9111 expires during timeout http-keep-alive, no error is returned, the connection
9112 just closes. If the connection expires in "http-request" while waiting for a
9113 connection to complete, a HTTP 408 error is returned.
9114
9115 In general it is optimal to set this value to a few tens to hundreds of
9116 milliseconds, to allow users to fetch all objects of a page at once but
9117 without waiting for further clicks. Also, if set to a very small value (eg:
9118 1 millisecond) it will probably only accept pipelined requests but not the
9119 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +02009120 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +01009121
9122 If this parameter is not set, the "http-request" timeout applies, and if both
9123 are not set, "timeout client" still applies at the lower level. It should be
9124 set in the frontend to take effect, unless the frontend is in TCP mode, in
9125 which case the HTTP backend's timeout will be used.
9126
9127 See also : "timeout http-request", "timeout client".
9128
9129
Willy Tarreau036fae02008-01-06 13:24:40 +01009130timeout http-request <timeout>
9131 Set the maximum allowed time to wait for a complete HTTP request
9132 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +02009133 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +01009134 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01009135 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +01009136 can be in any other unit if the number is suffixed by the unit,
9137 as explained at the top of this document.
9138
9139 In order to offer DoS protection, it may be required to lower the maximum
9140 accepted time to receive a complete HTTP request without affecting the client
9141 timeout. This helps protecting against established connections on which
9142 nothing is sent. The client timeout cannot offer a good protection against
9143 this abuse because it is an inactivity timeout, which means that if the
9144 attacker sends one character every now and then, the timeout will not
9145 trigger. With the HTTP request timeout, no matter what speed the client
Willy Tarreau2705a612014-05-23 17:38:34 +02009146 types, the request will be aborted if it does not complete in time. When the
9147 timeout expires, an HTTP 408 response is sent to the client to inform it
9148 about the problem, and the connection is closed. The logs will report
9149 termination codes "cR". Some recent browsers are having problems with this
9150 standard, well-documented behaviour, so it might be needed to hide the 408
Willy Tarreau0f228a02015-05-01 15:37:53 +02009151 code using "option http-ignore-probes" or "errorfile 408 /dev/null". See
9152 more details in the explanations of the "cR" termination code in section 8.5.
Willy Tarreau036fae02008-01-06 13:24:40 +01009153
Baptiste Assmanneccdf432015-10-28 13:49:01 +01009154 By default, this timeout only applies to the header part of the request,
9155 and not to any data. As soon as the empty line is received, this timeout is
9156 not used anymore. When combined with "option http-buffer-request", this
9157 timeout also applies to the body of the request..
9158 It is used again on keep-alive connections to wait for a second
Willy Tarreaub16a5742010-01-10 14:46:16 +01009159 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +01009160
9161 Generally it is enough to set it to a few seconds, as most clients send the
9162 full request immediately upon connection. Add 3 or more seconds to cover TCP
9163 retransmits but that's all. Setting it to very low values (eg: 50 ms) will
9164 generally work on local networks as long as there are no packet losses. This
9165 will prevent people from sending bare HTTP requests using telnet.
9166
9167 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +02009168 chunk of the incoming request. It should be set in the frontend to take
9169 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
9170 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +01009171
Willy Tarreau0f228a02015-05-01 15:37:53 +02009172 See also : "errorfile", "http-ignore-probes", "timeout http-keep-alive", and
Baptiste Assmanneccdf432015-10-28 13:49:01 +01009173 "timeout client", "option http-buffer-request".
Willy Tarreau036fae02008-01-06 13:24:40 +01009174
Willy Tarreau844e3c52008-01-11 16:28:18 +01009175
9176timeout queue <timeout>
9177 Set the maximum time to wait in the queue for a connection slot to be free
9178 May be used in sections : defaults | frontend | listen | backend
9179 yes | no | yes | yes
9180 Arguments :
9181 <timeout> is the timeout value specified in milliseconds by default, but
9182 can be in any other unit if the number is suffixed by the unit,
9183 as explained at the top of this document.
9184
9185 When a server's maxconn is reached, connections are left pending in a queue
9186 which may be server-specific or global to the backend. In order not to wait
9187 indefinitely, a timeout is applied to requests pending in the queue. If the
9188 timeout is reached, it is considered that the request will almost never be
9189 served, so it is dropped and a 503 error is returned to the client.
9190
9191 The "timeout queue" statement allows to fix the maximum time for a request to
9192 be left pending in a queue. If unspecified, the same value as the backend's
9193 connection timeout ("timeout connect") is used, for backwards compatibility
9194 with older versions with no "timeout queue" parameter.
9195
9196 See also : "timeout connect", "contimeout".
9197
9198
9199timeout server <timeout>
9200timeout srvtimeout <timeout> (deprecated)
9201 Set the maximum inactivity time on the server side.
9202 May be used in sections : defaults | frontend | listen | backend
9203 yes | no | yes | yes
9204 Arguments :
9205 <timeout> is the timeout value specified in milliseconds by default, but
9206 can be in any other unit if the number is suffixed by the unit,
9207 as explained at the top of this document.
9208
9209 The inactivity timeout applies when the server is expected to acknowledge or
9210 send data. In HTTP mode, this timeout is particularly important to consider
9211 during the first phase of the server's response, when it has to send the
9212 headers, as it directly represents the server's processing time for the
9213 request. To find out what value to put there, it's often good to start with
9214 what would be considered as unacceptable response times, then check the logs
9215 to observe the response time distribution, and adjust the value accordingly.
9216
9217 The value is specified in milliseconds by default, but can be in any other
9218 unit if the number is suffixed by the unit, as specified at the top of this
9219 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
9220 recommended that the client timeout remains equal to the server timeout in
9221 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01009222 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01009223 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreauce887fd2012-05-12 12:50:00 +02009224 seconds (eg: 4 or 5 seconds minimum). If some long-lived sessions are mixed
9225 with short-lived sessions (eg: WebSocket and HTTP), it's worth considering
9226 "timeout tunnel", which overrides "timeout client" and "timeout server" for
9227 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009228
9229 This parameter is specific to backends, but can be specified once for all in
9230 "defaults" sections. This is in fact one of the easiest solutions not to
9231 forget about it. An unspecified timeout results in an infinite timeout, which
9232 is not recommended. Such a usage is accepted and works but reports a warning
9233 during startup because it may results in accumulation of expired sessions in
9234 the system if the system's timeouts are not configured either.
9235
9236 This parameter replaces the old, deprecated "srvtimeout". It is recommended
9237 to use it to write new configurations. The form "timeout srvtimeout" is
9238 provided only by backwards compatibility but its use is strongly discouraged.
9239
Willy Tarreauce887fd2012-05-12 12:50:00 +02009240 See also : "srvtimeout", "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +01009241
Willy Tarreau05cdd962014-05-10 14:30:07 +02009242
9243timeout server-fin <timeout>
9244 Set the inactivity timeout on the server side for half-closed connections.
9245 May be used in sections : defaults | frontend | listen | backend
9246 yes | no | yes | yes
9247 Arguments :
9248 <timeout> is the timeout value specified in milliseconds by default, but
9249 can be in any other unit if the number is suffixed by the unit,
9250 as explained at the top of this document.
9251
9252 The inactivity timeout applies when the server is expected to acknowledge or
9253 send data while one direction is already shut down. This timeout is different
9254 from "timeout server" in that it only applies to connections which are closed
9255 in one direction. This is particularly useful to avoid keeping connections in
9256 FIN_WAIT state for too long when a remote server does not disconnect cleanly.
9257 This problem is particularly common long connections such as RDP or WebSocket.
9258 Note that this timeout can override "timeout tunnel" when a connection shuts
9259 down in one direction. This setting was provided for completeness, but in most
9260 situations, it should not be needed.
9261
9262 This parameter is specific to backends, but can be specified once for all in
9263 "defaults" sections. By default it is not set, so half-closed connections
9264 will use the other timeouts (timeout.server or timeout.tunnel).
9265
9266 See also : "timeout client-fin", "timeout server", and "timeout tunnel".
9267
Willy Tarreau844e3c52008-01-11 16:28:18 +01009268
9269timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +01009270 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +01009271 May be used in sections : defaults | frontend | listen | backend
9272 yes | yes | yes | yes
9273 Arguments :
9274 <timeout> is the tarpit duration specified in milliseconds by default, but
9275 can be in any other unit if the number is suffixed by the unit,
9276 as explained at the top of this document.
9277
9278 When a connection is tarpitted using "reqtarpit", it is maintained open with
9279 no activity for a certain amount of time, then closed. "timeout tarpit"
9280 defines how long it will be maintained open.
9281
9282 The value is specified in milliseconds by default, but can be in any other
9283 unit if the number is suffixed by the unit, as specified at the top of this
9284 document. If unspecified, the same value as the backend's connection timeout
9285 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +01009286 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009287
9288 See also : "timeout connect", "contimeout".
9289
9290
Willy Tarreauce887fd2012-05-12 12:50:00 +02009291timeout tunnel <timeout>
9292 Set the maximum inactivity time on the client and server side for tunnels.
9293 May be used in sections : defaults | frontend | listen | backend
9294 yes | no | yes | yes
9295 Arguments :
9296 <timeout> is the timeout value specified in milliseconds by default, but
9297 can be in any other unit if the number is suffixed by the unit,
9298 as explained at the top of this document.
9299
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009300 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +02009301 between a client and a server, and the connection remains inactive in both
9302 directions. This timeout supersedes both the client and server timeouts once
9303 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
9304 analyser remains attached to either connection (eg: tcp content rules are
9305 accepted). In HTTP, this timeout is used when a connection is upgraded (eg:
9306 when switching to the WebSocket protocol, or forwarding a CONNECT request
9307 to a proxy), or after the first response when no keepalive/close option is
9308 specified.
9309
Willy Tarreau05cdd962014-05-10 14:30:07 +02009310 Since this timeout is usually used in conjunction with long-lived connections,
9311 it usually is a good idea to also set "timeout client-fin" to handle the
9312 situation where a client suddenly disappears from the net and does not
9313 acknowledge a close, or sends a shutdown and does not acknowledge pending
9314 data anymore. This can happen in lossy networks where firewalls are present,
9315 and is detected by the presence of large amounts of sessions in a FIN_WAIT
9316 state.
9317
Willy Tarreauce887fd2012-05-12 12:50:00 +02009318 The value is specified in milliseconds by default, but can be in any other
9319 unit if the number is suffixed by the unit, as specified at the top of this
9320 document. Whatever the expected normal idle time, it is a good practice to
9321 cover at least one or several TCP packet losses by specifying timeouts that
9322 are slightly above multiples of 3 seconds (eg: 4 or 5 seconds minimum).
9323
9324 This parameter is specific to backends, but can be specified once for all in
9325 "defaults" sections. This is in fact one of the easiest solutions not to
9326 forget about it.
9327
9328 Example :
9329 defaults http
9330 option http-server-close
9331 timeout connect 5s
9332 timeout client 30s
Willy Tarreau05cdd962014-05-10 14:30:07 +02009333 timeout client-fin 30s
Willy Tarreauce887fd2012-05-12 12:50:00 +02009334 timeout server 30s
9335 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
9336
Willy Tarreau05cdd962014-05-10 14:30:07 +02009337 See also : "timeout client", "timeout client-fin", "timeout server".
Willy Tarreauce887fd2012-05-12 12:50:00 +02009338
9339
Willy Tarreau844e3c52008-01-11 16:28:18 +01009340transparent (deprecated)
9341 Enable client-side transparent proxying
9342 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01009343 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01009344 Arguments : none
9345
9346 This keyword was introduced in order to provide layer 7 persistence to layer
9347 3 load balancers. The idea is to use the OS's ability to redirect an incoming
9348 connection for a remote address to a local process (here HAProxy), and let
9349 this process know what address was initially requested. When this option is
9350 used, sessions without cookies will be forwarded to the original destination
9351 IP address of the incoming request (which should match that of another
9352 equipment), while requests with cookies will still be forwarded to the
9353 appropriate server.
9354
9355 The "transparent" keyword is deprecated, use "option transparent" instead.
9356
9357 Note that contrary to a common belief, this option does NOT make HAProxy
9358 present the client's IP to the server when establishing the connection.
9359
Willy Tarreau844e3c52008-01-11 16:28:18 +01009360 See also: "option transparent"
9361
William Lallemanda73203e2012-03-12 12:48:57 +01009362unique-id-format <string>
9363 Generate a unique ID for each request.
9364 May be used in sections : defaults | frontend | listen | backend
9365 yes | yes | yes | no
9366 Arguments :
9367 <string> is a log-format string.
9368
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009369 This keyword creates a ID for each request using the custom log format. A
9370 unique ID is useful to trace a request passing through many components of
9371 a complex infrastructure. The newly created ID may also be logged using the
9372 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +01009373
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009374 The format should be composed from elements that are guaranteed to be
9375 unique when combined together. For instance, if multiple haproxy instances
9376 are involved, it might be important to include the node name. It is often
9377 needed to log the incoming connection's source and destination addresses
9378 and ports. Note that since multiple requests may be performed over the same
9379 connection, including a request counter may help differentiate them.
9380 Similarly, a timestamp may protect against a rollover of the counter.
9381 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +01009382
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009383 It is recommended to use hexadecimal notation for many fields since it
9384 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +01009385
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009386 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01009387
Julien Vehentf21be322014-03-07 08:27:34 -05009388 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +01009389
9390 will generate:
9391
9392 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
9393
9394 See also: "unique-id-header"
9395
9396unique-id-header <name>
9397 Add a unique ID header in the HTTP request.
9398 May be used in sections : defaults | frontend | listen | backend
9399 yes | yes | yes | no
9400 Arguments :
9401 <name> is the name of the header.
9402
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009403 Add a unique-id header in the HTTP request sent to the server, using the
9404 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +01009405
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009406 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01009407
Julien Vehentf21be322014-03-07 08:27:34 -05009408 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +01009409 unique-id-header X-Unique-ID
9410
9411 will generate:
9412
9413 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
9414
9415 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +01009416
Willy Tarreauf51658d2014-04-23 01:21:56 +02009417use_backend <backend> [{if | unless} <condition>]
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02009418 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009419 May be used in sections : defaults | frontend | listen | backend
9420 no | yes | yes | no
9421 Arguments :
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01009422 <backend> is the name of a valid backend or "listen" section, or a
9423 "log-format" string resolving to a backend name.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009424
Willy Tarreauf51658d2014-04-23 01:21:56 +02009425 <condition> is a condition composed of ACLs, as described in section 7. If
9426 it is omitted, the rule is unconditionally applied.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009427
9428 When doing content-switching, connections arrive on a frontend and are then
9429 dispatched to various backends depending on a number of conditions. The
9430 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02009431 "use_backend" keyword. While it is normally used with HTTP processing, it can
9432 also be used in pure TCP, either without content using stateless ACLs (eg:
9433 source address validation) or combined with a "tcp-request" rule to wait for
9434 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009435
9436 There may be as many "use_backend" rules as desired. All of these rules are
9437 evaluated in their declaration order, and the first one which matches will
9438 assign the backend.
9439
9440 In the first form, the backend will be used if the condition is met. In the
9441 second form, the backend will be used if the condition is not met. If no
9442 condition is valid, the backend defined with "default_backend" will be used.
9443 If no default backend is defined, either the servers in the same section are
9444 used (in case of a "listen" section) or, in case of a frontend, no server is
9445 used and a 503 service unavailable response is returned.
9446
Willy Tarreau51aecc72009-07-12 09:47:04 +02009447 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009448 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +02009449 and backend processing will immediately follow, or the backend will wait for
9450 a complete HTTP request to get in. This feature is useful when a frontend
9451 must decode several protocols on a unique port, one of them being HTTP.
9452
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01009453 When <backend> is a simple name, it is resolved at configuration time, and an
9454 error is reported if the specified backend does not exist. If <backend> is
9455 a log-format string instead, no check may be done at configuration time, so
9456 the backend name is resolved dynamically at run time. If the resulting
9457 backend name does not correspond to any valid backend, no other rule is
9458 evaluated, and the default_backend directive is applied instead. Note that
9459 when using dynamic backend names, it is highly recommended to use a prefix
9460 that no other backend uses in order to ensure that an unauthorized backend
9461 cannot be forced from the request.
9462
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009463 It is worth mentioning that "use_backend" rules with an explicit name are
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01009464 used to detect the association between frontends and backends to compute the
9465 backend's "fullconn" setting. This cannot be done for dynamic names.
9466
9467 See also: "default_backend", "tcp-request", "fullconn", "log-format", and
9468 section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +01009469
Willy Tarreau036fae02008-01-06 13:24:40 +01009470
Willy Tarreau4a5cade2012-04-05 21:09:48 +02009471use-server <server> if <condition>
9472use-server <server> unless <condition>
9473 Only use a specific server if/unless an ACL-based condition is matched.
9474 May be used in sections : defaults | frontend | listen | backend
9475 no | no | yes | yes
9476 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009477 <server> is the name of a valid server in the same backend section.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02009478
9479 <condition> is a condition composed of ACLs, as described in section 7.
9480
9481 By default, connections which arrive to a backend are load-balanced across
9482 the available servers according to the configured algorithm, unless a
9483 persistence mechanism such as a cookie is used and found in the request.
9484
9485 Sometimes it is desirable to forward a particular request to a specific
9486 server without having to declare a dedicated backend for this server. This
9487 can be achieved using the "use-server" rules. These rules are evaluated after
9488 the "redirect" rules and before evaluating cookies, and they have precedence
9489 on them. There may be as many "use-server" rules as desired. All of these
9490 rules are evaluated in their declaration order, and the first one which
9491 matches will assign the server.
9492
9493 If a rule designates a server which is down, and "option persist" is not used
9494 and no force-persist rule was validated, it is ignored and evaluation goes on
9495 with the next rules until one matches.
9496
9497 In the first form, the server will be used if the condition is met. In the
9498 second form, the server will be used if the condition is not met. If no
9499 condition is valid, the processing continues and the server will be assigned
9500 according to other persistence mechanisms.
9501
9502 Note that even if a rule is matched, cookie processing is still performed but
9503 does not assign the server. This allows prefixed cookies to have their prefix
9504 stripped.
9505
9506 The "use-server" statement works both in HTTP and TCP mode. This makes it
9507 suitable for use with content-based inspection. For instance, a server could
9508 be selected in a farm according to the TLS SNI field. And if these servers
9509 have their weight set to zero, they will not be used for other traffic.
9510
9511 Example :
9512 # intercept incoming TLS requests based on the SNI field
9513 use-server www if { req_ssl_sni -i www.example.com }
9514 server www 192.168.0.1:443 weight 0
9515 use-server mail if { req_ssl_sni -i mail.example.com }
9516 server mail 192.168.0.1:587 weight 0
9517 use-server imap if { req_ssl_sni -i imap.example.com }
9518 server mail 192.168.0.1:993 weight 0
9519 # all the rest is forwarded to this server
9520 server default 192.168.0.2:443 check
9521
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009522 See also: "use_backend", section 5 about server and section 7 about ACLs.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02009523
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009524
95255. Bind and Server options
9526--------------------------
9527
9528The "bind", "server" and "default-server" keywords support a number of settings
9529depending on some build options and on the system HAProxy was built on. These
9530settings generally each consist in one word sometimes followed by a value,
9531written on the same line as the "bind" or "server" line. All these options are
9532described in this section.
9533
9534
95355.1. Bind options
9536-----------------
9537
9538The "bind" keyword supports a certain number of settings which are all passed
9539as arguments on the same line. The order in which those arguments appear makes
9540no importance, provided that they appear after the bind address. All of these
9541parameters are optional. Some of them consist in a single words (booleans),
9542while other ones expect a value after them. In this case, the value must be
9543provided immediately after the setting name.
9544
9545The currently supported settings are the following ones.
9546
9547accept-proxy
9548 Enforces the use of the PROXY protocol over any connection accepted by any of
Willy Tarreau77992672014-06-14 11:06:17 +02009549 the sockets declared on the same line. Versions 1 and 2 of the PROXY protocol
9550 are supported and correctly detected. The PROXY protocol dictates the layer
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009551 3/4 addresses of the incoming connection to be used everywhere an address is
9552 used, with the only exception of "tcp-request connection" rules which will
9553 only see the real connection address. Logs will reflect the addresses
9554 indicated in the protocol, unless it is violated, in which case the real
9555 address will still be used. This keyword combined with support from external
9556 components can be used as an efficient and reliable alternative to the
9557 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +02009558 usable. See also "tcp-request connection expect-proxy" for a finer-grained
9559 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009560
Willy Tarreauab861d32013-04-02 02:30:41 +02009561alpn <protocols>
9562 This enables the TLS ALPN extension and advertises the specified protocol
9563 list as supported on top of ALPN. The protocol list consists in a comma-
9564 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
9565 quotes). This requires that the SSL library is build with support for TLS
9566 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
9567 initial NPN extension.
9568
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009569backlog <backlog>
9570 Sets the socket's backlog to this value. If unspecified, the frontend's
9571 backlog is used instead, which generally defaults to the maxconn value.
9572
Emeric Brun7fb34422012-09-28 15:26:15 +02009573ecdhe <named curve>
9574 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +01009575 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
9576 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +02009577
Emeric Brunfd33a262012-10-11 16:28:27 +02009578ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +02009579 This setting is only available when support for OpenSSL was built in. It
9580 designates a PEM file from which to load CA certificates used to verify
9581 client's certificate.
9582
Emeric Brunb6dc9342012-09-28 17:55:37 +02009583ca-ignore-err [all|<errorID>,...]
9584 This setting is only available when support for OpenSSL was built in.
9585 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
9586 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
9587 error is ignored.
9588
Christopher Faulet31af49d2015-06-09 17:29:50 +02009589ca-sign-file <cafile>
9590 This setting is only available when support for OpenSSL was built in. It
9591 designates a PEM file containing both the CA certificate and the CA private
9592 key used to create and sign server's certificates. This is a mandatory
9593 setting when the dynamic generation of certificates is enabled. See
9594 'generate-certificates' for details.
9595
9596ca-sign-passphrase <passphrase>
9597 This setting is only available when support for OpenSSL was built in. It is
9598 the CA private key passphrase. This setting is optional and used only when
9599 the dynamic generation of certificates is enabled. See
9600 'generate-certificates' for details.
9601
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009602ciphers <ciphers>
9603 This setting is only available when support for OpenSSL was built in. It sets
9604 the string describing the list of cipher algorithms ("cipher suite") that are
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009605 negotiated during the SSL/TLS handshake. The format of the string is defined
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009606 in "man 1 ciphers" from OpenSSL man pages, and can be for instance a string
9607 such as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes).
9608
Emeric Brunfd33a262012-10-11 16:28:27 +02009609crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +02009610 This setting is only available when support for OpenSSL was built in. It
9611 designates a PEM file from which to load certificate revocation list used
9612 to verify client's certificate.
9613
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009614crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +00009615 This setting is only available when support for OpenSSL was built in. It
9616 designates a PEM file containing both the required certificates and any
9617 associated private keys. This file can be built by concatenating multiple
9618 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
9619 requires an intermediate certificate, this can also be concatenated into this
9620 file.
9621
9622 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
9623 are loaded.
9624
9625 If a directory name is used instead of a PEM file, then all files found in
Cyril Bonté3180f7b2015-01-25 00:16:08 +01009626 that directory will be loaded in alphabetic order unless their name ends with
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +01009627 '.issuer', '.ocsp' or '.sctl' (reserved extensions). This directive may be
9628 specified multiple times in order to load certificates from multiple files or
9629 directories. The certificates will be presented to clients who provide a
9630 valid TLS Server Name Indication field matching one of their CN or alt
9631 subjects. Wildcards are supported, where a wildcard character '*' is used
9632 instead of the first hostname component (eg: *.example.org matches
9633 www.example.org but not www.sub.example.org).
Alex Davies0fbf0162013-03-02 16:04:50 +00009634
9635 If no SNI is provided by the client or if the SSL library does not support
9636 TLS extensions, or if the client provides an SNI hostname which does not
9637 match any certificate, then the first loaded certificate will be presented.
9638 This means that when loading certificates from a directory, it is highly
Cyril Bonté3180f7b2015-01-25 00:16:08 +01009639 recommended to load the default one first as a file or to ensure that it will
9640 always be the first one in the directory.
Alex Davies0fbf0162013-03-02 16:04:50 +00009641
Emeric Brune032bfa2012-09-28 13:01:45 +02009642 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009643
Alex Davies0fbf0162013-03-02 16:04:50 +00009644 Some CAs (such as Godaddy) offer a drop down list of server types that do not
9645 include HAProxy when obtaining a certificate. If this happens be sure to
Godbach8bf60a12014-04-21 21:42:41 +08009646 choose a webserver that the CA believes requires an intermediate CA (for
Alex Davies0fbf0162013-03-02 16:04:50 +00009647 Godaddy, selection Apache Tomcat will get the correct bundle, but many
9648 others, e.g. nginx, result in a wrong bundle that will not work for some
9649 clients).
9650
Emeric Brun4147b2e2014-06-16 18:36:30 +02009651 For each PEM file, haproxy checks for the presence of file at the same path
9652 suffixed by ".ocsp". If such file is found, support for the TLS Certificate
9653 Status Request extension (also known as "OCSP stapling") is automatically
9654 enabled. The content of this file is optional. If not empty, it must contain
9655 a valid OCSP Response in DER format. In order to be valid an OCSP Response
9656 must comply with the following rules: it has to indicate a good status,
9657 it has to be a single response for the certificate of the PEM file, and it
9658 has to be valid at the moment of addition. If these rules are not respected
9659 the OCSP Response is ignored and a warning is emitted. In order to identify
9660 which certificate an OCSP Response applies to, the issuer's certificate is
9661 necessary. If the issuer's certificate is not found in the PEM file, it will
9662 be loaded from a file at the same path as the PEM file suffixed by ".issuer"
9663 if it exists otherwise it will fail with an error.
9664
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +01009665 For each PEM file, haproxy also checks for the presence of file at the same
9666 path suffixed by ".sctl". If such file is found, support for Certificate
9667 Transparency (RFC6962) TLS extension is enabled. The file must contain a
9668 valid Signed Certificate Timestamp List, as described in RFC. File is parsed
9669 to check basic syntax, but no signatures are verified.
9670
yanbzhud19630c2015-12-14 15:10:25 -05009671 There are cases where it is desirable support multiple key types (RSA/ECDSA)
9672 in the cipher suites offered to the clients. This allows clients that support
9673 EC certificates to be able to use EC ciphers, while simultaneously supporting
9674 older, RSA only clients.
9675
9676 In order to provide this functionality, multiple PEM files, each with a
9677 different key type, are required. To associate these PEM files into a
9678 "cert bundle" that is recognized by haproxy, they must be named in the
9679 following way: All PEM files that are to be bundled must have the same base
9680 name, with a suffix indicating the key type. Currently, three suffixes are
9681 supported: rsa, dsa and ecdsa. For example, if www.example.com has two PEM
9682 files, an RSA file and an ECDSA file, they must be named: "example.pem.rsa"
9683 and "example.pem.ecdsa". The first part of the filename is arbitrary; only the
9684 suffix matters. To load this bundle into haproxy, specify the base name only:
9685
9686 Example : bind :8443 ssl crt example.pem
9687
9688 Note that the suffix is not given to haproxy, this tells haproxy to look for
9689 a cert bundle.
9690
9691 Haproxy will load all PEM files in the bundle at the same time to try to
9692 support multiple key types. PEM files are combined based on Common Name
9693 (CN) and Subject Alternative Name (SAN) to support SNI lookups. This means
9694 that even if you give haproxy a cert bundle, if there are no shared CN/SAN
9695 entries in the certificates in that bundle, haproxy will not be able to
9696 provide multi-cert support.
9697
9698 Assuming bundle in the example above contained the following:
9699
9700 Filename | CN | SAN
9701 -------------------+-----------------+-------------------
9702 example.pem.rsa | www.example.com | rsa.example.com
9703 -------------------+-----------------+--+----------------
9704 example.pem.ecdsa | www.example.com | ecdsa.example.com
9705 -------------------+-----------------+-------------------
9706
9707 Users connecting with an SNI of "www.example.com" will be able
9708 to use both RSA and ECDSA cipher suites. Users connecting with an SNI of
9709 "rsa.example.com" will only be able to use RSA cipher suites, and users
9710 connecting with "ecdsa.example.com" will only be able to use ECDSA cipher
9711 suites.
9712
9713 If a directory name is given as the <cert> argument, haproxy will
9714 automatically search and load bundled files in that directory.
9715
9716 OSCP files (.ocsp) and issuer files (.issuer) are supported with multi-cert
9717 bundling. Each certificate can have its own .ocsp and .issuer file. At this
9718 time, sctl is not supported in multi-certificate bundling.
9719
Emeric Brunb6dc9342012-09-28 17:55:37 +02009720crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +00009721 This setting is only available when support for OpenSSL was built in. Sets a
9722 comma separated list of errorIDs to ignore during verify at depth == 0. If
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009723 set to 'all', all errors are ignored. SSL handshake is not aborted if an error
Alex Davies0fbf0162013-03-02 16:04:50 +00009724 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +02009725
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009726crt-list <file>
9727 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02009728 designates a list of PEM file with an optional list of SNI filter per
9729 certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009730
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02009731 <crtfile> [[!]<snifilter> ...]
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009732
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02009733 Wildcards are supported in the SNI filter. Negative filter are also supported,
9734 only useful in combination with a wildcard filter to exclude a particular SNI.
9735 The certificates will be presented to clients who provide a valid TLS Server
9736 Name Indication field matching one of the SNI filters. If no SNI filter is
9737 specified, the CN and alt subjects are used. This directive may be specified
9738 multiple times. See the "crt" option for more information. The default
9739 certificate is still needed to meet OpenSSL expectations. If it is not used,
9740 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01009741
yanbzhud19630c2015-12-14 15:10:25 -05009742 Multi-cert bundling (see "crt") is support with crt-list, as long as only the
9743 base name is given in the crt-list. Due to the nature of bundling, all SNI
9744 filters given to a multi-cert bundle entry are ignored.
9745
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009746defer-accept
9747 Is an optional keyword which is supported only on certain Linux kernels. It
9748 states that a connection will only be accepted once some data arrive on it,
9749 or at worst after the first retransmit. This should be used only on protocols
9750 for which the client talks first (eg: HTTP). It can slightly improve
9751 performance by ensuring that most of the request is already available when
9752 the connection is accepted. On the other hand, it will not be able to detect
9753 connections which don't talk. It is important to note that this option is
9754 broken in all kernels up to 2.6.31, as the connection is never accepted until
9755 the client talks. This can cause issues with front firewalls which would see
9756 an established connection while the proxy will only see it in SYN_RECV. This
9757 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
9758
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009759force-sslv3
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009760 This option enforces use of SSLv3 only on SSL connections instantiated from
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009761 this listener. SSLv3 is generally less expensive than the TLS counterparts
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009762 for high connection rates. This option is also available on global statement
9763 "ssl-default-bind-options". See also "no-tlsv*" and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009764
9765force-tlsv10
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009766 This option enforces use of TLSv1.0 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009767 this listener. This option is also available on global statement
9768 "ssl-default-bind-options". See also "no-tlsv*" and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009769
9770force-tlsv11
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009771 This option enforces use of TLSv1.1 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009772 this listener. This option is also available on global statement
9773 "ssl-default-bind-options". See also "no-tlsv*", and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009774
9775force-tlsv12
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009776 This option enforces use of TLSv1.2 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009777 this listener. This option is also available on global statement
9778 "ssl-default-bind-options". See also "no-tlsv*", and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009779
Christopher Faulet31af49d2015-06-09 17:29:50 +02009780generate-certificates
9781 This setting is only available when support for OpenSSL was built in. It
9782 enables the dynamic SSL certificates generation. A CA certificate and its
9783 private key are necessary (see 'ca-sign-file'). When HAProxy is configured as
9784 a transparent forward proxy, SSL requests generate errors because of a common
9785 name mismatch on the certificate presented to the client. With this option
9786 enabled, HAProxy will try to forge a certificate using the SNI hostname
9787 indicated by the client. This is done only if no certificate matches the SNI
9788 hostname (see 'crt-list'). If an error occurs, the default certificate is
9789 used, else the 'strict-sni' option is set.
9790 It can also be used when HAProxy is configured as a reverse proxy to ease the
9791 deployment of an architecture with many backends.
9792
9793 Creating a SSL certificate is an expensive operation, so a LRU cache is used
9794 to store forged certificates (see 'tune.ssl.ssl-ctx-cache-size'). It
9795 increases the HAProxy's memroy footprint to reduce latency when the same
9796 certificate is used many times.
9797
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009798gid <gid>
9799 Sets the group of the UNIX sockets to the designated system gid. It can also
9800 be set by default in the global section's "unix-bind" statement. Note that
9801 some platforms simply ignore this. This setting is equivalent to the "group"
9802 setting except that the group ID is used instead of its name. This setting is
9803 ignored by non UNIX sockets.
9804
9805group <group>
9806 Sets the group of the UNIX sockets to the designated system group. It can
9807 also be set by default in the global section's "unix-bind" statement. Note
9808 that some platforms simply ignore this. This setting is equivalent to the
9809 "gid" setting except that the group name is used instead of its gid. This
9810 setting is ignored by non UNIX sockets.
9811
9812id <id>
9813 Fixes the socket ID. By default, socket IDs are automatically assigned, but
9814 sometimes it is more convenient to fix them to ease monitoring. This value
9815 must be strictly positive and unique within the listener/frontend. This
9816 option can only be used when defining only a single socket.
9817
9818interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +01009819 Restricts the socket to a specific interface. When specified, only packets
9820 received from that particular interface are processed by the socket. This is
9821 currently only supported on Linux. The interface must be a primary system
9822 interface, not an aliased interface. It is also possible to bind multiple
9823 frontends to the same address if they are bound to different interfaces. Note
9824 that binding to a network interface requires root privileges. This parameter
9825 is only compatible with TCPv4/TCPv6 sockets.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009826
Willy Tarreauabb175f2012-09-24 12:43:26 +02009827level <level>
9828 This setting is used with the stats sockets only to restrict the nature of
9829 the commands that can be issued on the socket. It is ignored by other
9830 sockets. <level> can be one of :
9831 - "user" is the least privileged level ; only non-sensitive stats can be
9832 read, and no change is allowed. It would make sense on systems where it
9833 is not easy to restrict access to the socket.
9834 - "operator" is the default level and fits most common uses. All data can
9835 be read, and only non-sensitive changes are permitted (eg: clear max
9836 counters).
9837 - "admin" should be used with care, as everything is permitted (eg: clear
9838 all counters).
9839
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009840maxconn <maxconn>
9841 Limits the sockets to this number of concurrent connections. Extraneous
9842 connections will remain in the system's backlog until a connection is
9843 released. If unspecified, the limit will be the same as the frontend's
9844 maxconn. Note that in case of port ranges or multiple addresses, the same
9845 value will be applied to each socket. This setting enables different
9846 limitations on expensive sockets, for instance SSL entries which may easily
9847 eat all memory.
9848
9849mode <mode>
9850 Sets the octal mode used to define access permissions on the UNIX socket. It
9851 can also be set by default in the global section's "unix-bind" statement.
9852 Note that some platforms simply ignore this. This setting is ignored by non
9853 UNIX sockets.
9854
9855mss <maxseg>
9856 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
9857 connections. This can be used to force a lower MSS for certain specific
9858 ports, for instance for connections passing through a VPN. Note that this
9859 relies on a kernel feature which is theoretically supported under Linux but
9860 was buggy in all versions prior to 2.6.28. It may or may not work on other
9861 operating systems. It may also not change the advertised value but change the
9862 effective size of outgoing segments. The commonly advertised value for TCPv4
9863 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
9864 positive, it will be used as the advertised MSS. If it is negative, it will
9865 indicate by how much to reduce the incoming connection's advertised MSS for
9866 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
9867
9868name <name>
9869 Sets an optional name for these sockets, which will be reported on the stats
9870 page.
9871
Willy Tarreaud72f0f32015-10-13 14:50:22 +02009872namespace <name>
9873 On Linux, it is possible to specify which network namespace a socket will
9874 belong to. This directive makes it possible to explicitly bind a listener to
9875 a namespace different from the default one. Please refer to your operating
9876 system's documentation to find more details about network namespaces.
9877
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009878nice <nice>
9879 Sets the 'niceness' of connections initiated from the socket. Value must be
9880 in the range -1024..1024 inclusive, and defaults to zero. Positive values
9881 means that such connections are more friendly to others and easily offer
9882 their place in the scheduler. On the opposite, negative values mean that
9883 connections want to run with a higher priority than others. The difference
9884 only happens under high loads when the system is close to saturation.
9885 Negative values are appropriate for low-latency or administration services,
9886 and high values are generally recommended for CPU intensive tasks such as SSL
9887 processing or bulk transfers which are less sensible to latency. For example,
9888 it may make sense to use a positive value for an SMTP socket and a negative
9889 one for an RDP socket.
9890
Emeric Brun9b3009b2012-10-05 11:55:06 +02009891no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009892 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009893 disables support for SSLv3 on any sockets instantiated from the listener when
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009894 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009895 be enabled using any configuration option. This option is also available on
9896 global statement "ssl-default-bind-options". See also "force-tls*",
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009897 and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009898
Emeric Brun90ad8722012-10-02 14:00:59 +02009899no-tls-tickets
9900 This setting is only available when support for OpenSSL was built in. It
9901 disables the stateless session resumption (RFC 5077 TLS Ticket
9902 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009903 session resumption is more expensive in CPU usage. This option is also
9904 available on global statement "ssl-default-bind-options".
Emeric Brun90ad8722012-10-02 14:00:59 +02009905
Emeric Brun9b3009b2012-10-05 11:55:06 +02009906no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009907 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009908 disables support for TLSv1.0 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009909 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009910 cannot be enabled using any configuration option. This option is also
9911 available on global statement "ssl-default-bind-options". See also
9912 "force-tlsv*", and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009913
Emeric Brun9b3009b2012-10-05 11:55:06 +02009914no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +02009915 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009916 disables support for TLSv1.1 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009917 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009918 cannot be enabled using any configuration option. This option is also
9919 available on global statement "ssl-default-bind-options". See also
9920 "force-tlsv*", and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +02009921
Emeric Brun9b3009b2012-10-05 11:55:06 +02009922no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +02009923 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009924 disables support for TLSv1.2 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +02009925 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009926 cannot be enabled using any configuration option. This option is also
9927 available on global statement "ssl-default-bind-options". See also
9928 "force-tlsv*", and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +02009929
Willy Tarreau6c9a3d52012-10-18 18:57:14 +02009930npn <protocols>
9931 This enables the NPN TLS extension and advertises the specified protocol list
9932 as supported on top of NPN. The protocol list consists in a comma-delimited
9933 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
9934 This requires that the SSL library is build with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +02009935 enabled (check with haproxy -vv). Note that the NPN extension has been
9936 replaced with the ALPN extension (see the "alpn" keyword).
Willy Tarreau6c9a3d52012-10-18 18:57:14 +02009937
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02009938process [ all | odd | even | <number 1-64>[-<number 1-64>] ]
9939 This restricts the list of processes on which this listener is allowed to
9940 run. It does not enforce any process but eliminates those which do not match.
9941 If the frontend uses a "bind-process" setting, the intersection between the
9942 two is applied. If in the end the listener is not allowed to run on any
9943 remaining process, a warning is emitted, and the listener will either run on
9944 the first process of the listener if a single process was specified, or on
9945 all of its processes if multiple processes were specified. For the unlikely
Willy Tarreauae302532014-05-07 19:22:24 +02009946 case where several ranges are needed, this directive may be repeated. The
9947 main purpose of this directive is to be used with the stats sockets and have
9948 one different socket per process. The second purpose is to have multiple bind
9949 lines sharing the same IP:port but not the same process in a listener, so
9950 that the system can distribute the incoming connections into multiple queues
9951 and allow a smoother inter-process load balancing. Currently Linux 3.9 and
9952 above is known for supporting this. See also "bind-process" and "nbproc".
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02009953
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009954ssl
9955 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009956 enables SSL deciphering on connections instantiated from this listener. A
Willy Tarreaub6205fd2012-09-24 12:27:33 +02009957 certificate is necessary (see "crt" above). All contents in the buffers will
9958 appear in clear text, so that ACLs and HTTP processing will only have access
9959 to deciphered contents.
9960
Emmanuel Hocdet65623372013-01-24 17:17:15 +01009961strict-sni
9962 This setting is only available when support for OpenSSL was built in. The
9963 SSL/TLS negotiation is allow only if the client provided an SNI which match
9964 a certificate. The default certificate is not used.
9965 See the "crt" option for more information.
9966
Willy Tarreau2af207a2015-02-04 00:45:58 +01009967tcp-ut <delay>
9968 Sets the TCP User Timeout for all incoming connections instanciated from this
9969 listening socket. This option is available on Linux since version 2.6.37. It
9970 allows haproxy to configure a timeout for sockets which contain data not
9971 receiving an acknoledgement for the configured delay. This is especially
9972 useful on long-lived connections experiencing long idle periods such as
9973 remote terminals or database connection pools, where the client and server
9974 timeouts must remain high to allow a long period of idle, but where it is
9975 important to detect that the client has disappeared in order to release all
9976 resources associated with its connection (and the server's session). The
9977 argument is a delay expressed in milliseconds by default. This only works
9978 for regular TCP connections, and is ignored for other protocols.
9979
Willy Tarreau1c862c52012-10-05 16:21:00 +02009980tfo
Lukas Tribus0defb902013-02-13 23:35:39 +01009981 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +02009982 enables TCP Fast Open on the listening socket, which means that clients which
9983 support this feature will be able to send a request and receive a response
9984 during the 3-way handshake starting from second connection, thus saving one
9985 round-trip after the first connection. This only makes sense with protocols
9986 that use high connection rates and where each round trip matters. This can
9987 possibly cause issues with many firewalls which do not accept data on SYN
9988 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +02009989 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
9990 need to build HAProxy with USE_TFO=1 if your libc doesn't define
9991 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +02009992
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +01009993tls-ticket-keys <keyfile>
9994 Sets the TLS ticket keys file to load the keys from. The keys need to be 48
9995 bytes long, encoded with base64 (ex. openssl rand -base64 48). Number of keys
9996 is specified by the TLS_TICKETS_NO build option (default 3) and at least as
9997 many keys need to be present in the file. Last TLS_TICKETS_NO keys will be
9998 used for decryption and the penultimate one for encryption. This enables easy
9999 key rotation by just appending new key to the file and reloading the process.
10000 Keys must be periodically rotated (ex. every 12h) or Perfect Forward Secrecy
10001 is compromised. It is also a good idea to keep the keys off any permanent
10002 storage such as hard drives (hint: use tmpfs and don't swap those files).
10003 Lifetime hint can be changed using tune.ssl.timeout.
10004
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010005transparent
10006 Is an optional keyword which is supported only on certain Linux kernels. It
10007 indicates that the addresses will be bound even if they do not belong to the
10008 local machine, and that packets targeting any of these addresses will be
10009 intercepted just as if the addresses were locally configured. This normally
10010 requires that IP forwarding is enabled. Caution! do not use this with the
10011 default address '*', as it would redirect any traffic for the specified port.
10012 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
10013 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
10014 kernel version. Some distribution kernels include backports of the feature,
10015 so check for support with your vendor.
10016
Willy Tarreau77e3af92012-11-24 15:07:23 +010010017v4v6
10018 Is an optional keyword which is supported only on most recent systems
10019 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
10020 and IPv6 when it uses the default address. Doing so is sometimes necessary
10021 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010022 sockets, and is overridden by the "v6only" option.
Willy Tarreau77e3af92012-11-24 15:07:23 +010010023
Willy Tarreau9b6700f2012-11-24 11:55:28 +010010024v6only
10025 Is an optional keyword which is supported only on most recent systems
10026 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
10027 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +010010028 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
10029 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +010010030
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010031uid <uid>
10032 Sets the owner of the UNIX sockets to the designated system uid. It can also
10033 be set by default in the global section's "unix-bind" statement. Note that
10034 some platforms simply ignore this. This setting is equivalent to the "user"
10035 setting except that the user numeric ID is used instead of its name. This
10036 setting is ignored by non UNIX sockets.
10037
10038user <user>
10039 Sets the owner of the UNIX sockets to the designated system user. It can also
10040 be set by default in the global section's "unix-bind" statement. Note that
10041 some platforms simply ignore this. This setting is equivalent to the "uid"
10042 setting except that the user name is used instead of its uid. This setting is
10043 ignored by non UNIX sockets.
10044
Emeric Brun1a073b42012-09-28 17:07:34 +020010045verify [none|optional|required]
10046 This setting is only available when support for OpenSSL was built in. If set
10047 to 'none', client certificate is not requested. This is the default. In other
10048 cases, a client certificate is requested. If the client does not provide a
10049 certificate after the request and if 'verify' is set to 'required', then the
10050 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +020010051 certificate provided by the client is always verified using CAs from
10052 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
10053 is aborted, regardless of the 'verify' option, unless the error code exactly
10054 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020010055
Willy Tarreaub6205fd2012-09-24 12:27:33 +0200100565.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +010010057------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020010058
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010010059The "server" and "default-server" keywords support a certain number of settings
10060which are all passed as arguments on the server line. The order in which those
10061arguments appear does not count, and they are all optional. Some of those
10062settings are single words (booleans) while others expect one or several values
10063after them. In this case, the values must immediately follow the setting name.
10064Except default-server, all those settings must be specified after the server's
10065address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +020010066
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010067 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010010068 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +020010069
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010070The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010071
Willy Tarreauceb4ac92012-04-28 00:41:46 +020010072addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010073 Using the "addr" parameter, it becomes possible to use a different IP address
10074 to send health-checks. On some servers, it may be desirable to dedicate an IP
10075 address to specific component able to perform complex tests which are more
10076 suitable to health-checks than the application. This parameter is ignored if
10077 the "check" parameter is not set. See also the "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010078
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010079 Supported in default-server: No
10080
Simon Hormand60d6912013-11-25 10:46:36 +090010081agent-check
10082 Enable an auxiliary agent check which is run independently of a regular
Willy Tarreau81f5d942013-12-09 20:51:51 +010010083 health check. An agent health check is performed by making a TCP connection
10084 to the port set by the "agent-port" parameter and reading an ASCII string.
10085 The string is made of a series of words delimited by spaces, tabs or commas
10086 in any order, optionally terminated by '\r' and/or '\n', each consisting of :
Simon Hormand60d6912013-11-25 10:46:36 +090010087
Willy Tarreau81f5d942013-12-09 20:51:51 +010010088 - An ASCII representation of a positive integer percentage, e.g. "75%".
Simon Hormand60d6912013-11-25 10:46:36 +090010089 Values in this format will set the weight proportional to the initial
Willy Tarreauc5af3a62014-10-07 15:27:33 +020010090 weight of a server as configured when haproxy starts. Note that a zero
10091 weight is reported on the stats page as "DRAIN" since it has the same
10092 effect on the server (it's removed from the LB farm).
Simon Hormand60d6912013-11-25 10:46:36 +090010093
Willy Tarreau81f5d942013-12-09 20:51:51 +010010094 - The word "ready". This will turn the server's administrative state to the
10095 READY mode, thus cancelling any DRAIN or MAINT state
Simon Hormand60d6912013-11-25 10:46:36 +090010096
Willy Tarreau81f5d942013-12-09 20:51:51 +010010097 - The word "drain". This will turn the server's administrative state to the
10098 DRAIN mode, thus it will not accept any new connections other than those
10099 that are accepted via persistence.
Simon Hormand60d6912013-11-25 10:46:36 +090010100
Willy Tarreau81f5d942013-12-09 20:51:51 +010010101 - The word "maint". This will turn the server's administrative state to the
10102 MAINT mode, thus it will not accept any new connections at all, and health
10103 checks will be stopped.
Simon Hormand60d6912013-11-25 10:46:36 +090010104
Willy Tarreau81f5d942013-12-09 20:51:51 +010010105 - The words "down", "failed", or "stopped", optionally followed by a
10106 description string after a sharp ('#'). All of these mark the server's
10107 operating state as DOWN, but since the word itself is reported on the stats
10108 page, the difference allows an administrator to know if the situation was
10109 expected or not : the service may intentionally be stopped, may appear up
10110 but fail some validity tests, or may be seen as down (eg: missing process,
10111 or port not responding).
Simon Hormand60d6912013-11-25 10:46:36 +090010112
Willy Tarreau81f5d942013-12-09 20:51:51 +010010113 - The word "up" sets back the server's operating state as UP if health checks
10114 also report that the service is accessible.
Simon Hormand60d6912013-11-25 10:46:36 +090010115
Willy Tarreau81f5d942013-12-09 20:51:51 +010010116 Parameters which are not advertised by the agent are not changed. For
10117 example, an agent might be designed to monitor CPU usage and only report a
10118 relative weight and never interact with the operating status. Similarly, an
10119 agent could be designed as an end-user interface with 3 radio buttons
10120 allowing an administrator to change only the administrative state. However,
10121 it is important to consider that only the agent may revert its own actions,
10122 so if a server is set to DRAIN mode or to DOWN state using the agent, the
10123 agent must implement the other equivalent actions to bring the service into
10124 operations again.
Simon Hormand60d6912013-11-25 10:46:36 +090010125
Simon Horman2f1f9552013-11-25 10:46:37 +090010126 Failure to connect to the agent is not considered an error as connectivity
10127 is tested by the regular health check which is enabled by the "check"
Willy Tarreau81f5d942013-12-09 20:51:51 +010010128 parameter. Warning though, it is not a good idea to stop an agent after it
10129 reports "down", since only an agent reporting "up" will be able to turn the
10130 server up again. Note that the CLI on the Unix stats socket is also able to
10131 force an agent's result in order to workaround a bogus agent if needed.
Simon Horman2f1f9552013-11-25 10:46:37 +090010132
Willy Tarreau81f5d942013-12-09 20:51:51 +010010133 Requires the "agent-port" parameter to be set. See also the "agent-inter"
10134 parameter.
Simon Hormand60d6912013-11-25 10:46:36 +090010135
10136 Supported in default-server: No
10137
James Brown55f9ff12015-10-21 18:19:05 -070010138agent-send <string>
10139 If this option is specified, haproxy will send the given string (verbatim)
10140 to the agent server upon connection. You could, for example, encode
10141 the backend name into this string, which would enable your agent to send
10142 different responses based on the backend. Make sure to include a '\n' if
10143 you want to terminate your request with a newline.
10144
Simon Hormand60d6912013-11-25 10:46:36 +090010145agent-inter <delay>
10146 The "agent-inter" parameter sets the interval between two agent checks
10147 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
10148
10149 Just as with every other time-based parameter, it may be entered in any
10150 other explicit unit among { us, ms, s, m, h, d }. The "agent-inter"
10151 parameter also serves as a timeout for agent checks "timeout check" is
10152 not set. In order to reduce "resonance" effects when multiple servers are
10153 hosted on the same hardware, the agent and health checks of all servers
10154 are started with a small time offset between them. It is also possible to
10155 add some random noise in the agent and health checks interval using the
10156 global "spread-checks" keyword. This makes sense for instance when a lot
10157 of backends use the same servers.
10158
10159 See also the "agent-check" and "agent-port" parameters.
10160
10161 Supported in default-server: Yes
10162
10163agent-port <port>
10164 The "agent-port" parameter sets the TCP port used for agent checks.
10165
10166 See also the "agent-check" and "agent-inter" parameters.
10167
10168 Supported in default-server: Yes
10169
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010170backup
10171 When "backup" is present on a server line, the server is only used in load
10172 balancing when all other non-backup servers are unavailable. Requests coming
10173 with a persistence cookie referencing the server will always be served
10174 though. By default, only the first operational backup server is used, unless
10175 the "allbackups" option is set in the backend. See also the "allbackups"
10176 option.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010177
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010178 Supported in default-server: No
10179
Emeric Brunef42d922012-10-11 16:11:36 +020010180ca-file <cafile>
10181 This setting is only available when support for OpenSSL was built in. It
10182 designates a PEM file from which to load CA certificates used to verify
10183 server's certificate.
10184
10185 Supported in default-server: No
10186
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010187check
10188 This option enables health checks on the server. By default, a server is
Patrick Mézardb7aeec62012-01-22 16:01:22 +010010189 always considered available. If "check" is set, the server is available when
10190 accepting periodic TCP connections, to ensure that it is really able to serve
10191 requests. The default address and port to send the tests to are those of the
10192 server, and the default source is the same as the one defined in the
10193 backend. It is possible to change the address using the "addr" parameter, the
10194 port using the "port" parameter, the source address using the "source"
10195 address, and the interval and timers using the "inter", "rise" and "fall"
Simon Hormanafc47ee2013-11-25 10:46:35 +090010196 parameters. The request method is define in the backend using the "httpchk",
10197 "smtpchk", "mysql-check", "pgsql-check" and "ssl-hello-chk" options. Please
10198 refer to those options and parameters for more information.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010199
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010200 Supported in default-server: No
10201
Willy Tarreau6c16adc2012-10-05 00:04:16 +020010202check-send-proxy
10203 This option forces emission of a PROXY protocol line with outgoing health
10204 checks, regardless of whether the server uses send-proxy or not for the
10205 normal traffic. By default, the PROXY protocol is enabled for health checks
10206 if it is already enabled for normal traffic and if no "port" nor "addr"
10207 directive is present. However, if such a directive is present, the
10208 "check-send-proxy" option needs to be used to force the use of the
10209 protocol. See also the "send-proxy" option for more information.
10210
10211 Supported in default-server: No
10212
Willy Tarreau763a95b2012-10-04 23:15:39 +020010213check-ssl
10214 This option forces encryption of all health checks over SSL, regardless of
10215 whether the server uses SSL or not for the normal traffic. This is generally
10216 used when an explicit "port" or "addr" directive is specified and SSL health
10217 checks are not inherited. It is important to understand that this option
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010218 inserts an SSL transport layer below the checks, so that a simple TCP connect
Willy Tarreau763a95b2012-10-04 23:15:39 +020010219 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
10220 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
10221 All SSL settings are common to health checks and traffic (eg: ciphers).
10222 See the "ssl" option for more information.
10223
10224 Supported in default-server: No
10225
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010226ciphers <ciphers>
10227 This option sets the string describing the list of cipher algorithms that is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010228 is negotiated during the SSL/TLS handshake with the server. The format of the
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010229 string is defined in "man 1 ciphers". When SSL is used to communicate with
10230 servers on the local network, it is common to see a weaker set of algorithms
10231 than what is used over the internet. Doing so reduces CPU usage on both the
10232 server and haproxy while still keeping it compatible with deployed software.
10233 Some algorithms such as RC4-SHA1 are reasonably cheap. If no security at all
10234 is needed and just connectivity, using DES can be appropriate.
10235
Willy Tarreau763a95b2012-10-04 23:15:39 +020010236 Supported in default-server: No
10237
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010238cookie <value>
10239 The "cookie" parameter sets the cookie value assigned to the server to
10240 <value>. This value will be checked in incoming requests, and the first
10241 operational server possessing the same value will be selected. In return, in
10242 cookie insertion or rewrite modes, this value will be assigned to the cookie
10243 sent to the client. There is nothing wrong in having several servers sharing
10244 the same cookie value, and it is in fact somewhat common between normal and
10245 backup servers. See also the "cookie" keyword in backend section.
10246
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010247 Supported in default-server: No
10248
Emeric Brunef42d922012-10-11 16:11:36 +020010249crl-file <crlfile>
10250 This setting is only available when support for OpenSSL was built in. It
10251 designates a PEM file from which to load certificate revocation list used
10252 to verify server's certificate.
10253
10254 Supported in default-server: No
10255
Emeric Bruna7aa3092012-10-26 12:58:00 +020010256crt <cert>
10257 This setting is only available when support for OpenSSL was built in.
10258 It designates a PEM file from which to load both a certificate and the
10259 associated private key. This file can be built by concatenating both PEM
10260 files into one. This certificate will be sent if the server send a client
10261 certificate request.
10262
10263 Supported in default-server: No
10264
Willy Tarreau96839092010-03-29 10:02:24 +020010265disabled
10266 The "disabled" keyword starts the server in the "disabled" state. That means
10267 that it is marked down in maintenance mode, and no connection other than the
10268 ones allowed by persist mode will reach it. It is very well suited to setup
10269 new servers, because normal traffic will never reach them, while it is still
10270 possible to test the service by making use of the force-persist mechanism.
10271
10272 Supported in default-server: No
10273
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010274error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +010010275 If health observing is enabled, the "error-limit" parameter specifies the
10276 number of consecutive errors that triggers event selected by the "on-error"
10277 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010278
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010279 Supported in default-server: Yes
10280
10281 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010282
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010283fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010284 The "fall" parameter states that a server will be considered as dead after
10285 <count> consecutive unsuccessful health checks. This value defaults to 3 if
10286 unspecified. See also the "check", "inter" and "rise" parameters.
10287
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010288 Supported in default-server: Yes
10289
Emeric Brun8694b9a2012-10-05 14:39:07 +020010290force-sslv3
10291 This option enforces use of SSLv3 only when SSL is used to communicate with
10292 the server. SSLv3 is generally less expensive than the TLS counterparts for
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010293 high connection rates. This option is also available on global statement
10294 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010295
10296 Supported in default-server: No
10297
10298force-tlsv10
10299 This option enforces use of TLSv1.0 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010300 the server. This option is also available on global statement
10301 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010302
10303 Supported in default-server: No
10304
10305force-tlsv11
10306 This option enforces use of TLSv1.1 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010307 the server. This option is also available on global statement
10308 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010309
10310 Supported in default-server: No
10311
10312force-tlsv12
10313 This option enforces use of TLSv1.2 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010314 the server. This option is also available on global statement
10315 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +020010316
10317 Supported in default-server: No
10318
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010319id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +020010320 Set a persistent ID for the server. This ID must be positive and unique for
10321 the proxy. An unused ID will automatically be assigned if unset. The first
10322 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010323
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010324 Supported in default-server: No
10325
10326inter <delay>
10327fastinter <delay>
10328downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010329 The "inter" parameter sets the interval between two consecutive health checks
10330 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
10331 It is also possible to use "fastinter" and "downinter" to optimize delays
10332 between checks depending on the server state :
10333
Pieter Baauw44fc9df2015-09-17 21:30:46 +020010334 Server state | Interval used
10335 ----------------------------------------+----------------------------------
10336 UP 100% (non-transitional) | "inter"
10337 ----------------------------------------+----------------------------------
10338 Transitionally UP (going down "fall"), | "fastinter" if set,
10339 Transitionally DOWN (going up "rise"), | "inter" otherwise.
10340 or yet unchecked. |
10341 ----------------------------------------+----------------------------------
10342 DOWN 100% (non-transitional) | "downinter" if set,
10343 | "inter" otherwise.
10344 ----------------------------------------+----------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +010010345
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010346 Just as with every other time-based parameter, they can be entered in any
10347 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
10348 serves as a timeout for health checks sent to servers if "timeout check" is
10349 not set. In order to reduce "resonance" effects when multiple servers are
Simon Hormand60d6912013-11-25 10:46:36 +090010350 hosted on the same hardware, the agent and health checks of all servers
10351 are started with a small time offset between them. It is also possible to
10352 add some random noise in the agent and health checks interval using the
10353 global "spread-checks" keyword. This makes sense for instance when a lot
10354 of backends use the same servers.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010355
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010356 Supported in default-server: Yes
10357
10358maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010359 The "maxconn" parameter specifies the maximal number of concurrent
10360 connections that will be sent to this server. If the number of incoming
10361 concurrent requests goes higher than this value, they will be queued, waiting
10362 for a connection to be released. This parameter is very important as it can
10363 save fragile servers from going down under extreme loads. If a "minconn"
10364 parameter is specified, the limit becomes dynamic. The default value is "0"
10365 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
10366 the backend's "fullconn" keyword.
10367
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010368 Supported in default-server: Yes
10369
10370maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010371 The "maxqueue" parameter specifies the maximal number of connections which
10372 will wait in the queue for this server. If this limit is reached, next
10373 requests will be redispatched to other servers instead of indefinitely
10374 waiting to be served. This will break persistence but may allow people to
10375 quickly re-log in when the server they try to connect to is dying. The
10376 default value is "0" which means the queue is unlimited. See also the
10377 "maxconn" and "minconn" parameters.
10378
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010379 Supported in default-server: Yes
10380
10381minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010382 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
10383 limit following the backend's load. The server will always accept at least
10384 <minconn> connections, never more than <maxconn>, and the limit will be on
10385 the ramp between both values when the backend has less than <fullconn>
10386 concurrent connections. This makes it possible to limit the load on the
10387 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010388 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010389 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010390
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010391 Supported in default-server: Yes
10392
Willy Tarreaud72f0f32015-10-13 14:50:22 +020010393namespace <name>
10394 On Linux, it is possible to specify which network namespace a socket will
10395 belong to. This directive makes it possible to explicitly bind a server to
10396 a namespace different from the default one. Please refer to your operating
10397 system's documentation to find more details about network namespaces.
10398
Willy Tarreau2a3fb1c2015-02-05 16:47:07 +010010399no-ssl-reuse
10400 This option disables SSL session reuse when SSL is used to communicate with
10401 the server. It will force the server to perform a full handshake for every
10402 new connection. It's probably only useful for benchmarking, troubleshooting,
10403 and for paranoid users.
10404
10405 Supported in default-server: No
10406
Emeric Brun9b3009b2012-10-05 11:55:06 +020010407no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010408 This option disables support for SSLv3 when SSL is used to communicate with
10409 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emeric Brun8694b9a2012-10-05 14:39:07 +020010410 using any configuration option. See also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010411
Willy Tarreau763a95b2012-10-04 23:15:39 +020010412 Supported in default-server: No
10413
Emeric Brunf9c5c472012-10-11 15:28:34 +020010414no-tls-tickets
10415 This setting is only available when support for OpenSSL was built in. It
10416 disables the stateless session resumption (RFC 5077 TLS Ticket
10417 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010418 session resumption is more expensive in CPU usage for servers. This option
10419 is also available on global statement "ssl-default-server-options".
Emeric Brunf9c5c472012-10-11 15:28:34 +020010420
10421 Supported in default-server: No
10422
Emeric Brun9b3009b2012-10-05 11:55:06 +020010423no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +020010424 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020010425 the server. Note that SSLv2 is disabled in the code and cannot be enabled
10426 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010427 often makes sense to disable it when communicating with local servers. This
10428 option is also available on global statement "ssl-default-server-options".
10429 See also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +020010430
Willy Tarreau763a95b2012-10-04 23:15:39 +020010431 Supported in default-server: No
10432
Emeric Brun9b3009b2012-10-05 11:55:06 +020010433no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +020010434 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020010435 the server. Note that SSLv2 is disabled in the code and cannot be enabled
10436 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010437 often makes sense to disable it when communicating with local servers. This
10438 option is also available on global statement "ssl-default-server-options".
10439 See also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +020010440
Willy Tarreau763a95b2012-10-04 23:15:39 +020010441 Supported in default-server: No
10442
Emeric Brun9b3009b2012-10-05 11:55:06 +020010443no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +020010444 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010445 the server. Note that SSLv2 is disabled in the code and cannot be enabled
10446 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010447 often makes sense to disable it when communicating with local servers. This
10448 option is also available on global statement "ssl-default-server-options".
10449 See also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010450
Willy Tarreau763a95b2012-10-04 23:15:39 +020010451 Supported in default-server: No
10452
Simon Hormanfa461682011-06-25 09:39:49 +090010453non-stick
10454 Never add connections allocated to this sever to a stick-table.
10455 This may be used in conjunction with backup to ensure that
10456 stick-table persistence is disabled for backup servers.
10457
Willy Tarreau763a95b2012-10-04 23:15:39 +020010458 Supported in default-server: No
10459
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010460observe <mode>
10461 This option enables health adjusting based on observing communication with
10462 the server. By default this functionality is disabled and enabling it also
10463 requires to enable health checks. There are two supported modes: "layer4" and
10464 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
10465 significant. In layer7, which is only allowed for http proxies, responses
10466 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +010010467 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010468
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010469 Supported in default-server: No
10470
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010471 See also the "check", "on-error" and "error-limit".
10472
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010473on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010474 Select what should happen when enough consecutive errors are detected.
10475 Currently, four modes are available:
10476 - fastinter: force fastinter
10477 - fail-check: simulate a failed check, also forces fastinter (default)
10478 - sudden-death: simulate a pre-fatal failed health check, one more failed
10479 check will mark a server down, forces fastinter
10480 - mark-down: mark the server immediately down and force fastinter
10481
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010482 Supported in default-server: Yes
10483
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010010484 See also the "check", "observe" and "error-limit".
10485
Simon Hormane0d1bfb2011-06-21 14:34:58 +090010486on-marked-down <action>
10487 Modify what occurs when a server is marked down.
10488 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070010489 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
10490 all connections to the server are immediately terminated when the server
10491 goes down. It might be used if the health check detects more complex cases
10492 than a simple connection status, and long timeouts would cause the service
10493 to remain unresponsive for too long a time. For instance, a health check
10494 might detect that a database is stuck and that there's no chance to reuse
10495 existing connections anymore. Connections killed this way are logged with
10496 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +090010497
10498 Actions are disabled by default
10499
10500 Supported in default-server: Yes
10501
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070010502on-marked-up <action>
10503 Modify what occurs when a server is marked up.
10504 Currently one action is available:
10505 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
10506 done only if the server is not in backup state and if it is not disabled
10507 (it must have an effective weight > 0). This can be used sometimes to force
10508 an active server to take all the traffic back after recovery when dealing
10509 with long sessions (eg: LDAP, SQL, ...). Doing this can cause more trouble
10510 than it tries to solve (eg: incomplete transactions), so use this feature
10511 with extreme care. Sessions killed because a server comes up are logged
10512 with an 'U' termination code (for "Up").
10513
10514 Actions are disabled by default
10515
10516 Supported in default-server: Yes
10517
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010518port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010519 Using the "port" parameter, it becomes possible to use a different port to
10520 send health-checks. On some servers, it may be desirable to dedicate a port
10521 to a specific component able to perform complex tests which are more suitable
10522 to health-checks than the application. It is common to run a simple script in
10523 inetd for instance. This parameter is ignored if the "check" parameter is not
10524 set. See also the "addr" parameter.
10525
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010526 Supported in default-server: Yes
10527
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010528redir <prefix>
10529 The "redir" parameter enables the redirection mode for all GET and HEAD
10530 requests addressing this server. This means that instead of having HAProxy
10531 forward the request to the server, it will send an "HTTP 302" response with
10532 the "Location" header composed of this prefix immediately followed by the
10533 requested URI beginning at the leading '/' of the path component. That means
10534 that no trailing slash should be used after <prefix>. All invalid requests
10535 will be rejected, and all non-GET or HEAD requests will be normally served by
10536 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010537 mangling nor cookie insertion is possible in the response. However, cookies in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010538 requests are still analysed, making this solution completely usable to direct
10539 users to a remote location in case of local disaster. Main use consists in
10540 increasing bandwidth for static servers by having the clients directly
10541 connect to them. Note: never use a relative location here, it would cause a
10542 loop between the client and HAProxy!
10543
10544 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
10545
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010546 Supported in default-server: No
10547
10548rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010549 The "rise" parameter states that a server will be considered as operational
10550 after <count> consecutive successful health checks. This value defaults to 2
10551 if unspecified. See also the "check", "inter" and "fall" parameters.
10552
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010553 Supported in default-server: Yes
10554
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010555resolve-prefer <family>
10556 When DNS resolution is enabled for a server and multiple IP addresses from
10557 different families are returned, HAProxy will prefer using an IP address
10558 from the family mentioned in the "resolve-prefer" parameter.
10559 Available families: "ipv4" and "ipv6"
10560
Baptiste Assmannc4aabae2015-08-04 22:43:06 +020010561 Default value: ipv6
10562
10563 Supported in default-server: Yes
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010564
10565 Example: server s1 app1.domain.com:80 resolvers mydns resolve-prefer ipv6
10566
10567resolvers <id>
10568 Points to an existing "resolvers" section to resolve current server's
10569 hostname.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020010570 In order to be operational, DNS resolution requires that health check is
10571 enabled on the server. Actually, health checks triggers the DNS resolution.
10572 You must precise one 'resolvers' parameter on each server line where DNS
10573 resolution is required.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010574
Baptiste Assmann62b75b42015-09-09 01:11:36 +020010575 Supported in default-server: No
10576
10577 Example: server s1 app1.domain.com:80 check resolvers mydns
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010578
10579 See also chapter 5.3
10580
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010010581send-proxy
10582 The "send-proxy" parameter enforces use of the PROXY protocol over any
10583 connection established to this server. The PROXY protocol informs the other
10584 end about the layer 3/4 addresses of the incoming connection, so that it can
10585 know the client's address or the public address it accessed to, whatever the
10586 upper layer protocol. For connections accepted by an "accept-proxy" listener,
10587 the advertised address will be used. Only TCPv4 and TCPv6 address families
10588 are supported. Other families such as Unix sockets, will report an UNKNOWN
10589 family. Servers using this option can fully be chained to another instance of
10590 haproxy listening with an "accept-proxy" setting. This setting must not be
Willy Tarreau6c16adc2012-10-05 00:04:16 +020010591 used if the server isn't aware of the protocol. When health checks are sent
10592 to the server, the PROXY protocol is automatically used when this option is
10593 set, unless there is an explicit "port" or "addr" directive, in which case an
10594 explicit "check-send-proxy" directive would also be needed to use the PROXY
10595 protocol. See also the "accept-proxy" option of the "bind" keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010010596
10597 Supported in default-server: No
10598
David Safb76832014-05-08 23:42:08 -040010599send-proxy-v2
10600 The "send-proxy-v2" parameter enforces use of the PROXY protocol version 2
10601 over any connection established to this server. The PROXY protocol informs
10602 the other end about the layer 3/4 addresses of the incoming connection, so
10603 that it can know the client's address or the public address it accessed to,
10604 whatever the upper layer protocol. This setting must not be used if the
10605 server isn't aware of this version of the protocol. See also the "send-proxy"
10606 option of the "bind" keyword.
10607
10608 Supported in default-server: No
10609
10610send-proxy-v2-ssl
10611 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
10612 2 over any connection established to this server. The PROXY protocol informs
10613 the other end about the layer 3/4 addresses of the incoming connection, so
10614 that it can know the client's address or the public address it accessed to,
10615 whatever the upper layer protocol. In addition, the SSL information extension
10616 of the PROXY protocol is added to the PROXY protocol header. This setting
10617 must not be used if the server isn't aware of this version of the protocol.
10618 See also the "send-proxy-v2" option of the "bind" keyword.
10619
10620 Supported in default-server: No
10621
10622send-proxy-v2-ssl-cn
10623 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
10624 2 over any connection established to this server. The PROXY protocol informs
10625 the other end about the layer 3/4 addresses of the incoming connection, so
10626 that it can know the client's address or the public address it accessed to,
10627 whatever the upper layer protocol. In addition, the SSL information extension
10628 of the PROXY protocol, along along with the Common Name from the subject of
10629 the client certificate (if any), is added to the PROXY protocol header. This
10630 setting must not be used if the server isn't aware of this version of the
10631 protocol. See also the "send-proxy-v2" option of the "bind" keyword.
10632
10633 Supported in default-server: No
10634
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010635slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010636 The "slowstart" parameter for a server accepts a value in milliseconds which
10637 indicates after how long a server which has just come back up will run at
10638 full speed. Just as with every other time-based parameter, it can be entered
10639 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
10640 linearly from 0 to 100% during this time. The limitation applies to two
10641 parameters :
10642
10643 - maxconn: the number of connections accepted by the server will grow from 1
10644 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
10645
10646 - weight: when the backend uses a dynamic weighted algorithm, the weight
10647 grows linearly from 1 to 100%. In this case, the weight is updated at every
10648 health-check. For this reason, it is important that the "inter" parameter
10649 is smaller than the "slowstart", in order to maximize the number of steps.
10650
10651 The slowstart never applies when haproxy starts, otherwise it would cause
10652 trouble to running servers. It only applies when a server has been previously
10653 seen as failed.
10654
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010655 Supported in default-server: Yes
10656
Willy Tarreau732eac42015-07-09 11:40:25 +020010657sni <expression>
10658 The "sni" parameter evaluates the sample fetch expression, converts it to a
10659 string and uses the result as the host name sent in the SNI TLS extension to
10660 the server. A typical use case is to send the SNI received from the client in
10661 a bridged HTTPS scenario, using the "ssl_fc_sni" sample fetch for the
10662 expression, though alternatives such as req.hdr(host) can also make sense.
10663
10664 Supported in default-server: no
10665
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020010666source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +020010667source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020010668source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010669 The "source" parameter sets the source address which will be used when
10670 connecting to the server. It follows the exact same parameters and principle
10671 as the backend "source" keyword, except that it only applies to the server
10672 referencing it. Please consult the "source" keyword for details.
10673
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020010674 Additionally, the "source" statement on a server line allows one to specify a
10675 source port range by indicating the lower and higher bounds delimited by a
10676 dash ('-'). Some operating systems might require a valid IP address when a
10677 source port range is specified. It is permitted to have the same IP/range for
10678 several servers. Doing so makes it possible to bypass the maximum of 64k
10679 total concurrent connections. The limit will then reach 64k connections per
10680 server.
10681
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010682 Supported in default-server: No
10683
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010684ssl
Willy Tarreau44f65392013-06-25 07:56:20 +020010685 This option enables SSL ciphering on outgoing connections to the server. It
10686 is critical to verify server certificates using "verify" when using SSL to
10687 connect to servers, otherwise the communication is prone to trivial man in
10688 the-middle attacks rendering SSL useless. When this option is used, health
10689 checks are automatically sent in SSL too unless there is a "port" or an
10690 "addr" directive indicating the check should be sent to a different location.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010691 See the "check-ssl" option to force SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +020010692
10693 Supported in default-server: No
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020010694
Willy Tarreau163d4622015-10-13 16:16:41 +020010695tcp-ut <delay>
10696 Sets the TCP User Timeout for all outgoing connections to this server. This
10697 option is available on Linux since version 2.6.37. It allows haproxy to
10698 configure a timeout for sockets which contain data not receiving an
10699 acknoledgement for the configured delay. This is especially useful on
10700 long-lived connections experiencing long idle periods such as remote
10701 terminals or database connection pools, where the client and server timeouts
10702 must remain high to allow a long period of idle, but where it is important to
10703 detect that the server has disappeared in order to release all resources
10704 associated with its connection (and the client's session). One typical use
10705 case is also to force dead server connections to die when health checks are
10706 too slow or during a soft reload since health checks are then disabled. The
10707 argument is a delay expressed in milliseconds by default. This only works for
10708 regular TCP connections, and is ignored for other protocols.
10709
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010710track [<proxy>/]<server>
Willy Tarreau32091232014-05-16 13:52:00 +020010711 This option enables ability to set the current state of the server by tracking
10712 another one. It is possible to track a server which itself tracks another
10713 server, provided that at the end of the chain, a server has health checks
10714 enabled. If <proxy> is omitted the current one is used. If disable-on-404 is
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010715 used, it has to be enabled on both proxies.
10716
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010717 Supported in default-server: No
10718
Emeric Brunef42d922012-10-11 16:11:36 +020010719verify [none|required]
10720 This setting is only available when support for OpenSSL was built in. If set
Emeric Brun850efd52014-01-29 12:24:34 +010010721 to 'none', server certificate is not verified. In the other case, The
10722 certificate provided by the server is verified using CAs from 'ca-file'
10723 and optional CRLs from 'crl-file'. If 'ssl_server_verify' is not specified
10724 in global section, this is the default. On verify failure the handshake
Willy Tarreau44f65392013-06-25 07:56:20 +020010725 is aborted. It is critically important to verify server certificates when
10726 using SSL to connect to servers, otherwise the communication is prone to
10727 trivial man-in-the-middle attacks rendering SSL totally useless.
Emeric Brunef42d922012-10-11 16:11:36 +020010728
10729 Supported in default-server: No
10730
Evan Broderbe554312013-06-27 00:05:25 -070010731verifyhost <hostname>
10732 This setting is only available when support for OpenSSL was built in, and
10733 only takes effect if 'verify required' is also specified. When set, the
10734 hostnames in the subject and subjectAlternateNames of the certificate
10735 provided by the server are checked. If none of the hostnames in the
10736 certificate match the specified hostname, the handshake is aborted. The
10737 hostnames in the server-provided certificate may include wildcards.
10738
10739 Supported in default-server: No
10740
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010741weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010742 The "weight" parameter is used to adjust the server's weight relative to
10743 other servers. All servers will receive a load proportional to their weight
10744 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +020010745 load. The default weight is 1, and the maximal value is 256. A value of 0
10746 means the server will not participate in load-balancing but will still accept
10747 persistent connections. If this parameter is used to distribute the load
10748 according to server's capacity, it is recommended to start with values which
10749 can both grow and shrink, for instance between 10 and 100 to leave enough
10750 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010751
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010010752 Supported in default-server: Yes
10753
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010754
Cyril Bonté46175dd2015-07-02 22:45:32 +0200107555.3. Server IP address resolution using DNS
10756-------------------------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010757
Baptiste Assmann62b75b42015-09-09 01:11:36 +020010758HAProxy allows using a host name on the server line to retrieve its IP address
10759using name servers. By default, HAProxy resolves the name when parsing the
10760configuration file, at startup and cache the result for the process' life.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010761This is not sufficient in some cases, such as in Amazon where a server's IP
10762can change after a reboot or an ELB Virtual IP can change based on current
10763workload.
10764This chapter describes how HAProxy can be configured to process server's name
10765resolution at run time.
10766Whether run time server name resolution has been enable or not, HAProxy will
10767carry on doing the first resolution when parsing the configuration.
10768
Baptiste Assmann62b75b42015-09-09 01:11:36 +020010769Bear in mind that DNS resolution is triggered by health checks. This makes
10770health checks mandatory to allow DNS resolution.
10771
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010772
Cyril Bonté46175dd2015-07-02 22:45:32 +0200107735.3.1. Global overview
10774----------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010775
10776As we've seen in introduction, name resolution in HAProxy occurs at two
10777different steps of the process life:
10778
10779 1. when starting up, HAProxy parses the server line definition and matches a
10780 host name. It uses libc functions to get the host name resolved. This
10781 resolution relies on /etc/resolv.conf file.
10782
10783 2. at run time, when HAProxy gets prepared to run a health check on a server,
10784 it verifies if the current name resolution is still considered as valid.
10785 If not, it processes a new resolution, in parallel of the health check.
10786
10787A few other events can trigger a name resolution at run time:
10788 - when a server's health check ends up in a connection timeout: this may be
10789 because the server has a new IP address. So we need to trigger a name
10790 resolution to know this new IP.
10791
10792A few things important to notice:
10793 - all the name servers are queried in the mean time. HAProxy will process the
10794 first valid response.
10795
10796 - a resolution is considered as invalid (NX, timeout, refused), when all the
10797 servers return an error.
10798
10799
Cyril Bonté46175dd2015-07-02 22:45:32 +0200108005.3.2. The resolvers section
10801----------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010802
10803This section is dedicated to host information related to name resolution in
10804HAProxy.
10805There can be as many as resolvers section as needed. Each section can contain
10806many name servers.
10807
Baptiste Assmann62b75b42015-09-09 01:11:36 +020010808When multiple name servers are configured in a resolvers section, then HAProxy
10809uses the first valid response. In case of invalid responses, only the last one
10810is treated. Purpose is to give the chance to a slow server to deliver a valid
10811answer after a fast faulty or outdated server.
10812
10813When each server returns a different error type, then only the last error is
10814used by HAProxy to decide what type of behavior to apply.
10815
10816Two types of behavior can be applied:
10817 1. stop DNS resolution
10818 2. replay the DNS query with a new query type
10819 In such case, the following types are applied in this exact order:
10820 1. ANY query type
10821 2. query type corresponding to family pointed by resolve-prefer
10822 server's parameter
10823 3. remaining family type
10824
10825HAProxy stops DNS resolution when the following errors occur:
10826 - invalid DNS response packet
10827 - wrong name in the query section of the response
10828 - NX domain
10829 - Query refused by server
10830 - CNAME not pointing to an IP address
10831
10832HAProxy tries a new query type when the following errors occur:
10833 - no Answer records in the response
10834 - DNS response truncated
10835 - Error in DNS response
10836 - No expected DNS records found in the response
10837 - name server timeout
10838
10839For example, with 2 name servers configured in a resolvers section:
10840 - first response is valid and is applied directly, second response is ignored
10841 - first response is invalid and second one is valid, then second response is
10842 applied;
10843 - first response is a NX domain and second one a truncated response, then
10844 HAProxy replays the query with a new type;
10845 - first response is truncated and second one is a NX Domain, then HAProxy
10846 stops resolution.
10847
10848
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010849resolvers <resolvers id>
10850 Creates a new name server list labelled <resolvers id>
10851
10852A resolvers section accept the following parameters:
10853
10854nameserver <id> <ip>:<port>
10855 DNS server description:
10856 <id> : label of the server, should be unique
10857 <ip> : IP address of the server
10858 <port> : port where the DNS service actually runs
10859
10860hold <status> <period>
10861 Defines <period> during which the last name resolution should be kept based
10862 on last resolution <status>
10863 <status> : last name resolution status. Only "valid" is accepted for now.
10864 <period> : interval between two successive name resolution when the last
10865 answer was in <status>. It follows the HAProxy time format.
10866 <period> is in milliseconds by default.
10867
10868 Default value is 10s for "valid".
10869
10870 Note: since the name resolution is triggered by the health checks, a new
10871 resolution is triggered after <period> modulo the <inter> parameter of
10872 the healch check.
10873
10874resolve_retries <nb>
10875 Defines the number <nb> of queries to send to resolve a server name before
10876 giving up.
10877 Default value: 3
10878
Baptiste Assmann62b75b42015-09-09 01:11:36 +020010879 A retry occurs on name server timeout or when the full sequence of DNS query
10880 type failover is over and we need to start up from the default ANY query
10881 type.
10882
Baptiste Assmann1fa66662015-04-14 00:28:47 +020010883timeout <event> <time>
10884 Defines timeouts related to name resolution
10885 <event> : the event on which the <time> timeout period applies to.
10886 events available are:
10887 - retry: time between two DNS queries, when no response have
10888 been received.
10889 Default value: 1s
10890 <time> : time related to the event. It follows the HAProxy time format.
10891 <time> is expressed in milliseconds.
10892
10893Example of a resolvers section (with default values):
10894
10895 resolvers mydns
10896 nameserver dns1 10.0.0.1:53
10897 nameserver dns2 10.0.0.2:53
10898 resolve_retries 3
10899 timeout retry 1s
10900 hold valid 10s
10901
10902
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200109036. HTTP header manipulation
10904---------------------------
10905
10906In HTTP mode, it is possible to rewrite, add or delete some of the request and
10907response headers based on regular expressions. It is also possible to block a
10908request or a response if a particular header matches a regular expression,
10909which is enough to stop most elementary protocol attacks, and to protect
Willy Tarreau70dffda2014-01-30 03:07:23 +010010910against information leak from the internal network.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010911
Willy Tarreau70dffda2014-01-30 03:07:23 +010010912If HAProxy encounters an "Informational Response" (status code 1xx), it is able
10913to process all rsp* rules which can allow, deny, rewrite or delete a header,
10914but it will refuse to add a header to any such messages as this is not
10915HTTP-compliant. The reason for still processing headers in such responses is to
10916stop and/or fix any possible information leak which may happen, for instance
10917because another downstream equipment would unconditionally add a header, or if
10918a server name appears there. When such messages are seen, normal processing
10919still occurs on the next non-informational messages.
Willy Tarreau816b9792009-09-15 21:25:21 +020010920
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010921This section covers common usage of the following keywords, described in detail
10922in section 4.2 :
10923
10924 - reqadd <string>
10925 - reqallow <search>
10926 - reqiallow <search>
10927 - reqdel <search>
10928 - reqidel <search>
10929 - reqdeny <search>
10930 - reqideny <search>
10931 - reqpass <search>
10932 - reqipass <search>
10933 - reqrep <search> <replace>
10934 - reqirep <search> <replace>
10935 - reqtarpit <search>
10936 - reqitarpit <search>
10937 - rspadd <string>
10938 - rspdel <search>
10939 - rspidel <search>
10940 - rspdeny <search>
10941 - rspideny <search>
10942 - rsprep <search> <replace>
10943 - rspirep <search> <replace>
10944
10945With all these keywords, the same conventions are used. The <search> parameter
10946is a POSIX extended regular expression (regex) which supports grouping through
10947parenthesis (without the backslash). Spaces and other delimiters must be
10948prefixed with a backslash ('\') to avoid confusion with a field delimiter.
10949Other characters may be prefixed with a backslash to change their meaning :
10950
10951 \t for a tab
10952 \r for a carriage return (CR)
10953 \n for a new line (LF)
10954 \ to mark a space and differentiate it from a delimiter
10955 \# to mark a sharp and differentiate it from a comment
10956 \\ to use a backslash in a regex
10957 \\\\ to use a backslash in the text (*2 for regex, *2 for haproxy)
10958 \xXX to write the ASCII hex code XX as in the C language
10959
10960The <replace> parameter contains the string to be used to replace the largest
10961portion of text matching the regex. It can make use of the special characters
10962above, and can reference a substring which is delimited by parenthesis in the
10963regex, by writing a backslash ('\') immediately followed by one digit from 0 to
109649 indicating the group position (0 designating the entire line). This practice
10965is very common to users of the "sed" program.
10966
10967The <string> parameter represents the string which will systematically be added
10968after the last header line. It can also use special character sequences above.
10969
10970Notes related to these keywords :
10971---------------------------------
10972 - these keywords are not always convenient to allow/deny based on header
10973 contents. It is strongly recommended to use ACLs with the "block" keyword
10974 instead, resulting in far more flexible and manageable rules.
10975
10976 - lines are always considered as a whole. It is not possible to reference
10977 a header name only or a value only. This is important because of the way
10978 headers are written (notably the number of spaces after the colon).
10979
10980 - the first line is always considered as a header, which makes it possible to
10981 rewrite or filter HTTP requests URIs or response codes, but in turn makes
10982 it harder to distinguish between headers and request line. The regex prefix
10983 ^[^\ \t]*[\ \t] matches any HTTP method followed by a space, and the prefix
10984 ^[^ \t:]*: matches any header name followed by a colon.
10985
10986 - for performances reasons, the number of characters added to a request or to
10987 a response is limited at build time to values between 1 and 4 kB. This
10988 should normally be far more than enough for most usages. If it is too short
10989 on occasional usages, it is possible to gain some space by removing some
10990 useless headers before adding new ones.
10991
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010992 - keywords beginning with "reqi" and "rspi" are the same as their counterpart
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010993 without the 'i' letter except that they ignore case when matching patterns.
10994
10995 - when a request passes through a frontend then a backend, all req* rules
10996 from the frontend will be evaluated, then all req* rules from the backend
10997 will be evaluated. The reverse path is applied to responses.
10998
10999 - req* statements are applied after "block" statements, so that "block" is
11000 always the first one, but before "use_backend" in order to permit rewriting
Willy Tarreaud72758d2010-01-12 10:42:19 +010011001 before switching.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011002
11003
Willy Tarreau74ca5042013-06-11 23:12:07 +0200110047. Using ACLs and fetching samples
11005----------------------------------
11006
11007Haproxy is capable of extracting data from request or response streams, from
11008client or server information, from tables, environmental information etc...
11009The action of extracting such data is called fetching a sample. Once retrieved,
11010these samples may be used for various purposes such as a key to a stick-table,
11011but most common usages consist in matching them against predefined constant
11012data called patterns.
11013
11014
110157.1. ACL basics
11016---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011017
11018The use of Access Control Lists (ACL) provides a flexible solution to perform
11019content switching and generally to take decisions based on content extracted
11020from the request, the response or any environmental status. The principle is
11021simple :
11022
Willy Tarreau74ca5042013-06-11 23:12:07 +020011023 - extract a data sample from a stream, table or the environment
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011024 - optionally apply some format conversion to the extracted sample
Willy Tarreau74ca5042013-06-11 23:12:07 +020011025 - apply one or multiple pattern matching methods on this sample
11026 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011027
Willy Tarreau74ca5042013-06-11 23:12:07 +020011028The actions generally consist in blocking a request, selecting a backend, or
11029adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011030
11031In order to define a test, the "acl" keyword is used. The syntax is :
11032
Willy Tarreau74ca5042013-06-11 23:12:07 +020011033 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011034
11035This creates a new ACL <aclname> or completes an existing one with new tests.
11036Those tests apply to the portion of request/response specified in <criterion>
11037and may be adjusted with optional flags [flags]. Some criteria also support
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011038an operator which may be specified before the set of values. Optionally some
11039conversion operators may be applied to the sample, and they will be specified
11040as a comma-delimited list of keywords just after the first keyword. The values
11041are of the type supported by the criterion, and are separated by spaces.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011042
11043ACL names must be formed from upper and lower case letters, digits, '-' (dash),
11044'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
11045which means that "my_acl" and "My_Acl" are two different ACLs.
11046
11047There is no enforced limit to the number of ACLs. The unused ones do not affect
11048performance, they just consume a small amount of memory.
11049
Willy Tarreau74ca5042013-06-11 23:12:07 +020011050The criterion generally is the name of a sample fetch method, or one of its ACL
11051specific declinations. The default test method is implied by the output type of
11052this sample fetch method. The ACL declinations can describe alternate matching
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011053methods of a same sample fetch method. The sample fetch methods are the only
11054ones supporting a conversion.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011055
11056Sample fetch methods return data which can be of the following types :
11057 - boolean
11058 - integer (signed or unsigned)
11059 - IPv4 or IPv6 address
11060 - string
11061 - data block
11062
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011063Converters transform any of these data into any of these. For example, some
11064converters might convert a string to a lower-case string while other ones
11065would turn a string to an IPv4 address, or apply a netmask to an IP address.
11066The resulting sample is of the type of the last converter applied to the list,
11067which defaults to the type of the sample fetch method.
11068
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011069Each sample or converter returns data of a specific type, specified with its
11070keyword in this documentation. When an ACL is declared using a standard sample
11071fetch method, certain types automatically involved a default matching method
11072which are summarized in the table below :
11073
11074 +---------------------+-----------------+
11075 | Sample or converter | Default |
11076 | output type | matching method |
11077 +---------------------+-----------------+
11078 | boolean | bool |
11079 +---------------------+-----------------+
11080 | integer | int |
11081 +---------------------+-----------------+
11082 | ip | ip |
11083 +---------------------+-----------------+
11084 | string | str |
11085 +---------------------+-----------------+
11086 | binary | none, use "-m" |
11087 +---------------------+-----------------+
11088
11089Note that in order to match a binary samples, it is mandatory to specify a
11090matching method, see below.
11091
Willy Tarreau74ca5042013-06-11 23:12:07 +020011092The ACL engine can match these types against patterns of the following types :
11093 - boolean
11094 - integer or integer range
11095 - IP address / network
11096 - string (exact, substring, suffix, prefix, subdir, domain)
11097 - regular expression
11098 - hex block
11099
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011100The following ACL flags are currently supported :
11101
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011102 -i : ignore case during matching of all subsequent patterns.
11103 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011104 -m : use a specific pattern matching method
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010011105 -n : forbid the DNS resolutions
Thierry FOURNIER9860c412014-01-29 14:23:29 +010011106 -M : load the file pointed by -f like a map file.
Thierry FOURNIER3534d882014-01-20 17:01:44 +010011107 -u : force the unique id of the ACL
Willy Tarreau6a06a402007-07-15 20:15:28 +020011108 -- : force end of flags. Useful when a string looks like one of the flags.
11109
Willy Tarreau74ca5042013-06-11 23:12:07 +020011110The "-f" flag is followed by the name of a file from which all lines will be
11111read as individual values. It is even possible to pass multiple "-f" arguments
11112if the patterns are to be loaded from multiple files. Empty lines as well as
11113lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
11114will be stripped. If it is absolutely necessary to insert a valid pattern
11115beginning with a sharp, just prefix it with a space so that it is not taken for
11116a comment. Depending on the data type and match method, haproxy may load the
11117lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
11118exact string matching. In this case, duplicates will automatically be removed.
11119
Thierry FOURNIER9860c412014-01-29 14:23:29 +010011120The "-M" flag allows an ACL to use a map file. If this flag is set, the file is
11121parsed as two column file. The first column contains the patterns used by the
11122ACL, and the second column contain the samples. The sample can be used later by
11123a map. This can be useful in some rare cases where an ACL would just be used to
11124check for the existence of a pattern in a map before a mapping is applied.
11125
Thierry FOURNIER3534d882014-01-20 17:01:44 +010011126The "-u" flag forces the unique id of the ACL. This unique id is used with the
11127socket interface to identify ACL and dynamically change its values. Note that a
11128file is always identified by its name even if an id is set.
11129
Willy Tarreau74ca5042013-06-11 23:12:07 +020011130Also, note that the "-i" flag applies to subsequent entries and not to entries
11131loaded from files preceding it. For instance :
11132
11133 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
11134
11135In this example, each line of "exact-ua.lst" will be exactly matched against
11136the "user-agent" header of the request. Then each line of "generic-ua" will be
11137case-insensitively matched. Then the word "test" will be insensitively matched
11138as well.
11139
11140The "-m" flag is used to select a specific pattern matching method on the input
11141sample. All ACL-specific criteria imply a pattern matching method and generally
11142do not need this flag. However, this flag is useful with generic sample fetch
11143methods to describe how they're going to be matched against the patterns. This
11144is required for sample fetches which return data type for which there is no
11145obvious matching method (eg: string or binary). When "-m" is specified and
11146followed by a pattern matching method name, this method is used instead of the
11147default one for the criterion. This makes it possible to match contents in ways
11148that were not initially planned, or with sample fetch methods which return a
11149string. The matching method also affects the way the patterns are parsed.
11150
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010011151The "-n" flag forbids the dns resolutions. It is used with the load of ip files.
11152By default, if the parser cannot parse ip address it considers that the parsed
11153string is maybe a domain name and try dns resolution. The flag "-n" disable this
11154resolution. It is useful for detecting malformed ip lists. Note that if the DNS
11155server is not reachable, the haproxy configuration parsing may last many minutes
11156waiting fir the timeout. During this time no error messages are displayed. The
11157flag "-n" disable this behavior. Note also that during the runtime, this
11158function is disabled for the dynamic acl modifications.
11159
Willy Tarreau74ca5042013-06-11 23:12:07 +020011160There are some restrictions however. Not all methods can be used with all
11161sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
11162be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +020011163
11164 - "found" : only check if the requested sample could be found in the stream,
11165 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +020011166 to pass any pattern to avoid confusion. This matching method is
11167 particularly useful to detect presence of certain contents such
11168 as headers, cookies, etc... even if they are empty and without
11169 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011170
11171 - "bool" : check the value as a boolean. It can only be applied to fetches
11172 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011173 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011174
11175 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +020011176 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011177
11178 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +020011179 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011180
11181 - "bin" : match the contents against an hexadecimal string representing a
11182 binary sequence. This may be used with binary or string samples.
11183
11184 - "len" : match the sample's length as an integer. This may be used with
11185 binary or string samples.
11186
Willy Tarreau74ca5042013-06-11 23:12:07 +020011187 - "str" : exact match : match the contents against a string. This may be
11188 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011189
Willy Tarreau74ca5042013-06-11 23:12:07 +020011190 - "sub" : substring match : check that the contents contain at least one of
11191 the provided string patterns. This may be used with binary or
11192 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011193
Willy Tarreau74ca5042013-06-11 23:12:07 +020011194 - "reg" : regex match : match the contents against a list of regular
11195 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011196
Willy Tarreau74ca5042013-06-11 23:12:07 +020011197 - "beg" : prefix match : check that the contents begin like the provided
11198 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011199
Willy Tarreau74ca5042013-06-11 23:12:07 +020011200 - "end" : suffix match : check that the contents end like the provided
11201 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011202
Willy Tarreau74ca5042013-06-11 23:12:07 +020011203 - "dir" : subdir match : check that a slash-delimited portion of the
11204 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011205 This may be used with binary or string samples.
11206
Willy Tarreau74ca5042013-06-11 23:12:07 +020011207 - "dom" : domain match : check that a dot-delimited portion of the contents
11208 exactly match one of the provided string patterns. This may be
11209 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020011210
11211For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
11212request, it is possible to do :
11213
11214 acl jsess_present cook(JSESSIONID) -m found
11215
11216In order to apply a regular expression on the 500 first bytes of data in the
11217buffer, one would use the following acl :
11218
11219 acl script_tag payload(0,500) -m reg -i <script>
11220
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011221On systems where the regex library is much slower when using "-i", it is
11222possible to convert the sample to lowercase before matching, like this :
11223
11224 acl script_tag payload(0,500),lower -m reg <script>
11225
Willy Tarreau74ca5042013-06-11 23:12:07 +020011226All ACL-specific criteria imply a default matching method. Most often, these
11227criteria are composed by concatenating the name of the original sample fetch
11228method and the matching method. For example, "hdr_beg" applies the "beg" match
11229to samples retrieved using the "hdr" fetch method. Since all ACL-specific
11230criteria rely on a sample fetch method, it is always possible instead to use
11231the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011232
Willy Tarreau74ca5042013-06-11 23:12:07 +020011233If an alternate match is specified using "-m" on an ACL-specific criterion,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011234the matching method is simply applied to the underlying sample fetch method.
11235For example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011236
Willy Tarreau74ca5042013-06-11 23:12:07 +020011237 acl short_form hdr_beg(host) www.
11238 acl alternate1 hdr_beg(host) -m beg www.
11239 acl alternate2 hdr_dom(host) -m beg www.
11240 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011241
Willy Tarreau2b5285d2010-05-09 23:45:24 +020011242
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011243The table below summarizes the compatibility matrix between sample or converter
11244types and the pattern types to fetch against. It indicates for each compatible
11245combination the name of the matching method to be used, surrounded with angle
11246brackets ">" and "<" when the method is the default one and will work by
11247default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +010011248
Willy Tarreau74ca5042013-06-11 23:12:07 +020011249 +-------------------------------------------------+
11250 | Input sample type |
11251 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011252 | pattern type | boolean | integer | ip | string | binary |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011253 +----------------------+---------+---------+---------+---------+---------+
11254 | none (presence only) | found | found | found | found | found |
11255 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011256 | none (boolean value) |> bool <| bool | | bool | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011257 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011258 | integer (value) | int |> int <| int | int | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011259 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011260 | integer (length) | len | len | len | len | len |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011261 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011262 | IP address | | |> ip <| ip | ip |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011263 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020011264 | exact string | str | str | str |> str <| str |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011265 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011266 | prefix | beg | beg | beg | beg | beg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011267 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011268 | suffix | end | end | end | end | end |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011269 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011270 | substring | sub | sub | sub | sub | sub |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011271 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011272 | subdir | dir | dir | dir | dir | dir |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011273 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011274 | domain | dom | dom | dom | dom | dom |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011275 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010011276 | regex | reg | reg | reg | reg | reg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020011277 +----------------------+---------+---------+---------+---------+---------+
11278 | hex block | | | | bin | bin |
11279 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +020011280
11281
Willy Tarreau74ca5042013-06-11 23:12:07 +0200112827.1.1. Matching booleans
11283------------------------
11284
11285In order to match a boolean, no value is needed and all values are ignored.
11286Boolean matching is used by default for all fetch methods of type "boolean".
11287When boolean matching is used, the fetched value is returned as-is, which means
11288that a boolean "true" will always match and a boolean "false" will never match.
11289
11290Boolean matching may also be enforced using "-m bool" on fetch methods which
11291return an integer value. Then, integer value 0 is converted to the boolean
11292"false" and all other values are converted to "true".
11293
Willy Tarreau6a06a402007-07-15 20:15:28 +020011294
Willy Tarreau74ca5042013-06-11 23:12:07 +0200112957.1.2. Matching integers
11296------------------------
11297
11298Integer matching applies by default to integer fetch methods. It can also be
11299enforced on boolean fetches using "-m int". In this case, "false" is converted
11300to the integer 0, and "true" is converted to the integer 1.
11301
11302Integer matching also supports integer ranges and operators. Note that integer
11303matching only applies to positive values. A range is a value expressed with a
11304lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011305
11306For instance, "1024:65535" is a valid range to represent a range of
11307unprivileged ports, and "1024:" would also work. "0:1023" is a valid
11308representation of privileged ports, and ":1023" would also work.
11309
Willy Tarreau62644772008-07-16 18:36:06 +020011310As a special case, some ACL functions support decimal numbers which are in fact
11311two integers separated by a dot. This is used with some version checks for
11312instance. All integer properties apply to those decimal numbers, including
11313ranges and operators.
11314
Willy Tarreau6a06a402007-07-15 20:15:28 +020011315For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +010011316operators with ranges does not make much sense and is strongly discouraged.
11317Similarly, it does not make much sense to perform order comparisons with a set
11318of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011319
Willy Tarreau0ba27502007-12-24 16:55:16 +010011320Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +020011321
11322 eq : true if the tested value equals at least one value
11323 ge : true if the tested value is greater than or equal to at least one value
11324 gt : true if the tested value is greater than at least one value
11325 le : true if the tested value is less than or equal to at least one value
11326 lt : true if the tested value is less than at least one value
11327
Willy Tarreau0ba27502007-12-24 16:55:16 +010011328For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +020011329
11330 acl negative-length hdr_val(content-length) lt 0
11331
Willy Tarreau62644772008-07-16 18:36:06 +020011332This one matches SSL versions between 3.0 and 3.1 (inclusive) :
11333
11334 acl sslv3 req_ssl_ver 3:3.1
11335
Willy Tarreau6a06a402007-07-15 20:15:28 +020011336
Willy Tarreau74ca5042013-06-11 23:12:07 +0200113377.1.3. Matching strings
11338-----------------------
11339
11340String matching applies to string or binary fetch methods, and exists in 6
11341different forms :
11342
11343 - exact match (-m str) : the extracted string must exactly match the
11344 patterns ;
11345
11346 - substring match (-m sub) : the patterns are looked up inside the
11347 extracted string, and the ACL matches if any of them is found inside ;
11348
11349 - prefix match (-m beg) : the patterns are compared with the beginning of
11350 the extracted string, and the ACL matches if any of them matches.
11351
11352 - suffix match (-m end) : the patterns are compared with the end of the
11353 extracted string, and the ACL matches if any of them matches.
11354
11355 - subdir match (-m sub) : the patterns are looked up inside the extracted
11356 string, delimited with slashes ("/"), and the ACL matches if any of them
11357 matches.
11358
11359 - domain match (-m dom) : the patterns are looked up inside the extracted
11360 string, delimited with dots ("."), and the ACL matches if any of them
11361 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011362
11363String matching applies to verbatim strings as they are passed, with the
11364exception of the backslash ("\") which makes it possible to escape some
11365characters such as the space. If the "-i" flag is passed before the first
11366string, then the matching will be performed ignoring the case. In order
11367to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +010011368before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020011369
11370
Willy Tarreau74ca5042013-06-11 23:12:07 +0200113717.1.4. Matching regular expressions (regexes)
11372---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020011373
11374Just like with string matching, regex matching applies to verbatim strings as
11375they are passed, with the exception of the backslash ("\") which makes it
11376possible to escape some characters such as the space. If the "-i" flag is
11377passed before the first regex, then the matching will be performed ignoring
11378the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +010011379the "--" flag before the first string. Same principle applies of course to
11380match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020011381
11382
Willy Tarreau74ca5042013-06-11 23:12:07 +0200113837.1.5. Matching arbitrary data blocks
11384-------------------------------------
11385
11386It is possible to match some extracted samples against a binary block which may
11387not safely be represented as a string. For this, the patterns must be passed as
11388a series of hexadecimal digits in an even number, when the match method is set
11389to binary. Each sequence of two digits will represent a byte. The hexadecimal
11390digits may be used upper or lower case.
11391
11392Example :
11393 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
11394 acl hello payload(0,6) -m bin 48656c6c6f0a
11395
11396
113977.1.6. Matching IPv4 and IPv6 addresses
11398---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020011399
11400IPv4 addresses values can be specified either as plain addresses or with a
11401netmask appended, in which case the IPv4 address matches whenever it is
11402within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010011403host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +010011404difficult to read and debug configurations. If hostnames are used, you should
11405at least ensure that they are present in /etc/hosts so that the configuration
11406does not depend on any random DNS match at the moment the configuration is
11407parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011408
Willy Tarreauceb4ac92012-04-28 00:41:46 +020011409IPv6 may be entered in their usual form, with or without a netmask appended.
11410Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
11411trouble with randomly resolved IP addresses, host names are never allowed in
11412IPv6 patterns.
11413
11414HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
11415following situations :
11416 - tested address is IPv4, pattern address is IPv4, the match applies
11417 in IPv4 using the supplied mask if any.
11418 - tested address is IPv6, pattern address is IPv6, the match applies
11419 in IPv6 using the supplied mask if any.
11420 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
11421 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
11422 ::IPV4 or ::ffff:IPV4, otherwise it fails.
11423 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
11424 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
11425 applied in IPv6 using the supplied IPv6 mask.
11426
Willy Tarreau74ca5042013-06-11 23:12:07 +020011427
114287.2. Using ACLs to form conditions
11429----------------------------------
11430
11431Some actions are only performed upon a valid condition. A condition is a
11432combination of ACLs with operators. 3 operators are supported :
11433
11434 - AND (implicit)
11435 - OR (explicit with the "or" keyword or the "||" operator)
11436 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +020011437
Willy Tarreau74ca5042013-06-11 23:12:07 +020011438A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +020011439
Willy Tarreau74ca5042013-06-11 23:12:07 +020011440 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +020011441
Willy Tarreau74ca5042013-06-11 23:12:07 +020011442Such conditions are generally used after an "if" or "unless" statement,
11443indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +020011444
Willy Tarreau74ca5042013-06-11 23:12:07 +020011445For instance, to block HTTP requests to the "*" URL with methods other than
11446"OPTIONS", as well as POST requests without content-length, and GET or HEAD
11447requests with a content-length greater than 0, and finally every request which
11448is not either GET/HEAD/POST/OPTIONS !
11449
11450 acl missing_cl hdr_cnt(Content-length) eq 0
11451 block if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
11452 block if METH_GET HTTP_CONTENT
11453 block unless METH_GET or METH_POST or METH_OPTIONS
11454
11455To select a different backend for requests to static contents on the "www" site
11456and to every request on the "img", "video", "download" and "ftp" hosts :
11457
11458 acl url_static path_beg /static /images /img /css
11459 acl url_static path_end .gif .png .jpg .css .js
11460 acl host_www hdr_beg(host) -i www
11461 acl host_static hdr_beg(host) -i img. video. download. ftp.
11462
11463 # now use backend "static" for all static-only hosts, and for static urls
11464 # of host "www". Use backend "www" for the rest.
11465 use_backend static if host_static or host_www url_static
11466 use_backend www if host_www
11467
11468It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
11469expressions that are built on the fly without needing to be declared. They must
11470be enclosed between braces, with a space before and after each brace (because
11471the braces must be seen as independent words). Example :
11472
11473 The following rule :
11474
11475 acl missing_cl hdr_cnt(Content-length) eq 0
11476 block if METH_POST missing_cl
11477
11478 Can also be written that way :
11479
11480 block if METH_POST { hdr_cnt(Content-length) eq 0 }
11481
11482It is generally not recommended to use this construct because it's a lot easier
11483to leave errors in the configuration when written that way. However, for very
11484simple rules matching only one source IP address for instance, it can make more
11485sense to use them than to declare ACLs with random names. Another example of
11486good use is the following :
11487
11488 With named ACLs :
11489
11490 acl site_dead nbsrv(dynamic) lt 2
11491 acl site_dead nbsrv(static) lt 2
11492 monitor fail if site_dead
11493
11494 With anonymous ACLs :
11495
11496 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
11497
11498See section 4.2 for detailed help on the "block" and "use_backend" keywords.
11499
11500
115017.3. Fetching samples
11502---------------------
11503
11504Historically, sample fetch methods were only used to retrieve data to match
11505against patterns using ACLs. With the arrival of stick-tables, a new class of
11506sample fetch methods was created, most often sharing the same syntax as their
11507ACL counterpart. These sample fetch methods are also known as "fetches". As
11508of now, ACLs and fetches have converged. All ACL fetch methods have been made
11509available as fetch methods, and ACLs may use any sample fetch method as well.
11510
11511This section details all available sample fetch methods and their output type.
11512Some sample fetch methods have deprecated aliases that are used to maintain
11513compatibility with existing configurations. They are then explicitly marked as
11514deprecated and should not be used in new setups.
11515
11516The ACL derivatives are also indicated when available, with their respective
11517matching methods. These ones all have a well defined default pattern matching
11518method, so it is never necessary (though allowed) to pass the "-m" option to
11519indicate how the sample will be matched using ACLs.
11520
11521As indicated in the sample type versus matching compatibility matrix above,
11522when using a generic sample fetch method in an ACL, the "-m" option is
11523mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
11524the same keyword exists as an ACL keyword and as a standard fetch method, the
11525ACL engine will automatically pick the ACL-only one by default.
11526
11527Some of these keywords support one or multiple mandatory arguments, and one or
11528multiple optional arguments. These arguments are strongly typed and are checked
11529when the configuration is parsed so that there is no risk of running with an
11530incorrect argument (eg: an unresolved backend name). Fetch function arguments
11531are passed between parenthesis and are delimited by commas. When an argument
11532is optional, it will be indicated below between square brackets ('[ ]'). When
11533all arguments are optional, the parenthesis may be omitted.
11534
11535Thus, the syntax of a standard sample fetch method is one of the following :
11536 - name
11537 - name(arg1)
11538 - name(arg1,arg2)
11539
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011540
115417.3.1. Converters
11542-----------------
11543
Willy Tarreaue6b11e42013-11-26 19:02:32 +010011544Sample fetch methods may be combined with transformations to be applied on top
11545of the fetched sample (also called "converters"). These combinations form what
11546is called "sample expressions" and the result is a "sample". Initially this
11547was only supported by "stick on" and "stick store-request" directives but this
11548has now be extended to all places where samples may be used (acls, log-format,
11549unique-id-format, add-header, ...).
11550
11551These transformations are enumerated as a series of specific keywords after the
11552sample fetch method. These keywords may equally be appended immediately after
11553the fetch keyword's argument, delimited by a comma. These keywords can also
11554support some arguments (eg: a netmask) which must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +010011555
Willy Tarreau97707872015-01-27 15:12:13 +010011556A certain category of converters are bitwise and arithmetic operators which
11557support performing basic operations on integers. Some bitwise operations are
11558supported (and, or, xor, cpl) and some arithmetic operations are supported
11559(add, sub, mul, div, mod, neg). Some comparators are provided (odd, even, not,
11560bool) which make it possible to report a match without having to write an ACL.
11561
Willy Tarreau74ca5042013-06-11 23:12:07 +020011562The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +010011563
Willy Tarreau97707872015-01-27 15:12:13 +010011564add(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011565 Adds <value> to the input value of type signed integer, and returns the
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011566 result as a signed integer. <value> can be a numeric value or a variable
11567 name. The name of the variable starts by an indication about its scope. The
11568 allowed scopes are:
11569 "sess" : the variable is shared with all the session,
11570 "txn" : the variable is shared with all the transaction (request and
11571 response),
11572 "req" : the variable is shared only during the request processing,
11573 "res" : the variable is shared only during the response processing.
11574 This prefix is followed by a name. The separator is a '.'. The name may only
11575 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011576
11577and(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011578 Performs a bitwise "AND" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011579 integer, and returns the result as an signed integer. <value> can be a
11580 numeric value or a variable name. The name of the variable starts by an
11581 indication about its scope. The allowed scopes are:
11582 "sess" : the variable is shared with all the session,
11583 "txn" : the variable is shared with all the transaction (request and
11584 response),
11585 "req" : the variable is shared only during the request processing,
11586 "res" : the variable is shared only during the response processing.
11587 This prefix is followed by a name. The separator is a '.'. The name may only
11588 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011589
Emeric Brun53d1a982014-04-30 18:21:37 +020011590base64
11591 Converts a binary input sample to a base64 string. It is used to log or
11592 transfer binary content in a way that can be reliably transferred (eg:
11593 an SSL ID can be copied in a header).
11594
Willy Tarreau97707872015-01-27 15:12:13 +010011595bool
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011596 Returns a boolean TRUE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010011597 non-null, otherwise returns FALSE. Used in conjunction with and(), it can be
11598 used to report true/false for bit testing on input values (eg: verify the
11599 presence of a flag).
11600
Emeric Brun54c4ac82014-11-03 15:32:43 +010011601bytes(<offset>[,<length>])
11602 Extracts some bytes from an input binary sample. The result is a binary
11603 sample starting at an offset (in bytes) of the original sample and
11604 optionnaly truncated at the given length.
11605
Willy Tarreau97707872015-01-27 15:12:13 +010011606cpl
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011607 Takes the input value of type signed integer, applies a ones-complement
11608 (flips all bits) and returns the result as an signed integer.
Willy Tarreau97707872015-01-27 15:12:13 +010011609
Willy Tarreau80599772015-01-20 19:35:24 +010011610crc32([<avalanche>])
11611 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32
11612 hash function. Optionally, it is possible to apply a full avalanche hash
11613 function to the output if the optional <avalanche> argument equals 1. This
11614 converter uses the same functions as used by the various hash-based load
11615 balancing algorithms, so it will provide exactly the same results. It is
11616 provided for compatibility with other software which want a CRC32 to be
11617 computed on some input keys, so it follows the most common implementation as
11618 found in Ethernet, Gzip, PNG, etc... It is slower than the other algorithms
11619 but may provide a better or at least less predictable distribution. It must
11620 not be used for security purposes as a 32-bit hash is trivial to break. See
11621 also "djb2", "sdbm", "wt6" and the "hash-type" directive.
11622
David Carlier29b3ca32015-09-25 14:09:21 +010011623da-csv-conv(<prop>[,<prop>*])
David Carlier4542b102015-06-01 13:54:29 +020011624 Asks the DeviceAtlas converter to identify the User Agent string passed on
11625 input, and to emit a string made of the concatenation of the properties
11626 enumerated in argument, delimited by the separator defined by the global
11627 keyword "deviceatlas-property-separator", or by default the pipe character
11628 ('|'). There's a limit of 5 different properties imposed by the haproxy
11629 configuration language.
11630
11631 Example:
11632 frontend www
Cyril Bonté307ee1e2015-09-28 23:16:06 +020011633 bind *:8881
11634 default_backend servers
11635 http-request set-header X-DeviceAtlas-Data %[req.fhdr(User-Agent),da-csv(primaryHardwareType,osName,osVersion,browserName,browserVersion)]
David Carlier4542b102015-06-01 13:54:29 +020011636
Thierry FOURNIER9687c772015-05-07 15:46:29 +020011637debug
11638 This converter is used as debug tool. It dumps on screen the content and the
11639 type of the input sample. The sample is returned as is on its output. This
11640 converter only exists when haproxy was built with debugging enabled.
11641
Willy Tarreau97707872015-01-27 15:12:13 +010011642div(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011643 Divides the input value of type signed integer by <value>, and returns the
11644 result as an signed integer. If <value> is null, the largest unsigned
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011645 integer is returned (typically 2^63-1). <value> can be a numeric value or a
11646 variable name. The name of the variable starts by an indication about it
11647 scope. The scope allowed are:
11648 "sess" : the variable is shared with all the session,
11649 "txn" : the variable is shared with all the transaction (request and
11650 response),
11651 "req" : the variable is shared only during the request processing,
11652 "res" : the variable is shared only during the response processing.
11653 This prefix is followed by a name. The separator is a '.'. The name may only
11654 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011655
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020011656djb2([<avalanche>])
11657 Hashes a binary input sample into an unsigned 32-bit quantity using the DJB2
11658 hash function. Optionally, it is possible to apply a full avalanche hash
11659 function to the output if the optional <avalanche> argument equals 1. This
11660 converter uses the same functions as used by the various hash-based load
11661 balancing algorithms, so it will provide exactly the same results. It is
11662 mostly intended for debugging, but can be used as a stick-table entry to
11663 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010011664 32-bit hash is trivial to break. See also "crc32", "sdbm", "wt6" and the
11665 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020011666
Willy Tarreau97707872015-01-27 15:12:13 +010011667even
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011668 Returns a boolean TRUE if the input value of type signed integer is even
Willy Tarreau97707872015-01-27 15:12:13 +010011669 otherwise returns FALSE. It is functionally equivalent to "not,and(1),bool".
11670
Emeric Brunf399b0d2014-11-03 17:07:03 +010011671field(<index>,<delimiters>)
11672 Extracts the substring at the given index considering given delimiters from
11673 an input string. Indexes start at 1 and delimiters are a string formatted
11674 list of chars.
11675
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011676hex
11677 Converts a binary input sample to an hex string containing two hex digits per
11678 input byte. It is used to log or transfer hex dumps of some binary input data
11679 in a way that can be reliably transferred (eg: an SSL ID can be copied in a
11680 header).
Thierry FOURNIER2f49d6d2014-03-12 15:01:52 +010011681
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011682http_date([<offset>])
11683 Converts an integer supposed to contain a date since epoch to a string
11684 representing this date in a format suitable for use in HTTP header fields. If
11685 an offset value is specified, then it is a number of seconds that is added to
11686 the date before the conversion is operated. This is particularly useful to
11687 emit Date header fields, Expires values in responses when combined with a
11688 positive offset, or Last-Modified values when the offset is negative.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011689
Willy Tarreaud9f316a2014-07-10 14:03:38 +020011690in_table(<table>)
11691 Uses the string representation of the input sample to perform a look up in
11692 the specified table. If the key is not found in the table, a boolean false
11693 is returned. Otherwise a boolean true is returned. This can be used to verify
11694 the presence of a certain key in a table tracking some elements (eg: whether
11695 or not a source IP address or an Authorization header was already seen).
11696
Willy Tarreauffcb2e42014-07-10 16:29:08 +020011697ipmask(<mask>)
11698 Apply a mask to an IPv4 address, and use the result for lookups and storage.
11699 This can be used to make all hosts within a certain mask to share the same
11700 table entries and as such use the same server. The mask can be passed in
11701 dotted form (eg: 255.255.255.0) or in CIDR form (eg: 24).
11702
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020011703json([<input-code>])
11704 Escapes the input string and produces an ASCII ouput string ready to use as a
11705 JSON string. The converter tries to decode the input string according to the
11706 <input-code> parameter. It can be "ascii", "utf8", "utf8s", "utf8"" or
11707 "utf8ps". The "ascii" decoder never fails. The "utf8" decoder detects 3 types
11708 of errors:
11709 - bad UTF-8 sequence (lone continuation byte, bad number of continuation
11710 bytes, ...)
11711 - invalid range (the decoded value is within a UTF-8 prohibited range),
11712 - code overlong (the value is encoded with more bytes than necessary).
11713
11714 The UTF-8 JSON encoding can produce a "too long value" error when the UTF-8
11715 character is greater than 0xffff because the JSON string escape specification
11716 only authorizes 4 hex digits for the value encoding. The UTF-8 decoder exists
11717 in 4 variants designated by a combination of two suffix letters : "p" for
11718 "permissive" and "s" for "silently ignore". The behaviors of the decoders
11719 are :
11720 - "ascii" : never fails ;
11721 - "utf8" : fails on any detected errors ;
11722 - "utf8s" : never fails, but removes characters corresponding to errors ;
11723 - "utf8p" : accepts and fixes the overlong errors, but fails on any other
11724 error ;
11725 - "utf8ps" : never fails, accepts and fixes the overlong errors, but removes
11726 characters corresponding to the other errors.
11727
11728 This converter is particularly useful for building properly escaped JSON for
11729 logging to servers which consume JSON-formated traffic logs.
11730
11731 Example:
11732 capture request header user-agent len 150
11733 capture request header Host len 15
11734 log-format {"ip":"%[src]","user-agent":"%[capture.req.hdr(1),json]"}
11735
11736 Input request from client 127.0.0.1:
11737 GET / HTTP/1.0
11738 User-Agent: Very "Ugly" UA 1/2
11739
11740 Output log:
11741 {"ip":"127.0.0.1","user-agent":"Very \"Ugly\" UA 1\/2"}
11742
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011743language(<value>[,<default>])
11744 Returns the value with the highest q-factor from a list as extracted from the
11745 "accept-language" header using "req.fhdr". Values with no q-factor have a
11746 q-factor of 1. Values with a q-factor of 0 are dropped. Only values which
11747 belong to the list of semi-colon delimited <values> will be considered. The
11748 argument <value> syntax is "lang[;lang[;lang[;...]]]". If no value matches the
11749 given list and a default value is provided, it is returned. Note that language
11750 names may have a variant after a dash ('-'). If this variant is present in the
11751 list, it will be matched, but if it is not, only the base language is checked.
11752 The match is case-sensitive, and the output string is always one of those
11753 provided in arguments. The ordering of arguments is meaningless, only the
11754 ordering of the values in the request counts, as the first value among
11755 multiple sharing the same q-factor is used.
Thierry FOURNIERad903512014-04-11 17:51:01 +020011756
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011757 Example :
Thierry FOURNIERad903512014-04-11 17:51:01 +020011758
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011759 # this configuration switches to the backend matching a
11760 # given language based on the request :
Thierry FOURNIERad903512014-04-11 17:51:01 +020011761
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011762 acl es req.fhdr(accept-language),language(es;fr;en) -m str es
11763 acl fr req.fhdr(accept-language),language(es;fr;en) -m str fr
11764 acl en req.fhdr(accept-language),language(es;fr;en) -m str en
11765 use_backend spanish if es
11766 use_backend french if fr
11767 use_backend english if en
11768 default_backend choose_your_language
Thierry FOURNIERad903512014-04-11 17:51:01 +020011769
Willy Tarreauffcb2e42014-07-10 16:29:08 +020011770lower
11771 Convert a string sample to lower case. This can only be placed after a string
11772 sample fetch function or after a transformation keyword returning a string
11773 type. The result is of type string.
11774
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020011775ltime(<format>[,<offset>])
11776 Converts an integer supposed to contain a date since epoch to a string
11777 representing this date in local time using a format defined by the <format>
11778 string using strftime(3). The purpose is to allow any date format to be used
11779 in logs. An optional <offset> in seconds may be applied to the input date
11780 (positive or negative). See the strftime() man page for the format supported
11781 by your operating system. See also the utime converter.
11782
11783 Example :
11784
11785 # Emit two colons, one with the local time and another with ip:port
11786 # Eg: 20140710162350 127.0.0.1:57325
11787 log-format %[date,ltime(%Y%m%d%H%M%S)]\ %ci:%cp
11788
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011789map(<map_file>[,<default_value>])
11790map_<match_type>(<map_file>[,<default_value>])
11791map_<match_type>_<output_type>(<map_file>[,<default_value>])
11792 Search the input value from <map_file> using the <match_type> matching method,
11793 and return the associated value converted to the type <output_type>. If the
11794 input value cannot be found in the <map_file>, the converter returns the
11795 <default_value>. If the <default_value> is not set, the converter fails and
11796 acts as if no input value could be fetched. If the <match_type> is not set, it
11797 defaults to "str". Likewise, if the <output_type> is not set, it defaults to
11798 "str". For convenience, the "map" keyword is an alias for "map_str" and maps a
11799 string to another string.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010011800
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011801 It is important to avoid overlapping between the keys : IP addresses and
11802 strings are stored in trees, so the first of the finest match will be used.
11803 Other keys are stored in lists, so the first matching occurrence will be used.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010011804
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011805 The following array contains the list of all map functions avalaible sorted by
11806 input type, match type and output type.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010011807
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011808 input type | match method | output type str | output type int | output type ip
11809 -----------+--------------+-----------------+-----------------+---------------
11810 str | str | map_str | map_str_int | map_str_ip
11811 -----------+--------------+-----------------+-----------------+---------------
Willy Tarreau787a4c02014-05-10 07:55:30 +020011812 str | beg | map_beg | map_beg_int | map_end_ip
11813 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011814 str | sub | map_sub | map_sub_int | map_sub_ip
11815 -----------+--------------+-----------------+-----------------+---------------
11816 str | dir | map_dir | map_dir_int | map_dir_ip
11817 -----------+--------------+-----------------+-----------------+---------------
11818 str | dom | map_dom | map_dom_int | map_dom_ip
11819 -----------+--------------+-----------------+-----------------+---------------
11820 str | end | map_end | map_end_int | map_end_ip
11821 -----------+--------------+-----------------+-----------------+---------------
11822 str | reg | map_reg | map_reg_int | map_reg_ip
11823 -----------+--------------+-----------------+-----------------+---------------
11824 int | int | map_int | map_int_int | map_int_ip
11825 -----------+--------------+-----------------+-----------------+---------------
11826 ip | ip | map_ip | map_ip_int | map_ip_ip
11827 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010011828
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011829 The file contains one key + value per line. Lines which start with '#' are
11830 ignored, just like empty lines. Leading tabs and spaces are stripped. The key
11831 is then the first "word" (series of non-space/tabs characters), and the value
11832 is what follows this series of space/tab till the end of the line excluding
11833 trailing spaces/tabs.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010011834
Thierry FOURNIER060762e2014-04-23 13:29:15 +020011835 Example :
11836
11837 # this is a comment and is ignored
11838 2.22.246.0/23 United Kingdom \n
11839 <-><-----------><--><------------><---->
11840 | | | | `- trailing spaces ignored
11841 | | | `---------- value
11842 | | `-------------------- middle spaces ignored
11843 | `---------------------------- key
11844 `------------------------------------ leading spaces ignored
11845
Willy Tarreau97707872015-01-27 15:12:13 +010011846mod(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011847 Divides the input value of type signed integer by <value>, and returns the
11848 remainder as an signed integer. If <value> is null, then zero is returned.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011849 <value> can be a numeric value or a variable name. The name of the variable
11850 starts by an indication about its scope. The allowed scopes are:
11851 "sess" : the variable is shared with all the session,
11852 "txn" : the variable is shared with all the transaction (request and
11853 response),
11854 "req" : the variable is shared only during the request processing,
11855 "res" : the variable is shared only during the response processing.
11856 This prefix is followed by a name. The separator is a '.'. The name may only
11857 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011858
11859mul(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011860 Multiplies the input value of type signed integer by <value>, and returns
Thierry FOURNIER00c005c2015-07-08 01:10:21 +020011861 the product as an signed integer. In case of overflow, the largest possible
11862 value for the sign is returned so that the operation doesn't wrap around.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011863 <value> can be a numeric value or a variable name. The name of the variable
11864 starts by an indication about its scope. The allowed scopes are:
11865 "sess" : the variable is shared with all the session,
11866 "txn" : the variable is shared with all the transaction (request and
11867 response),
11868 "req" : the variable is shared only during the request processing,
11869 "res" : the variable is shared only during the response processing.
11870 This prefix is followed by a name. The separator is a '.'. The name may only
11871 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011872
11873neg
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011874 Takes the input value of type signed integer, computes the opposite value,
11875 and returns the remainder as an signed integer. 0 is identity. This operator
11876 is provided for reversed subtracts : in order to subtract the input from a
11877 constant, simply perform a "neg,add(value)".
Willy Tarreau97707872015-01-27 15:12:13 +010011878
11879not
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011880 Returns a boolean FALSE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010011881 non-null, otherwise returns TRUE. Used in conjunction with and(), it can be
11882 used to report true/false for bit testing on input values (eg: verify the
11883 absence of a flag).
11884
11885odd
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011886 Returns a boolean TRUE if the input value of type signed integer is odd
Willy Tarreau97707872015-01-27 15:12:13 +010011887 otherwise returns FALSE. It is functionally equivalent to "and(1),bool".
11888
11889or(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011890 Performs a bitwise "OR" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011891 integer, and returns the result as an signed integer. <value> can be a
11892 numeric value or a variable name. The name of the variable starts by an
11893 indication about its scope. The allowed scopes are:
11894 "sess" : the variable is shared with all the session,
11895 "txn" : the variable is shared with all the transaction (request and
11896 response),
11897 "req" : the variable is shared only during the request processing,
11898 "res" : the variable is shared only during the response processing.
11899 This prefix is followed by a name. The separator is a '.'. The name may only
11900 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011901
Willy Tarreauc4dc3502015-01-23 20:39:28 +010011902regsub(<regex>,<subst>[,<flags>])
Willy Tarreau7eda8492015-01-20 19:47:06 +010011903 Applies a regex-based substitution to the input string. It does the same
11904 operation as the well-known "sed" utility with "s/<regex>/<subst>/". By
11905 default it will replace in the input string the first occurrence of the
11906 largest part matching the regular expression <regex> with the substitution
11907 string <subst>. It is possible to replace all occurrences instead by adding
11908 the flag "g" in the third argument <flags>. It is also possible to make the
11909 regex case insensitive by adding the flag "i" in <flags>. Since <flags> is a
11910 string, it is made up from the concatenation of all desired flags. Thus if
11911 both "i" and "g" are desired, using "gi" or "ig" will have the same effect.
11912 It is important to note that due to the current limitations of the
11913 configuration parser, some characters such as closing parenthesis or comma
11914 are not possible to use in the arguments. The first use of this converter is
11915 to replace certain characters or sequence of characters with other ones.
11916
11917 Example :
11918
11919 # de-duplicate "/" in header "x-path".
11920 # input: x-path: /////a///b/c/xzxyz/
11921 # output: x-path: /a/b/c/xzxyz/
11922 http-request set-header x-path %[hdr(x-path),regsub(/+,/,g)]
11923
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020011924capture-req(<id>)
11925 Capture the string entry in the request slot <id> and returns the entry as
11926 is. If the slot doesn't exist, the capture fails silently.
11927
11928 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020011929 "http-response capture", "capture.req.hdr" and
11930 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020011931
11932capture-res(<id>)
11933 Capture the string entry in the response slot <id> and returns the entry as
11934 is. If the slot doesn't exist, the capture fails silently.
11935
11936 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020011937 "http-response capture", "capture.req.hdr" and
11938 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020011939
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020011940sdbm([<avalanche>])
11941 Hashes a binary input sample into an unsigned 32-bit quantity using the SDBM
11942 hash function. Optionally, it is possible to apply a full avalanche hash
11943 function to the output if the optional <avalanche> argument equals 1. This
11944 converter uses the same functions as used by the various hash-based load
11945 balancing algorithms, so it will provide exactly the same results. It is
11946 mostly intended for debugging, but can be used as a stick-table entry to
11947 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010011948 32-bit hash is trivial to break. See also "crc32", "djb2", "wt6" and the
11949 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020011950
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020011951set-var(<var name>)
11952 Sets a variable with the input content and return the content on the output as
11953 is. The variable keep the value and the associated input type. The name of the
11954 variable starts by an indication about it scope. The scope allowed are:
11955 "sess" : the variable is shared with all the session,
11956 "txn" : the variable is shared with all the transaction (request and
11957 response),
11958 "req" : the variable is shared only during the request processing,
11959 "res" : the variable is shared only during the response processing.
11960 This prefix is followed by a name. The separator is a '.'. The name may only
11961 contain characters 'a-z', 'A-Z', '0-9' and '_'.
11962
Willy Tarreau97707872015-01-27 15:12:13 +010011963sub(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020011964 Subtracts <value> from the input value of type signed integer, and returns
11965 the result as an signed integer. Note: in order to subtract the input from
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020011966 a constant, simply perform a "neg,add(value)". <value> can be a numeric value
11967 or a variable name. The name of the variable starts by an indication about its
11968 scope. The allowed scopes are:
11969 "sess" : the variable is shared with all the session,
11970 "txn" : the variable is shared with all the transaction (request and
11971 response),
11972 "req" : the variable is shared only during the request processing,
11973 "res" : the variable is shared only during the response processing.
11974 This prefix is followed by a name. The separator is a '.'. The name may only
11975 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010011976
Willy Tarreaud9f316a2014-07-10 14:03:38 +020011977table_bytes_in_rate(<table>)
11978 Uses the string representation of the input sample to perform a look up in
11979 the specified table. If the key is not found in the table, integer value zero
11980 is returned. Otherwise the converter returns the average client-to-server
11981 bytes rate associated with the input sample in the designated table, measured
11982 in amount of bytes over the period configured in the table. See also the
11983 sc_bytes_in_rate sample fetch keyword.
11984
11985
11986table_bytes_out_rate(<table>)
11987 Uses the string representation of the input sample to perform a look up in
11988 the specified table. If the key is not found in the table, integer value zero
11989 is returned. Otherwise the converter returns the average server-to-client
11990 bytes rate associated with the input sample in the designated table, measured
11991 in amount of bytes over the period configured in the table. See also the
11992 sc_bytes_out_rate sample fetch keyword.
11993
11994table_conn_cnt(<table>)
11995 Uses the string representation of the input sample to perform a look up in
11996 the specified table. If the key is not found in the table, integer value zero
11997 is returned. Otherwise the converter returns the cumulated amount of incoming
11998 connections associated with the input sample in the designated table. See
11999 also the sc_conn_cnt sample fetch keyword.
12000
12001table_conn_cur(<table>)
12002 Uses the string representation of the input sample to perform a look up in
12003 the specified table. If the key is not found in the table, integer value zero
12004 is returned. Otherwise the converter returns the current amount of concurrent
12005 tracked connections associated with the input sample in the designated table.
12006 See also the sc_conn_cur sample fetch keyword.
12007
12008table_conn_rate(<table>)
12009 Uses the string representation of the input sample to perform a look up in
12010 the specified table. If the key is not found in the table, integer value zero
12011 is returned. Otherwise the converter returns the average incoming connection
12012 rate associated with the input sample in the designated table. See also the
12013 sc_conn_rate sample fetch keyword.
12014
Thierry FOURNIER236657b2015-08-19 08:25:14 +020012015table_gpt0(<table>)
12016 Uses the string representation of the input sample to perform a look up in
12017 the specified table. If the key is not found in the table, boolean value zero
12018 is returned. Otherwise the converter returns the current value of the first
12019 general purpose tag associated with the input sample in the designated table.
12020 See also the sc_get_gpt0 sample fetch keyword.
12021
Willy Tarreaud9f316a2014-07-10 14:03:38 +020012022table_gpc0(<table>)
12023 Uses the string representation of the input sample to perform a look up in
12024 the specified table. If the key is not found in the table, integer value zero
12025 is returned. Otherwise the converter returns the current value of the first
12026 general purpose counter associated with the input sample in the designated
12027 table. See also the sc_get_gpc0 sample fetch keyword.
12028
12029table_gpc0_rate(<table>)
12030 Uses the string representation of the input sample to perform a look up in
12031 the specified table. If the key is not found in the table, integer value zero
12032 is returned. Otherwise the converter returns the frequency which the gpc0
12033 counter was incremented over the configured period in the table, associated
12034 with the input sample in the designated table. See also the sc_get_gpc0_rate
12035 sample fetch keyword.
12036
12037table_http_err_cnt(<table>)
12038 Uses the string representation of the input sample to perform a look up in
12039 the specified table. If the key is not found in the table, integer value zero
12040 is returned. Otherwise the converter returns the cumulated amount of HTTP
12041 errors associated with the input sample in the designated table. See also the
12042 sc_http_err_cnt sample fetch keyword.
12043
12044table_http_err_rate(<table>)
12045 Uses the string representation of the input sample to perform a look up in
12046 the specified table. If the key is not found in the table, integer value zero
12047 is returned. Otherwise the average rate of HTTP errors associated with the
12048 input sample in the designated table, measured in amount of errors over the
12049 period configured in the table. See also the sc_http_err_rate sample fetch
12050 keyword.
12051
12052table_http_req_cnt(<table>)
12053 Uses the string representation of the input sample to perform a look up in
12054 the specified table. If the key is not found in the table, integer value zero
12055 is returned. Otherwise the converter returns the cumulated amount of HTTP
12056 requests associated with the input sample in the designated table. See also
12057 the sc_http_req_cnt sample fetch keyword.
12058
12059table_http_req_rate(<table>)
12060 Uses the string representation of the input sample to perform a look up in
12061 the specified table. If the key is not found in the table, integer value zero
12062 is returned. Otherwise the average rate of HTTP requests associated with the
12063 input sample in the designated table, measured in amount of requests over the
12064 period configured in the table. See also the sc_http_req_rate sample fetch
12065 keyword.
12066
12067table_kbytes_in(<table>)
12068 Uses the string representation of the input sample to perform a look up in
12069 the specified table. If the key is not found in the table, integer value zero
12070 is returned. Otherwise the converter returns the cumulated amount of client-
12071 to-server data associated with the input sample in the designated table,
12072 measured in kilobytes. The test is currently performed on 32-bit integers,
12073 which limits values to 4 terabytes. See also the sc_kbytes_in sample fetch
12074 keyword.
12075
12076table_kbytes_out(<table>)
12077 Uses the string representation of the input sample to perform a look up in
12078 the specified table. If the key is not found in the table, integer value zero
12079 is returned. Otherwise the converter returns the cumulated amount of server-
12080 to-client data associated with the input sample in the designated table,
12081 measured in kilobytes. The test is currently performed on 32-bit integers,
12082 which limits values to 4 terabytes. See also the sc_kbytes_out sample fetch
12083 keyword.
12084
12085table_server_id(<table>)
12086 Uses the string representation of the input sample to perform a look up in
12087 the specified table. If the key is not found in the table, integer value zero
12088 is returned. Otherwise the converter returns the server ID associated with
12089 the input sample in the designated table. A server ID is associated to a
12090 sample by a "stick" rule when a connection to a server succeeds. A server ID
12091 zero means that no server is associated with this key.
12092
12093table_sess_cnt(<table>)
12094 Uses the string representation of the input sample to perform a look up in
12095 the specified table. If the key is not found in the table, integer value zero
12096 is returned. Otherwise the converter returns the cumulated amount of incoming
12097 sessions associated with the input sample in the designated table. Note that
12098 a session here refers to an incoming connection being accepted by the
12099 "tcp-request connection" rulesets. See also the sc_sess_cnt sample fetch
12100 keyword.
12101
12102table_sess_rate(<table>)
12103 Uses the string representation of the input sample to perform a look up in
12104 the specified table. If the key is not found in the table, integer value zero
12105 is returned. Otherwise the converter returns the average incoming session
12106 rate associated with the input sample in the designated table. Note that a
12107 session here refers to an incoming connection being accepted by the
12108 "tcp-request connection" rulesets. See also the sc_sess_rate sample fetch
12109 keyword.
12110
12111table_trackers(<table>)
12112 Uses the string representation of the input sample to perform a look up in
12113 the specified table. If the key is not found in the table, integer value zero
12114 is returned. Otherwise the converter returns the current amount of concurrent
12115 connections tracking the same key as the input sample in the designated
12116 table. It differs from table_conn_cur in that it does not rely on any stored
12117 information but on the table's reference count (the "use" value which is
12118 returned by "show table" on the CLI). This may sometimes be more suited for
12119 layer7 tracking. It can be used to tell a server how many concurrent
12120 connections there are from a given address for example. See also the
12121 sc_trackers sample fetch keyword.
12122
Willy Tarreauffcb2e42014-07-10 16:29:08 +020012123upper
12124 Convert a string sample to upper case. This can only be placed after a string
12125 sample fetch function or after a transformation keyword returning a string
12126 type. The result is of type string.
12127
Thierry FOURNIER82ff3c92015-05-07 15:46:20 +020012128url_dec
12129 Takes an url-encoded string provided as input and returns the decoded
12130 version as output. The input and the output are of type string.
12131
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020012132utime(<format>[,<offset>])
12133 Converts an integer supposed to contain a date since epoch to a string
12134 representing this date in UTC time using a format defined by the <format>
12135 string using strftime(3). The purpose is to allow any date format to be used
12136 in logs. An optional <offset> in seconds may be applied to the input date
12137 (positive or negative). See the strftime() man page for the format supported
12138 by your operating system. See also the ltime converter.
12139
12140 Example :
12141
12142 # Emit two colons, one with the UTC time and another with ip:port
12143 # Eg: 20140710162350 127.0.0.1:57325
12144 log-format %[date,utime(%Y%m%d%H%M%S)]\ %ci:%cp
12145
Emeric Brunc9a0f6d2014-11-25 14:09:01 +010012146word(<index>,<delimiters>)
12147 Extracts the nth word considering given delimiters from an input string.
12148 Indexes start at 1 and delimiters are a string formatted list of chars.
12149
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012150wt6([<avalanche>])
12151 Hashes a binary input sample into an unsigned 32-bit quantity using the WT6
12152 hash function. Optionally, it is possible to apply a full avalanche hash
12153 function to the output if the optional <avalanche> argument equals 1. This
12154 converter uses the same functions as used by the various hash-based load
12155 balancing algorithms, so it will provide exactly the same results. It is
12156 mostly intended for debugging, but can be used as a stick-table entry to
12157 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010012158 32-bit hash is trivial to break. See also "crc32", "djb2", "sdbm", and the
12159 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020012160
Willy Tarreau97707872015-01-27 15:12:13 +010012161xor(<value>)
12162 Performs a bitwise "XOR" (exclusive OR) between <value> and the input value
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012163 of type signed integer, and returns the result as an signed integer.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020012164 <value> can be a numeric value or a variable name. The name of the variable
12165 starts by an indication about its scope. The allowed scopes are:
12166 "sess" : the variable is shared with all the session,
12167 "txn" : the variable is shared with all the transaction (request and
12168 response),
12169 "req" : the variable is shared only during the request processing,
12170 "res" : the variable is shared only during the response processing.
12171 This prefix is followed by a name. The separator is a '.'. The name may only
12172 contain characters 'a-z', 'A-Z', '0-9' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010012173
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012174
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200121757.3.2. Fetching samples from internal states
Willy Tarreau74ca5042013-06-11 23:12:07 +020012176--------------------------------------------
12177
12178A first set of sample fetch methods applies to internal information which does
12179not even relate to any client information. These ones are sometimes used with
12180"monitor-fail" directives to report an internal status to external watchers.
12181The sample fetch methods described in this section are usable anywhere.
12182
12183always_false : boolean
12184 Always returns the boolean "false" value. It may be used with ACLs as a
12185 temporary replacement for another one when adjusting configurations.
12186
12187always_true : boolean
12188 Always returns the boolean "true" value. It may be used with ACLs as a
12189 temporary replacement for another one when adjusting configurations.
12190
12191avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010012192 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +020012193 divided by the number of active servers. The current backend is used if no
12194 backend is specified. This is very similar to "queue" except that the size of
12195 the farm is considered, in order to give a more accurate measurement of the
12196 time it may take for a new connection to be processed. The main usage is with
12197 ACL to return a sorry page to new users when it becomes certain they will get
12198 a degraded service, or to pass to the backend servers in a header so that
12199 they decide to work in degraded mode or to disable some functions to speed up
12200 the processing a bit. Note that in the event there would not be any active
12201 server anymore, twice the number of queued connections would be considered as
12202 the measured value. This is a fair estimate, as we expect one server to get
12203 back soon anyway, but we still prefer to send new traffic to another backend
12204 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
12205 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +010012206
Willy Tarreau74ca5042013-06-11 23:12:07 +020012207be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +020012208 Applies to the number of currently established connections on the backend,
12209 possibly including the connection being evaluated. If no backend name is
12210 specified, the current one is used. But it is also possible to check another
12211 backend. It can be used to use a specific farm when the nominal one is full.
12212 See also the "fe_conn", "queue" and "be_sess_rate" criteria.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012213
Willy Tarreau74ca5042013-06-11 23:12:07 +020012214be_sess_rate([<backend>]) : integer
12215 Returns an integer value corresponding to the sessions creation rate on the
12216 backend, in number of new sessions per second. This is used with ACLs to
12217 switch to an alternate backend when an expensive or fragile one reaches too
12218 high a session rate, or to limit abuse of service (eg. prevent sucking of an
12219 online dictionary). It can also be useful to add this element to logs using a
12220 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012221
12222 Example :
12223 # Redirect to an error page if the dictionary is requested too often
12224 backend dynamic
12225 mode http
12226 acl being_scanned be_sess_rate gt 100
12227 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +010012228
Thierry FOURNIERcc103292015-06-06 19:30:17 +020012229bin(<hexa>) : bin
12230 Returns a binary chain. The input is the hexadecimal representation
12231 of the string.
12232
12233bool(<bool>) : bool
12234 Returns a boolean value. <bool> can be 'true', 'false', '1' or '0'.
12235 'false' and '0' are the same. 'true' and '1' are the same.
12236
Willy Tarreau74ca5042013-06-11 23:12:07 +020012237connslots([<backend>]) : integer
12238 Returns an integer value corresponding to the number of connection slots
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012239 still available in the backend, by totaling the maximum amount of
Willy Tarreau74ca5042013-06-11 23:12:07 +020012240 connections on all servers and the maximum queue size. This is probably only
12241 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -050012242
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012243 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +020012244 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012245 usage; see "use_backend" keyword) can be redirected to a different backend.
12246
Willy Tarreau55165fe2009-05-10 12:02:55 +020012247 'connslots' = number of available server connection slots, + number of
12248 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012249
Willy Tarreaua36af912009-10-10 12:02:45 +020012250 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +020012251 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +020012252 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +020012253 you want to be able to differentiate between different backends, and their
12254 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +020012255 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +020012256 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012257
Willy Tarreau55165fe2009-05-10 12:02:55 +020012258 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
12259 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +020012260 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +020012261 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080012262
Willy Tarreau6236d3a2013-07-25 14:28:25 +020012263date([<offset>]) : integer
12264 Returns the current date as the epoch (number of seconds since 01/01/1970).
12265 If an offset value is specified, then it is a number of seconds that is added
12266 to the current date before returning the value. This is particularly useful
12267 to compute relative dates, as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +020012268 It is useful combined with the http_date converter.
12269
12270 Example :
12271
12272 # set an expires header to now+1 hour in every response
12273 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +020012274
Willy Tarreau595ec542013-06-12 21:34:28 +020012275env(<name>) : string
12276 Returns a string containing the value of environment variable <name>. As a
12277 reminder, environment variables are per-process and are sampled when the
12278 process starts. This can be useful to pass some information to a next hop
12279 server, or with ACLs to take specific action when the process is started a
12280 certain way.
12281
12282 Examples :
12283 # Pass the Via header to next hop with the local hostname in it
12284 http-request add-header Via 1.1\ %[env(HOSTNAME)]
12285
12286 # reject cookie-less requests when the STOP environment variable is set
12287 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
12288
Willy Tarreau74ca5042013-06-11 23:12:07 +020012289fe_conn([<frontend>]) : integer
12290 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +010012291 possibly including the connection being evaluated. If no frontend name is
12292 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020012293 frontend. It can be used to return a sorry page before hard-blocking, or to
12294 use a specific backend to drain new requests when the farm is considered
12295 full. This is mostly used with ACLs but can also be used to pass some
12296 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
12297 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +020012298
Willy Tarreau74ca5042013-06-11 23:12:07 +020012299fe_sess_rate([<frontend>]) : integer
12300 Returns an integer value corresponding to the sessions creation rate on the
12301 frontend, in number of new sessions per second. This is used with ACLs to
12302 limit the incoming session rate to an acceptable range in order to prevent
12303 abuse of service at the earliest moment, for example when combined with other
12304 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
12305 down below the limit. It can also be useful to add this element to logs using
12306 a log-format directive. See also the "rate-limit sessions" directive for use
12307 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010012308
12309 Example :
12310 # This frontend limits incoming mails to 10/s with a max of 100
12311 # concurrent connections. We accept any connection below 10/s, and
12312 # force excess clients to wait for 100 ms. Since clients are limited to
12313 # 100 max, there cannot be more than 10 incoming mails per second.
12314 frontend mail
12315 bind :25
12316 mode tcp
12317 maxconn 100
12318 acl too_fast fe_sess_rate ge 10
12319 tcp-request inspect-delay 100ms
12320 tcp-request content accept if ! too_fast
12321 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +010012322
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020012323int(<integer>) : signed integer
12324 Returns a signed integer.
12325
Thierry FOURNIERcc103292015-06-06 19:30:17 +020012326ipv4(<ipv4>) : ipv4
12327 Returns an ipv4.
12328
12329ipv6(<ipv6>) : ipv6
12330 Returns an ipv6.
12331
12332meth(<method>) : method
12333 Returns a method.
12334
Willy Tarreau0f30d262014-11-24 16:02:05 +010012335nbproc : integer
12336 Returns an integer value corresponding to the number of processes that were
12337 started (it equals the global "nbproc" setting). This is useful for logging
12338 and debugging purposes.
12339
Willy Tarreau74ca5042013-06-11 23:12:07 +020012340nbsrv([<backend>]) : integer
12341 Returns an integer value corresponding to the number of usable servers of
12342 either the current backend or the named backend. This is mostly used with
12343 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +010012344 switch to an alternate backend when the number of servers is too low to
12345 to handle some load. It is useful to report a failure when combined with
12346 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +010012347
Willy Tarreau0f30d262014-11-24 16:02:05 +010012348proc : integer
12349 Returns an integer value corresponding to the position of the process calling
12350 the function, between 1 and global.nbproc. This is useful for logging and
12351 debugging purposes.
12352
Willy Tarreau74ca5042013-06-11 23:12:07 +020012353queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010012354 Returns the total number of queued connections of the designated backend,
12355 including all the connections in server queues. If no backend name is
12356 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020012357 one. This is useful with ACLs or to pass statistics to backend servers. This
12358 can be used to take actions when queuing goes above a known level, generally
12359 indicating a surge of traffic or a massive slowdown on the servers. One
12360 possible action could be to reject new users but still accept old ones. See
12361 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
12362
Willy Tarreau84310e22014-02-14 11:59:04 +010012363rand([<range>]) : integer
12364 Returns a random integer value within a range of <range> possible values,
12365 starting at zero. If the range is not specified, it defaults to 2^32, which
12366 gives numbers between 0 and 4294967295. It can be useful to pass some values
12367 needed to take some routing decisions for example, or just for debugging
12368 purposes. This random must not be used for security purposes.
12369
Willy Tarreau74ca5042013-06-11 23:12:07 +020012370srv_conn([<backend>/]<server>) : integer
12371 Returns an integer value corresponding to the number of currently established
12372 connections on the designated server, possibly including the connection being
12373 evaluated. If <backend> is omitted, then the server is looked up in the
12374 current backend. It can be used to use a specific farm when one server is
12375 full, or to inform the server about our view of the number of active
12376 connections with it. See also the "fe_conn", "be_conn" and "queue" fetch
12377 methods.
12378
12379srv_is_up([<backend>/]<server>) : boolean
12380 Returns true when the designated server is UP, and false when it is either
12381 DOWN or in maintenance mode. If <backend> is omitted, then the server is
12382 looked up in the current backend. It is mainly used to take action based on
12383 an external status reported via a health check (eg: a geographical site's
12384 availability). Another possible use which is more of a hack consists in
12385 using dummy servers as boolean variables that can be enabled or disabled from
12386 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
12387
12388srv_sess_rate([<backend>/]<server>) : integer
12389 Returns an integer corresponding to the sessions creation rate on the
12390 designated server, in number of new sessions per second. If <backend> is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012391 omitted, then the server is looked up in the current backend. This is mostly
Willy Tarreau74ca5042013-06-11 23:12:07 +020012392 used with ACLs but can make sense with logs too. This is used to switch to an
12393 alternate backend when an expensive or fragile one reaches too high a session
12394 rate, or to limit abuse of service (eg. prevent latent requests from
12395 overloading servers).
12396
12397 Example :
12398 # Redirect to a separate back
12399 acl srv1_full srv_sess_rate(be1/srv1) gt 50
12400 acl srv2_full srv_sess_rate(be1/srv2) gt 50
12401 use_backend be2 if srv1_full or srv2_full
12402
Willy Tarreau0f30d262014-11-24 16:02:05 +010012403stopping : boolean
12404 Returns TRUE if the process calling the function is currently stopping. This
12405 can be useful for logging, or for relaxing certain checks or helping close
12406 certain connections upon graceful shutdown.
12407
Thierry FOURNIERcc103292015-06-06 19:30:17 +020012408str(<string>) : string
12409 Returns a string.
12410
Willy Tarreau74ca5042013-06-11 23:12:07 +020012411table_avl([<table>]) : integer
12412 Returns the total number of available entries in the current proxy's
12413 stick-table or in the designated stick-table. See also table_cnt.
12414
12415table_cnt([<table>]) : integer
12416 Returns the total number of entries currently in use in the current proxy's
12417 stick-table or in the designated stick-table. See also src_conn_cnt and
12418 table_avl for other entry counting methods.
12419
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012420var(<var-name>) : undefined
12421 Returns a variable with the stored type. If the variable is not set, the
12422 sample fetch fails. The name of the variable starts by an indication about its
12423 scope. The scope allowed are:
12424 "sess" : the variable is shared with all the session,
12425 "txn" : the variable is shared with all the transaction (request and
12426 response),
12427 "req" : the variable is shared only during the request processing,
12428 "res" : the variable is shared only during the response processing.
12429 This prefix is followed by a name. The separator is a '.'. The name may only
12430 contain characters 'a-z', 'A-Z', '0-9' and '_'.
12431
Willy Tarreau74ca5042013-06-11 23:12:07 +020012432
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200124337.3.3. Fetching samples at Layer 4
Willy Tarreau74ca5042013-06-11 23:12:07 +020012434----------------------------------
12435
12436The layer 4 usually describes just the transport layer which in haproxy is
12437closest to the connection, where no content is yet made available. The fetch
12438methods described here are usable as low as the "tcp-request connection" rule
12439sets unless they require some future information. Those generally include
12440TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012441the incoming connection. For retrieving a value from a sticky counters, the
12442counter number can be explicitly set as 0, 1, or 2 using the pre-defined
12443"sc0_", "sc1_", or "sc2_" prefix, or it can be specified as the first integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012444argument when using the "sc_" prefix. An optional table may be specified with
12445the "sc*" form, in which case the currently tracked key will be looked up into
12446this alternate table instead of the table currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012447
12448be_id : integer
12449 Returns an integer containing the current backend's id. It can be used in
12450 frontends with responses to check which backend processed the request.
12451
12452dst : ip
12453 This is the destination IPv4 address of the connection on the client side,
12454 which is the address the client connected to. It can be useful when running
12455 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
12456 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
12457 RFC 4291.
12458
12459dst_conn : integer
12460 Returns an integer value corresponding to the number of currently established
12461 connections on the same socket including the one being evaluated. It is
12462 normally used with ACLs but can as well be used to pass the information to
12463 servers in an HTTP header or in logs. It can be used to either return a sorry
12464 page before hard-blocking, or to use a specific backend to drain new requests
12465 when the socket is considered saturated. This offers the ability to assign
12466 different limits to different listening ports or addresses. See also the
12467 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012468
Willy Tarreau74ca5042013-06-11 23:12:07 +020012469dst_port : integer
12470 Returns an integer value corresponding to the destination TCP port of the
12471 connection on the client side, which is the port the client connected to.
12472 This might be used when running in transparent mode, when assigning dynamic
12473 ports to some clients for a whole application session, to stick all users to
12474 a same server, or to pass the destination port information to a server using
12475 an HTTP header.
12476
12477fe_id : integer
12478 Returns an integer containing the current frontend's id. It can be used in
12479 backends to check from which backend it was called, or to stick all users
12480 coming via a same frontend to the same server.
12481
Cyril Bonté62ba8702014-04-22 23:52:25 +020012482sc_bytes_in_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012483sc0_bytes_in_rate([<table>]) : integer
12484sc1_bytes_in_rate([<table>]) : integer
12485sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012486 Returns the average client-to-server bytes rate from the currently tracked
12487 counters, measured in amount of bytes over the period configured in the
12488 table. See also src_bytes_in_rate.
12489
Cyril Bonté62ba8702014-04-22 23:52:25 +020012490sc_bytes_out_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012491sc0_bytes_out_rate([<table>]) : integer
12492sc1_bytes_out_rate([<table>]) : integer
12493sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012494 Returns the average server-to-client bytes rate from the currently tracked
12495 counters, measured in amount of bytes over the period configured in the
12496 table. See also src_bytes_out_rate.
12497
Cyril Bonté62ba8702014-04-22 23:52:25 +020012498sc_clr_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012499sc0_clr_gpc0([<table>]) : integer
12500sc1_clr_gpc0([<table>]) : integer
12501sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +020012502 Clears the first General Purpose Counter associated to the currently tracked
12503 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +010012504 stored value is zero, so first invocation will always return zero. This is
12505 typically used as a second ACL in an expression in order to mark a connection
12506 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020012507
12508 # block if 5 consecutive requests continue to come faster than 10 sess
12509 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012510 acl abuse sc0_http_req_rate gt 10
12511 acl kill sc0_inc_gpc0 gt 5
12512 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020012513 tcp-request connection accept if !abuse save
12514 tcp-request connection reject if abuse kill
12515
Cyril Bonté62ba8702014-04-22 23:52:25 +020012516sc_conn_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012517sc0_conn_cnt([<table>]) : integer
12518sc1_conn_cnt([<table>]) : integer
12519sc2_conn_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012520 Returns the cumulated number of incoming connections from currently tracked
12521 counters. See also src_conn_cnt.
12522
Cyril Bonté62ba8702014-04-22 23:52:25 +020012523sc_conn_cur(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012524sc0_conn_cur([<table>]) : integer
12525sc1_conn_cur([<table>]) : integer
12526sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012527 Returns the current amount of concurrent connections tracking the same
12528 tracked counters. This number is automatically incremented when tracking
12529 begins and decremented when tracking stops. See also src_conn_cur.
12530
Cyril Bonté62ba8702014-04-22 23:52:25 +020012531sc_conn_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012532sc0_conn_rate([<table>]) : integer
12533sc1_conn_rate([<table>]) : integer
12534sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012535 Returns the average connection rate from the currently tracked counters,
12536 measured in amount of connections over the period configured in the table.
12537 See also src_conn_rate.
12538
Cyril Bonté62ba8702014-04-22 23:52:25 +020012539sc_get_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012540sc0_get_gpc0([<table>]) : integer
12541sc1_get_gpc0([<table>]) : integer
12542sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012543 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012544 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020012545
Thierry FOURNIER236657b2015-08-19 08:25:14 +020012546sc_get_gpt0(<ctr>[,<table>]) : integer
12547sc0_get_gpt0([<table>]) : integer
12548sc1_get_gpt0([<table>]) : integer
12549sc2_get_gpt0([<table>]) : integer
12550 Returns the value of the first General Purpose Tag associated to the
12551 currently tracked counters. See also src_get_gpt0.
12552
Cyril Bonté62ba8702014-04-22 23:52:25 +020012553sc_gpc0_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012554sc0_gpc0_rate([<table>]) : integer
12555sc1_gpc0_rate([<table>]) : integer
12556sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020012557 Returns the average increment rate of the first General Purpose Counter
12558 associated to the currently tracked counters. It reports the frequency
12559 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012560 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
12561 that the "gpc0_rate" counter must be stored in the stick-table for a value to
12562 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +020012563
Cyril Bonté62ba8702014-04-22 23:52:25 +020012564sc_http_err_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012565sc0_http_err_cnt([<table>]) : integer
12566sc1_http_err_cnt([<table>]) : integer
12567sc2_http_err_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012568 Returns the cumulated number of HTTP errors from the currently tracked
12569 counters. This includes the both request errors and 4xx error responses.
12570 See also src_http_err_cnt.
12571
Cyril Bonté62ba8702014-04-22 23:52:25 +020012572sc_http_err_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012573sc0_http_err_rate([<table>]) : integer
12574sc1_http_err_rate([<table>]) : integer
12575sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012576 Returns the average rate of HTTP errors from the currently tracked counters,
12577 measured in amount of errors over the period configured in the table. This
12578 includes the both request errors and 4xx error responses. See also
12579 src_http_err_rate.
12580
Cyril Bonté62ba8702014-04-22 23:52:25 +020012581sc_http_req_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012582sc0_http_req_cnt([<table>]) : integer
12583sc1_http_req_cnt([<table>]) : integer
12584sc2_http_req_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012585 Returns the cumulated number of HTTP requests from the currently tracked
12586 counters. This includes every started request, valid or not. See also
12587 src_http_req_cnt.
12588
Cyril Bonté62ba8702014-04-22 23:52:25 +020012589sc_http_req_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012590sc0_http_req_rate([<table>]) : integer
12591sc1_http_req_rate([<table>]) : integer
12592sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012593 Returns the average rate of HTTP requests from the currently tracked
12594 counters, measured in amount of requests over the period configured in
12595 the table. This includes every started request, valid or not. See also
12596 src_http_req_rate.
12597
Cyril Bonté62ba8702014-04-22 23:52:25 +020012598sc_inc_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012599sc0_inc_gpc0([<table>]) : integer
12600sc1_inc_gpc0([<table>]) : integer
12601sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012602 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +010012603 tracked counters, and returns its new value. Before the first invocation,
12604 the stored value is zero, so first invocation will increase it to 1 and will
12605 return 1. This is typically used as a second ACL in an expression in order
12606 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +020012607
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012608 acl abuse sc0_http_req_rate gt 10
12609 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020012610 tcp-request connection reject if abuse kill
12611
Cyril Bonté62ba8702014-04-22 23:52:25 +020012612sc_kbytes_in(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012613sc0_kbytes_in([<table>]) : integer
12614sc1_kbytes_in([<table>]) : integer
12615sc2_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020012616 Returns the total amount of client-to-server data from the currently tracked
12617 counters, measured in kilobytes. The test is currently performed on 32-bit
12618 integers, which limits values to 4 terabytes. See also src_kbytes_in.
Willy Tarreaue9656522010-08-17 15:40:09 +020012619
Cyril Bonté62ba8702014-04-22 23:52:25 +020012620sc_kbytes_out(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012621sc0_kbytes_out([<table>]) : integer
12622sc1_kbytes_out([<table>]) : integer
12623sc2_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020012624 Returns the total amount of server-to-client data from the currently tracked
12625 counters, measured in kilobytes. The test is currently performed on 32-bit
12626 integers, which limits values to 4 terabytes. See also src_kbytes_out.
Willy Tarreaue9656522010-08-17 15:40:09 +020012627
Cyril Bonté62ba8702014-04-22 23:52:25 +020012628sc_sess_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012629sc0_sess_cnt([<table>]) : integer
12630sc1_sess_cnt([<table>]) : integer
12631sc2_sess_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012632 Returns the cumulated number of incoming connections that were transformed
12633 into sessions, which means that they were accepted by a "tcp-request
12634 connection" rule, from the currently tracked counters. A backend may count
12635 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -040012636 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +020012637 with the client. See also src_sess_cnt.
12638
Cyril Bonté62ba8702014-04-22 23:52:25 +020012639sc_sess_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012640sc0_sess_rate([<table>]) : integer
12641sc1_sess_rate([<table>]) : integer
12642sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020012643 Returns the average session rate from the currently tracked counters,
12644 measured in amount of sessions over the period configured in the table. A
12645 session is a connection that got past the early "tcp-request connection"
12646 rules. A backend may count more sessions than connections because each
12647 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -040012648 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +020012649
Cyril Bonté62ba8702014-04-22 23:52:25 +020012650sc_tracked(<ctr>[,<table>]) : boolean
Willy Tarreau0f791d42013-07-23 19:56:43 +020012651sc0_tracked([<table>]) : boolean
12652sc1_tracked([<table>]) : boolean
12653sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +020012654 Returns true if the designated session counter is currently being tracked by
12655 the current session. This can be useful when deciding whether or not we want
12656 to set some values in a header passed to the server.
12657
Cyril Bonté62ba8702014-04-22 23:52:25 +020012658sc_trackers(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020012659sc0_trackers([<table>]) : integer
12660sc1_trackers([<table>]) : integer
12661sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +010012662 Returns the current amount of concurrent connections tracking the same
12663 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012664 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +010012665 that it does not rely on any stored information but on the table's reference
12666 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020012667 may sometimes be more suited for layer7 tracking. It can be used to tell a
12668 server how many concurrent connections there are from a given address for
12669 example.
Willy Tarreau2406db42012-12-09 12:16:43 +010012670
Willy Tarreau74ca5042013-06-11 23:12:07 +020012671so_id : integer
12672 Returns an integer containing the current listening socket's id. It is useful
12673 in frontends involving many "bind" lines, or to stick all users coming via a
12674 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012675
Willy Tarreau74ca5042013-06-11 23:12:07 +020012676src : ip
12677 This is the source IPv4 address of the client of the session. It is of type
12678 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
12679 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
12680 TCP-level source address which is used, and not the address of a client
12681 behind a proxy. However if the "accept-proxy" bind directive is used, it can
12682 be the address of a client behind another PROXY-protocol compatible component
12683 for all rule sets except "tcp-request connection" which sees the real address.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012684
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010012685 Example:
12686 # add an HTTP header in requests with the originating address' country
12687 http-request set-header X-Country %[src,map_ip(geoip.lst)]
12688
Willy Tarreau74ca5042013-06-11 23:12:07 +020012689src_bytes_in_rate([<table>]) : integer
12690 Returns the average bytes rate from the incoming connection's source address
12691 in the current proxy's stick-table or in the designated stick-table, measured
12692 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012693 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012694
Willy Tarreau74ca5042013-06-11 23:12:07 +020012695src_bytes_out_rate([<table>]) : integer
12696 Returns the average bytes rate to the incoming connection's source address in
12697 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +020012698 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012699 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012700
Willy Tarreau74ca5042013-06-11 23:12:07 +020012701src_clr_gpc0([<table>]) : integer
12702 Clears the first General Purpose Counter associated to the incoming
12703 connection's source address in the current proxy's stick-table or in the
12704 designated stick-table, and returns its previous value. If the address is not
12705 found, an entry is created and 0 is returned. This is typically used as a
12706 second ACL in an expression in order to mark a connection when a first ACL
12707 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020012708
12709 # block if 5 consecutive requests continue to come faster than 10 sess
12710 # per second, and reset the counter as soon as the traffic slows down.
12711 acl abuse src_http_req_rate gt 10
12712 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +010012713 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020012714 tcp-request connection accept if !abuse save
12715 tcp-request connection reject if abuse kill
12716
Willy Tarreau74ca5042013-06-11 23:12:07 +020012717src_conn_cnt([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020012718 Returns the cumulated number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +020012719 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020012720 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012721 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012722
Willy Tarreau74ca5042013-06-11 23:12:07 +020012723src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020012724 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +020012725 current incoming connection's source address in the current proxy's
12726 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012727 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012728
Willy Tarreau74ca5042013-06-11 23:12:07 +020012729src_conn_rate([<table>]) : integer
12730 Returns the average connection rate from the incoming connection's source
12731 address in the current proxy's stick-table or in the designated stick-table,
12732 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012733 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012734
Willy Tarreau74ca5042013-06-11 23:12:07 +020012735src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020012736 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +020012737 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020012738 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012739 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012740
Thierry FOURNIER236657b2015-08-19 08:25:14 +020012741src_get_gpt0([<table>]) : integer
12742 Returns the value of the first General Purpose Tag associated to the
12743 incoming connection's source address in the current proxy's stick-table or in
12744 the designated stick-table. If the address is not found, zero is returned.
12745 See also sc/sc0/sc1/sc2_get_gpt0.
12746
Willy Tarreau74ca5042013-06-11 23:12:07 +020012747src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020012748 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +020012749 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +020012750 stick-table or in the designated stick-table. It reports the frequency
12751 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012752 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
12753 that the "gpc0_rate" counter must be stored in the stick-table for a value to
12754 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020012755
Willy Tarreau74ca5042013-06-11 23:12:07 +020012756src_http_err_cnt([<table>]) : integer
12757 Returns the cumulated number of HTTP errors from the incoming connection's
12758 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +020012759 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012760 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +020012761 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012762
Willy Tarreau74ca5042013-06-11 23:12:07 +020012763src_http_err_rate([<table>]) : integer
12764 Returns the average rate of HTTP errors from the incoming connection's source
12765 address in the current proxy's stick-table or in the designated stick-table,
12766 measured in amount of errors over the period configured in the table. This
12767 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012768 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012769
Willy Tarreau74ca5042013-06-11 23:12:07 +020012770src_http_req_cnt([<table>]) : integer
12771 Returns the cumulated number of HTTP requests from the incoming connection's
12772 source address in the current proxy's stick-table or in the designated stick-
12773 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012774 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012775
Willy Tarreau74ca5042013-06-11 23:12:07 +020012776src_http_req_rate([<table>]) : integer
12777 Returns the average rate of HTTP requests from the incoming connection's
12778 source address in the current proxy's stick-table or in the designated stick-
12779 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +020012780 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012781 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012782
Willy Tarreau74ca5042013-06-11 23:12:07 +020012783src_inc_gpc0([<table>]) : integer
12784 Increments the first General Purpose Counter associated to the incoming
12785 connection's source address in the current proxy's stick-table or in the
12786 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012787 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012788 This is typically used as a second ACL in an expression in order to mark a
12789 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +020012790
12791 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +010012792 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020012793 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +020012794
Willy Tarreau74ca5042013-06-11 23:12:07 +020012795src_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020012796 Returns the total amount of data received from the incoming connection's
12797 source address in the current proxy's stick-table or in the designated
12798 stick-table, measured in kilobytes. If the address is not found, zero is
12799 returned. The test is currently performed on 32-bit integers, which limits
12800 values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012801
Willy Tarreau74ca5042013-06-11 23:12:07 +020012802src_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020012803 Returns the total amount of data sent to the incoming connection's source
12804 address in the current proxy's stick-table or in the designated stick-table,
12805 measured in kilobytes. If the address is not found, zero is returned. The
12806 test is currently performed on 32-bit integers, which limits values to 4
12807 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020012808
Willy Tarreau74ca5042013-06-11 23:12:07 +020012809src_port : integer
12810 Returns an integer value corresponding to the TCP source port of the
12811 connection on the client side, which is the port the client connected from.
12812 Usage of this function is very limited as modern protocols do not care much
12813 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010012814
Willy Tarreau74ca5042013-06-11 23:12:07 +020012815src_sess_cnt([<table>]) : integer
12816 Returns the cumulated number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +020012817 connection's source IPv4 address in the current proxy's stick-table or in the
12818 designated stick-table, that were transformed into sessions, which means that
12819 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012820 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012821
Willy Tarreau74ca5042013-06-11 23:12:07 +020012822src_sess_rate([<table>]) : integer
12823 Returns the average session rate from the incoming connection's source
12824 address in the current proxy's stick-table or in the designated stick-table,
12825 measured in amount of sessions over the period configured in the table. A
12826 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020012827 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020012828
Willy Tarreau74ca5042013-06-11 23:12:07 +020012829src_updt_conn_cnt([<table>]) : integer
12830 Creates or updates the entry associated to the incoming connection's source
12831 address in the current proxy's stick-table or in the designated stick-table.
12832 This table must be configured to store the "conn_cnt" data type, otherwise
12833 the match will be ignored. The current count is incremented by one, and the
12834 expiration timer refreshed. The updated count is returned, so this match
12835 can't return zero. This was used to reject service abusers based on their
12836 source address. Note: it is recommended to use the more complete "track-sc*"
12837 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020012838
12839 Example :
12840 # This frontend limits incoming SSH connections to 3 per 10 second for
12841 # each source address, and rejects excess connections until a 10 second
12842 # silence is observed. At most 20 addresses are tracked.
12843 listen ssh
12844 bind :22
12845 mode tcp
12846 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +020012847 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +020012848 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +020012849 server local 127.0.0.1:22
12850
Willy Tarreau74ca5042013-06-11 23:12:07 +020012851srv_id : integer
12852 Returns an integer containing the server's id when processing the response.
12853 While it's almost only used with ACLs, it may be used for logging or
12854 debugging.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +020012855
Hervé COMMOWICK35ed8012010-12-15 14:04:51 +010012856
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200128577.3.4. Fetching samples at Layer 5
Willy Tarreau74ca5042013-06-11 23:12:07 +020012858----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +020012859
Willy Tarreau74ca5042013-06-11 23:12:07 +020012860The layer 5 usually describes just the session layer which in haproxy is
12861closest to the session once all the connection handshakes are finished, but
12862when no content is yet made available. The fetch methods described here are
12863usable as low as the "tcp-request content" rule sets unless they require some
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012864future information. Those generally include the results of SSL negotiations.
Willy Tarreauc735a072011-03-29 00:57:02 +020012865
Emeric Brun645ae792014-04-30 14:21:06 +020012866ssl_bc : boolean
12867 Returns true when the back connection was made via an SSL/TLS transport
12868 layer and is locally deciphered. This means the outgoing connection was made
12869 other a server with the "ssl" option.
12870
12871ssl_bc_alg_keysize : integer
12872 Returns the symmetric cipher key size supported in bits when the outgoing
12873 connection was made over an SSL/TLS transport layer.
12874
12875ssl_bc_cipher : string
12876 Returns the name of the used cipher when the outgoing connection was made
12877 over an SSL/TLS transport layer.
12878
12879ssl_bc_protocol : string
12880 Returns the name of the used protocol when the outgoing connection was made
12881 over an SSL/TLS transport layer.
12882
Emeric Brunb73a9b02014-04-30 18:49:19 +020012883ssl_bc_unique_id : binary
Emeric Brun645ae792014-04-30 14:21:06 +020012884 When the outgoing connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020012885 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
12886 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
Emeric Brun645ae792014-04-30 14:21:06 +020012887
12888ssl_bc_session_id : binary
12889 Returns the SSL ID of the back connection when the outgoing connection was
12890 made over an SSL/TLS transport layer. It is useful to log if we want to know
12891 if session was reused or not.
12892
12893ssl_bc_use_keysize : integer
12894 Returns the symmetric cipher key size used in bits when the outgoing
12895 connection was made over an SSL/TLS transport layer.
12896
Willy Tarreau74ca5042013-06-11 23:12:07 +020012897ssl_c_ca_err : integer
12898 When the incoming connection was made over an SSL/TLS transport layer,
12899 returns the ID of the first error detected during verification of the client
12900 certificate at depth > 0, or 0 if no error was encountered during this
12901 verification process. Please refer to your SSL library's documentation to
12902 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +020012903
Willy Tarreau74ca5042013-06-11 23:12:07 +020012904ssl_c_ca_err_depth : integer
12905 When the incoming connection was made over an SSL/TLS transport layer,
12906 returns the depth in the CA chain of the first error detected during the
12907 verification of the client certificate. If no error is encountered, 0 is
12908 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +010012909
Emeric Brun43e79582014-10-29 19:03:26 +010012910ssl_c_der : binary
12911 Returns the DER formatted certificate presented by the client when the
12912 incoming connection was made over an SSL/TLS transport layer. When used for
12913 an ACL, the value(s) to match against can be passed in hexadecimal form.
12914
Willy Tarreau74ca5042013-06-11 23:12:07 +020012915ssl_c_err : integer
12916 When the incoming connection was made over an SSL/TLS transport layer,
12917 returns the ID of the first error detected during verification at depth 0, or
12918 0 if no error was encountered during this verification process. Please refer
12919 to your SSL library's documentation to find the exhaustive list of error
12920 codes.
Willy Tarreau62644772008-07-16 18:36:06 +020012921
Willy Tarreau74ca5042013-06-11 23:12:07 +020012922ssl_c_i_dn([<entry>[,<occ>]]) : string
12923 When the incoming connection was made over an SSL/TLS transport layer,
12924 returns the full distinguished name of the issuer of the certificate
12925 presented by the client when no <entry> is specified, or the value of the
12926 first given entry found from the beginning of the DN. If a positive/negative
12927 occurrence number is specified as the optional second argument, it returns
12928 the value of the nth given entry value from the beginning/end of the DN.
12929 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
12930 "ssl_c_i_dn(CN)" retrieves the common name.
Willy Tarreau62644772008-07-16 18:36:06 +020012931
Willy Tarreau74ca5042013-06-11 23:12:07 +020012932ssl_c_key_alg : string
12933 Returns the name of the algorithm used to generate the key of the certificate
12934 presented by the client when the incoming connection was made over an SSL/TLS
12935 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +020012936
Willy Tarreau74ca5042013-06-11 23:12:07 +020012937ssl_c_notafter : string
12938 Returns the end date presented by the client as a formatted string
12939 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
12940 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +020012941
Willy Tarreau74ca5042013-06-11 23:12:07 +020012942ssl_c_notbefore : string
12943 Returns the start date presented by the client as a formatted string
12944 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
12945 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +010012946
Willy Tarreau74ca5042013-06-11 23:12:07 +020012947ssl_c_s_dn([<entry>[,<occ>]]) : string
12948 When the incoming connection was made over an SSL/TLS transport layer,
12949 returns the full distinguished name of the subject of the certificate
12950 presented by the client when no <entry> is specified, or the value of the
12951 first given entry found from the beginning of the DN. If a positive/negative
12952 occurrence number is specified as the optional second argument, it returns
12953 the value of the nth given entry value from the beginning/end of the DN.
12954 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
12955 "ssl_c_s_dn(CN)" retrieves the common name.
Willy Tarreaub6672b52011-12-12 17:23:41 +010012956
Willy Tarreau74ca5042013-06-11 23:12:07 +020012957ssl_c_serial : binary
12958 Returns the serial of the certificate presented by the client when the
12959 incoming connection was made over an SSL/TLS transport layer. When used for
12960 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +020012961
Willy Tarreau74ca5042013-06-11 23:12:07 +020012962ssl_c_sha1 : binary
12963 Returns the SHA-1 fingerprint of the certificate presented by the client when
12964 the incoming connection was made over an SSL/TLS transport layer. This can be
12965 used to stick a client to a server, or to pass this information to a server.
Willy Tarreau2d0caa32014-07-02 19:01:22 +020012966 Note that the output is binary, so if you want to pass that signature to the
12967 server, you need to encode it in hex or base64, such as in the example below:
12968
12969 http-request set-header X-SSL-Client-SHA1 %[ssl_c_sha1,hex]
Emeric Brun2525b6b2012-10-18 15:59:43 +020012970
Willy Tarreau74ca5042013-06-11 23:12:07 +020012971ssl_c_sig_alg : string
12972 Returns the name of the algorithm used to sign the certificate presented by
12973 the client when the incoming connection was made over an SSL/TLS transport
12974 layer.
Emeric Brun87855892012-10-17 17:39:35 +020012975
Willy Tarreau74ca5042013-06-11 23:12:07 +020012976ssl_c_used : boolean
12977 Returns true if current SSL session uses a client certificate even if current
12978 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +020012979
Willy Tarreau74ca5042013-06-11 23:12:07 +020012980ssl_c_verify : integer
12981 Returns the verify result error ID when the incoming connection was made over
12982 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
12983 refer to your SSL library's documentation for an exhaustive list of error
12984 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +020012985
Willy Tarreau74ca5042013-06-11 23:12:07 +020012986ssl_c_version : integer
12987 Returns the version of the certificate presented by the client when the
12988 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +020012989
Emeric Brun43e79582014-10-29 19:03:26 +010012990ssl_f_der : binary
12991 Returns the DER formatted certificate presented by the frontend when the
12992 incoming connection was made over an SSL/TLS transport layer. When used for
12993 an ACL, the value(s) to match against can be passed in hexadecimal form.
12994
Willy Tarreau74ca5042013-06-11 23:12:07 +020012995ssl_f_i_dn([<entry>[,<occ>]]) : string
12996 When the incoming connection was made over an SSL/TLS transport layer,
12997 returns the full distinguished name of the issuer of the certificate
12998 presented by the frontend when no <entry> is specified, or the value of the
12999 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +020013000 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +020013001 the value of the nth given entry value from the beginning/end of the DN.
13002 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
13003 "ssl_f_i_dn(CN)" retrieves the common name.
Emeric Brun87855892012-10-17 17:39:35 +020013004
Willy Tarreau74ca5042013-06-11 23:12:07 +020013005ssl_f_key_alg : string
13006 Returns the name of the algorithm used to generate the key of the certificate
13007 presented by the frontend when the incoming connection was made over an
13008 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020013009
Willy Tarreau74ca5042013-06-11 23:12:07 +020013010ssl_f_notafter : string
13011 Returns the end date presented by the frontend as a formatted string
13012 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
13013 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +020013014
Willy Tarreau74ca5042013-06-11 23:12:07 +020013015ssl_f_notbefore : string
13016 Returns the start date presented by the frontend as a formatted string
13017 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
13018 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +020013019
Willy Tarreau74ca5042013-06-11 23:12:07 +020013020ssl_f_s_dn([<entry>[,<occ>]]) : string
13021 When the incoming connection was made over an SSL/TLS transport layer,
13022 returns the full distinguished name of the subject of the certificate
13023 presented by the frontend when no <entry> is specified, or the value of the
13024 first given entry found from the beginning of the DN. If a positive/negative
13025 occurrence number is specified as the optional second argument, it returns
13026 the value of the nth given entry value from the beginning/end of the DN.
13027 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
13028 "ssl_f_s_dn(CN)" retrieves the common name.
Emeric Brunce5ad802012-10-22 14:11:22 +020013029
Willy Tarreau74ca5042013-06-11 23:12:07 +020013030ssl_f_serial : binary
13031 Returns the serial of the certificate presented by the frontend when the
13032 incoming connection was made over an SSL/TLS transport layer. When used for
13033 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +020013034
Emeric Brun55f4fa82014-04-30 17:11:25 +020013035ssl_f_sha1 : binary
13036 Returns the SHA-1 fingerprint of the certificate presented by the frontend
13037 when the incoming connection was made over an SSL/TLS transport layer. This
13038 can be used to know which certificate was chosen using SNI.
13039
Willy Tarreau74ca5042013-06-11 23:12:07 +020013040ssl_f_sig_alg : string
13041 Returns the name of the algorithm used to sign the certificate presented by
13042 the frontend when the incoming connection was made over an SSL/TLS transport
13043 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020013044
Willy Tarreau74ca5042013-06-11 23:12:07 +020013045ssl_f_version : integer
13046 Returns the version of the certificate presented by the frontend when the
13047 incoming connection was made over an SSL/TLS transport layer.
13048
13049ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020013050 Returns true when the front connection was made via an SSL/TLS transport
13051 layer and is locally deciphered. This means it has matched a socket declared
13052 with a "bind" line having the "ssl" option.
13053
Willy Tarreau74ca5042013-06-11 23:12:07 +020013054 Example :
13055 # This passes "X-Proto: https" to servers when client connects over SSL
13056 listen http-https
13057 bind :80
13058 bind :443 ssl crt /etc/haproxy.pem
13059 http-request add-header X-Proto https if { ssl_fc }
13060
13061ssl_fc_alg_keysize : integer
13062 Returns the symmetric cipher key size supported in bits when the incoming
13063 connection was made over an SSL/TLS transport layer.
13064
13065ssl_fc_alpn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013066 This extracts the Application Layer Protocol Negotiation field from an
Willy Tarreau74ca5042013-06-11 23:12:07 +020013067 incoming connection made via a TLS transport layer and locally deciphered by
13068 haproxy. The result is a string containing the protocol name advertised by
13069 the client. The SSL library must have been built with support for TLS
13070 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
13071 not advertised unless the "alpn" keyword on the "bind" line specifies a
13072 protocol list. Also, nothing forces the client to pick a protocol from this
13073 list, any other one may be requested. The TLS ALPN extension is meant to
13074 replace the TLS NPN extension. See also "ssl_fc_npn".
13075
Willy Tarreau74ca5042013-06-11 23:12:07 +020013076ssl_fc_cipher : string
13077 Returns the name of the used cipher when the incoming connection was made
13078 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +020013079
Willy Tarreau74ca5042013-06-11 23:12:07 +020013080ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020013081 Returns true if a client certificate is present in an incoming connection over
13082 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +010013083 Note: on SSL session resumption with Session ID or TLS ticket, client
13084 certificate is not present in the current connection but may be retrieved
13085 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
13086 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +020013087
Willy Tarreau74ca5042013-06-11 23:12:07 +020013088ssl_fc_has_sni : boolean
13089 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020013090 in an incoming connection was made over an SSL/TLS transport layer. Returns
13091 true when the incoming connection presents a TLS SNI field. This requires
13092 that the SSL library is build with support for TLS extensions enabled (check
13093 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +020013094
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020013095ssl_fc_is_resumed: boolean
13096 Returns true if the SSL/TLS session has been resumed through the use of
13097 SSL session cache or TLS tickets.
13098
Willy Tarreau74ca5042013-06-11 23:12:07 +020013099ssl_fc_npn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013100 This extracts the Next Protocol Negotiation field from an incoming connection
Willy Tarreau74ca5042013-06-11 23:12:07 +020013101 made via a TLS transport layer and locally deciphered by haproxy. The result
13102 is a string containing the protocol name advertised by the client. The SSL
13103 library must have been built with support for TLS extensions enabled (check
13104 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
13105 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
13106 forces the client to pick a protocol from this list, any other one may be
13107 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +020013108
Willy Tarreau74ca5042013-06-11 23:12:07 +020013109ssl_fc_protocol : string
13110 Returns the name of the used protocol when the incoming connection was made
13111 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +020013112
Emeric Brunb73a9b02014-04-30 18:49:19 +020013113ssl_fc_unique_id : binary
David Sc1ad52e2014-04-08 18:48:47 -040013114 When the incoming connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020013115 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
13116 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
David Sc1ad52e2014-04-08 18:48:47 -040013117
Willy Tarreau74ca5042013-06-11 23:12:07 +020013118ssl_fc_session_id : binary
13119 Returns the SSL ID of the front connection when the incoming connection was
13120 made over an SSL/TLS transport layer. It is useful to stick a given client to
13121 a server. It is important to note that some browsers refresh their session ID
13122 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +020013123
Willy Tarreau74ca5042013-06-11 23:12:07 +020013124ssl_fc_sni : string
13125 This extracts the Server Name Indication TLS extension (SNI) field from an
13126 incoming connection made via an SSL/TLS transport layer and locally
13127 deciphered by haproxy. The result (when present) typically is a string
13128 matching the HTTPS host name (253 chars or less). The SSL library must have
13129 been built with support for TLS extensions enabled (check haproxy -vv).
13130
13131 This fetch is different from "req_ssl_sni" above in that it applies to the
13132 connection being deciphered by haproxy and not to SSL contents being blindly
13133 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +020013134 requires that the SSL library is build with support for TLS extensions
13135 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +020013136
Willy Tarreau74ca5042013-06-11 23:12:07 +020013137 ACL derivatives :
Willy Tarreau74ca5042013-06-11 23:12:07 +020013138 ssl_fc_sni_end : suffix match
13139 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +020013140
Willy Tarreau74ca5042013-06-11 23:12:07 +020013141ssl_fc_use_keysize : integer
13142 Returns the symmetric cipher key size used in bits when the incoming
13143 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020013144
Willy Tarreaub6fb4202008-07-20 11:18:28 +020013145
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200131467.3.5. Fetching samples from buffer contents (Layer 6)
Willy Tarreau74ca5042013-06-11 23:12:07 +020013147------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +020013148
Willy Tarreau74ca5042013-06-11 23:12:07 +020013149Fetching samples from buffer contents is a bit different from the previous
13150sample fetches above because the sampled data are ephemeral. These data can
13151only be used when they're available and will be lost when they're forwarded.
13152For this reason, samples fetched from buffer contents during a request cannot
13153be used in a response for example. Even while the data are being fetched, they
13154can change. Sometimes it is necessary to set some delays or combine multiple
13155sample fetch methods to ensure that the expected data are complete and usable,
13156for example through TCP request content inspection. Please see the "tcp-request
13157content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +020013158
Willy Tarreau74ca5042013-06-11 23:12:07 +020013159payload(<offset>,<length>) : binary (deprecated)
13160 This is an alias for "req.payload" when used in the context of a request (eg:
13161 "stick on", "stick match"), and for "res.payload" when used in the context of
13162 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010013163
Willy Tarreau74ca5042013-06-11 23:12:07 +020013164payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
13165 This is an alias for "req.payload_lv" when used in the context of a request
13166 (eg: "stick on", "stick match"), and for "res.payload_lv" when used in the
13167 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010013168
Willy Tarreau74ca5042013-06-11 23:12:07 +020013169req.len : integer
13170req_len : integer (deprecated)
13171 Returns an integer value corresponding to the number of bytes present in the
13172 request buffer. This is mostly used in ACL. It is important to understand
13173 that this test does not return false as long as the buffer is changing. This
13174 means that a check with equality to zero will almost always immediately match
13175 at the beginning of the session, while a test for more data will wait for
13176 that data to come in and return false only when haproxy is certain that no
13177 more data will come in. This test was designed to be used with TCP request
13178 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013179
Willy Tarreau74ca5042013-06-11 23:12:07 +020013180req.payload(<offset>,<length>) : binary
13181 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020013182 in the request buffer. As a special case, if the <length> argument is zero,
13183 the the whole buffer from <offset> to the end is extracted. This can be used
13184 with ACLs in order to check for the presence of some content in a buffer at
13185 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013186
Willy Tarreau74ca5042013-06-11 23:12:07 +020013187 ACL alternatives :
13188 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013189
Willy Tarreau74ca5042013-06-11 23:12:07 +020013190req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
13191 This extracts a binary block whose size is specified at <offset1> for <length>
13192 bytes, and which starts at <offset2> if specified or just after the length in
13193 the request buffer. The <offset2> parameter also supports relative offsets if
13194 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013195
Willy Tarreau74ca5042013-06-11 23:12:07 +020013196 ACL alternatives :
13197 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013198
Willy Tarreau74ca5042013-06-11 23:12:07 +020013199 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013200
Willy Tarreau74ca5042013-06-11 23:12:07 +020013201req.proto_http : boolean
13202req_proto_http : boolean (deprecated)
13203 Returns true when data in the request buffer look like HTTP and correctly
13204 parses as such. It is the same parser as the common HTTP request parser which
13205 is used so there should be no surprises. The test does not match until the
13206 request is complete, failed or timed out. This test may be used to report the
13207 protocol in TCP logs, but the biggest use is to block TCP request analysis
13208 until a complete HTTP request is present in the buffer, for example to track
13209 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013210
Willy Tarreau74ca5042013-06-11 23:12:07 +020013211 Example:
13212 # track request counts per "base" (concatenation of Host+URL)
13213 tcp-request inspect-delay 10s
13214 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020013215 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020013216
Willy Tarreau74ca5042013-06-11 23:12:07 +020013217req.rdp_cookie([<name>]) : string
13218rdp_cookie([<name>]) : string (deprecated)
13219 When the request buffer looks like the RDP protocol, extracts the RDP cookie
13220 <name>, or any cookie if unspecified. The parser only checks for the first
13221 cookie, as illustrated in the RDP protocol specification. The cookie name is
13222 case insensitive. Generally the "MSTS" cookie name will be used, as it can
13223 contain the user name of the client connecting to the server if properly
13224 configured on the client. The "MSTSHASH" cookie is often used as well for
13225 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013226
Willy Tarreau74ca5042013-06-11 23:12:07 +020013227 This differs from "balance rdp-cookie" in that any balancing algorithm may be
13228 used and thus the distribution of clients to backend servers is not linked to
13229 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
13230 such as "balance roundrobin" or "balance leastconn" will lead to a more even
13231 distribution of clients to backend servers than the hash used by "balance
13232 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013233
Willy Tarreau74ca5042013-06-11 23:12:07 +020013234 ACL derivatives :
13235 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013236
Willy Tarreau74ca5042013-06-11 23:12:07 +020013237 Example :
13238 listen tse-farm
13239 bind 0.0.0.0:3389
13240 # wait up to 5s for an RDP cookie in the request
13241 tcp-request inspect-delay 5s
13242 tcp-request content accept if RDP_COOKIE
13243 # apply RDP cookie persistence
13244 persist rdp-cookie
13245 # Persist based on the mstshash cookie
13246 # This is only useful makes sense if
13247 # balance rdp-cookie is not used
13248 stick-table type string size 204800
13249 stick on req.rdp_cookie(mstshash)
13250 server srv1 1.1.1.1:3389
13251 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013252
Willy Tarreau74ca5042013-06-11 23:12:07 +020013253 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
13254 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013255
Willy Tarreau74ca5042013-06-11 23:12:07 +020013256req.rdp_cookie_cnt([name]) : integer
13257rdp_cookie_cnt([name]) : integer (deprecated)
13258 Tries to parse the request buffer as RDP protocol, then returns an integer
13259 corresponding to the number of RDP cookies found. If an optional cookie name
13260 is passed, only cookies matching this name are considered. This is mostly
13261 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013262
Willy Tarreau74ca5042013-06-11 23:12:07 +020013263 ACL derivatives :
13264 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013265
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020013266req.ssl_ec_ext : boolean
13267 Returns a boolean identifying if client sent the Supported Elliptic Curves
13268 Extension as defined in RFC4492, section 5.1. within the SSL ClientHello
Cyril Bonté307ee1e2015-09-28 23:16:06 +020013269 message. This can be used to present ECC compatible clients with EC
13270 certificate and to use RSA for all others, on the same IP address. Note that
13271 this only applies to raw contents found in the request buffer and not to
13272 contents deciphered via an SSL data layer, so this will not work with "bind"
13273 lines having the "ssl" option.
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020013274
Willy Tarreau74ca5042013-06-11 23:12:07 +020013275req.ssl_hello_type : integer
13276req_ssl_hello_type : integer (deprecated)
13277 Returns an integer value containing the type of the SSL hello message found
13278 in the request buffer if the buffer contains data that parse as a complete
13279 SSL (v3 or superior) client hello message. Note that this only applies to raw
13280 contents found in the request buffer and not to contents deciphered via an
13281 SSL data layer, so this will not work with "bind" lines having the "ssl"
13282 option. This is mostly used in ACL to detect presence of an SSL hello message
13283 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013284
Willy Tarreau74ca5042013-06-11 23:12:07 +020013285req.ssl_sni : string
13286req_ssl_sni : string (deprecated)
13287 Returns a string containing the value of the Server Name TLS extension sent
13288 by a client in a TLS stream passing through the request buffer if the buffer
13289 contains data that parse as a complete SSL (v3 or superior) client hello
13290 message. Note that this only applies to raw contents found in the request
13291 buffer and not to contents deciphered via an SSL data layer, so this will not
13292 work with "bind" lines having the "ssl" option. SNI normally contains the
13293 name of the host the client tries to connect to (for recent browsers). SNI is
13294 useful for allowing or denying access to certain hosts when SSL/TLS is used
13295 by the client. This test was designed to be used with TCP request content
13296 inspection. If content switching is needed, it is recommended to first wait
13297 for a complete client hello (type 1), like in the example below. See also
13298 "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013299
Willy Tarreau74ca5042013-06-11 23:12:07 +020013300 ACL derivatives :
13301 req_ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013302
Willy Tarreau74ca5042013-06-11 23:12:07 +020013303 Examples :
13304 # Wait for a client hello for at most 5 seconds
13305 tcp-request inspect-delay 5s
13306 tcp-request content accept if { req_ssl_hello_type 1 }
13307 use_backend bk_allow if { req_ssl_sni -f allowed_sites }
13308 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +020013309
Pradeep Jindalbb2acf52015-09-29 10:12:57 +053013310req.ssl_st_ext : integer
13311 Returns 0 if the client didn't send a SessionTicket TLS Extension (RFC5077)
13312 Returns 1 if the client sent SessionTicket TLS Extension
13313 Returns 2 if the client also sent non-zero length TLS SessionTicket
13314 Note that this only applies to raw contents found in the request buffer and
13315 not to contents deciphered via an SSL data layer, so this will not work with
13316 "bind" lines having the "ssl" option. This can for example be used to detect
13317 whether the client sent a SessionTicket or not and stick it accordingly, if
13318 no SessionTicket then stick on SessionID or don't stick as there's no server
13319 side state is there when SessionTickets are in use.
13320
Willy Tarreau74ca5042013-06-11 23:12:07 +020013321req.ssl_ver : integer
13322req_ssl_ver : integer (deprecated)
13323 Returns an integer value containing the version of the SSL/TLS protocol of a
13324 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
13325 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
13326 composed of the major version multiplied by 65536, added to the minor
13327 version. Note that this only applies to raw contents found in the request
13328 buffer and not to contents deciphered via an SSL data layer, so this will not
13329 work with "bind" lines having the "ssl" option. The ACL version of the test
13330 matches against a decimal notation in the form MAJOR.MINOR (eg: 3.1). This
13331 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013332
Willy Tarreau74ca5042013-06-11 23:12:07 +020013333 ACL derivatives :
13334 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +010013335
Willy Tarreau47e8eba2013-09-11 23:28:46 +020013336res.len : integer
13337 Returns an integer value corresponding to the number of bytes present in the
13338 response buffer. This is mostly used in ACL. It is important to understand
13339 that this test does not return false as long as the buffer is changing. This
13340 means that a check with equality to zero will almost always immediately match
13341 at the beginning of the session, while a test for more data will wait for
13342 that data to come in and return false only when haproxy is certain that no
13343 more data will come in. This test was designed to be used with TCP response
13344 content inspection.
13345
Willy Tarreau74ca5042013-06-11 23:12:07 +020013346res.payload(<offset>,<length>) : binary
13347 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020013348 in the response buffer. As a special case, if the <length> argument is zero,
13349 the the whole buffer from <offset> to the end is extracted. This can be used
13350 with ACLs in order to check for the presence of some content in a buffer at
13351 any location.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013352
Willy Tarreau74ca5042013-06-11 23:12:07 +020013353res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
13354 This extracts a binary block whose size is specified at <offset1> for <length>
13355 bytes, and which starts at <offset2> if specified or just after the length in
13356 the response buffer. The <offset2> parameter also supports relative offsets
13357 if prepended with a '+' or '-' sign.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013358
Willy Tarreau74ca5042013-06-11 23:12:07 +020013359 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013360
Willy Tarreau971f7b62015-09-29 14:06:59 +020013361res.ssl_hello_type : integer
13362rep_ssl_hello_type : integer (deprecated)
13363 Returns an integer value containing the type of the SSL hello message found
13364 in the response buffer if the buffer contains data that parses as a complete
13365 SSL (v3 or superior) hello message. Note that this only applies to raw
13366 contents found in the response buffer and not to contents deciphered via an
13367 SSL data layer, so this will not work with "server" lines having the "ssl"
13368 option. This is mostly used in ACL to detect presence of an SSL hello message
13369 that is supposed to contain an SSL session ID usable for stickiness.
13370
Willy Tarreau74ca5042013-06-11 23:12:07 +020013371wait_end : boolean
13372 This fetch either returns true when the inspection period is over, or does
13373 not fetch. It is only used in ACLs, in conjunction with content analysis to
13374 avoid returning a wrong verdict early. It may also be used to delay some
13375 actions, such as a delayed reject for some special addresses. Since it either
13376 stops the rules evaluation or immediately returns true, it is recommended to
13377 use this acl as the last one in a rule. Please note that the default ACL
13378 "WAIT_END" is always usable without prior declaration. This test was designed
13379 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013380
Willy Tarreau74ca5042013-06-11 23:12:07 +020013381 Examples :
13382 # delay every incoming request by 2 seconds
13383 tcp-request inspect-delay 2s
13384 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +010013385
Willy Tarreau74ca5042013-06-11 23:12:07 +020013386 # don't immediately tell bad guys they are rejected
13387 tcp-request inspect-delay 10s
13388 acl goodguys src 10.0.0.0/24
13389 acl badguys src 10.0.1.0/24
13390 tcp-request content accept if goodguys
13391 tcp-request content reject if badguys WAIT_END
13392 tcp-request content reject
13393
13394
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200133957.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +020013396--------------------------------------
13397
13398It is possible to fetch samples from HTTP contents, requests and responses.
13399This application layer is also called layer 7. It is only possible to fetch the
13400data in this section when a full HTTP request or response has been parsed from
13401its respective request or response buffer. This is always the case with all
13402HTTP specific rules and for sections running with "mode http". When using TCP
13403content inspection, it may be necessary to support an inspection delay in order
13404to let the request or response come in first. These fetches may require a bit
13405more CPU resources than the layer 4 ones, but not much since the request and
13406response are indexed.
13407
13408base : string
13409 This returns the concatenation of the first Host header and the path part of
13410 the request, which starts at the first slash and ends before the question
13411 mark. It can be useful in virtual hosted environments to detect URL abuses as
13412 well as to improve shared caches efficiency. Using this with a limited size
13413 stick table also allows one to collect statistics about most commonly
13414 requested objects by host/path. With ACLs it can allow simple content
13415 switching rules involving the host and the path at the same time, such as
13416 "www.example.com/favicon.ico". See also "path" and "uri".
13417
13418 ACL derivatives :
13419 base : exact string match
13420 base_beg : prefix match
13421 base_dir : subdir match
13422 base_dom : domain match
13423 base_end : suffix match
13424 base_len : length match
13425 base_reg : regex match
13426 base_sub : substring match
13427
13428base32 : integer
13429 This returns a 32-bit hash of the value returned by the "base" fetch method
13430 above. This is useful to track per-URL activity on high traffic sites without
13431 having to store all URLs. Instead a shorter hash is stored, saving a lot of
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013432 memory. The output type is an unsigned integer. The hash function used is
13433 SDBM with full avalanche on the output. Technically, base32 is exactly equal
13434 to "base,sdbm(1)".
Willy Tarreau74ca5042013-06-11 23:12:07 +020013435
13436base32+src : binary
13437 This returns the concatenation of the base32 fetch above and the src fetch
13438 below. The resulting type is of type binary, with a size of 8 or 20 bytes
13439 depending on the source address family. This can be used to track per-IP,
13440 per-URL counters.
13441
William Lallemand65ad6e12014-01-31 15:08:02 +010013442capture.req.hdr(<idx>) : string
13443 This extracts the content of the header captured by the "capture request
13444 header", idx is the position of the capture keyword in the configuration.
13445 The first entry is an index of 0. See also: "capture request header".
13446
13447capture.req.method : string
13448 This extracts the METHOD of an HTTP request. It can be used in both request
13449 and response. Unlike "method", it can be used in both request and response
13450 because it's allocated.
13451
13452capture.req.uri : string
13453 This extracts the request's URI, which starts at the first slash and ends
13454 before the first space in the request (without the host part). Unlike "path"
13455 and "url", it can be used in both request and response because it's
13456 allocated.
13457
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020013458capture.req.ver : string
13459 This extracts the request's HTTP version and returns either "HTTP/1.0" or
13460 "HTTP/1.1". Unlike "req.ver", it can be used in both request, response, and
13461 logs because it relies on a persistent flag.
13462
William Lallemand65ad6e12014-01-31 15:08:02 +010013463capture.res.hdr(<idx>) : string
13464 This extracts the content of the header captured by the "capture response
13465 header", idx is the position of the capture keyword in the configuration.
13466 The first entry is an index of 0.
13467 See also: "capture response header"
13468
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020013469capture.res.ver : string
13470 This extracts the response's HTTP version and returns either "HTTP/1.0" or
13471 "HTTP/1.1". Unlike "res.ver", it can be used in logs because it relies on a
13472 persistent flag.
13473
Willy Tarreaua5910cc2015-05-02 00:46:08 +020013474req.body : binary
13475 This returns the HTTP request's available body as a block of data. It
13476 requires that the request body has been buffered made available using
13477 "option http-buffer-request". In case of chunked-encoded body, currently only
13478 the first chunk is analyzed.
13479
Thierry FOURNIER9826c772015-05-20 15:50:54 +020013480req.body_param([<name>) : string
13481 This fetch assumes that the body of the POST request is url-encoded. The user
13482 can check if the "content-type" contains the value
13483 "application/x-www-form-urlencoded". This extracts the first occurrence of the
13484 parameter <name> in the body, which ends before '&'. The parameter name is
13485 case-sensitive. If no name is given, any parameter will match, and the first
13486 one will be returned. The result is a string corresponding to the value of the
13487 parameter <name> as presented in the request body (no URL decoding is
13488 performed). Note that the ACL version of this fetch iterates over multiple
13489 parameters and will iteratively report all parameters values if no name is
13490 given.
13491
Willy Tarreaua5910cc2015-05-02 00:46:08 +020013492req.body_len : integer
13493 This returns the length of the HTTP request's available body in bytes. It may
13494 be lower than the advertised length if the body is larger than the buffer. It
13495 requires that the request body has been buffered made available using
13496 "option http-buffer-request".
13497
13498req.body_size : integer
13499 This returns the advertised length of the HTTP request's body in bytes. It
13500 will represent the advertised Content-Length header, or the size of the first
13501 chunk in case of chunked encoding. In order to parse the chunks, it requires
13502 that the request body has been buffered made available using
13503 "option http-buffer-request".
13504
Willy Tarreau74ca5042013-06-11 23:12:07 +020013505req.cook([<name>]) : string
13506cook([<name>]) : string (deprecated)
13507 This extracts the last occurrence of the cookie name <name> on a "Cookie"
13508 header line from the request, and returns its value as string. If no name is
13509 specified, the first cookie value is returned. When used with ACLs, all
13510 matching cookies are evaluated. Spaces around the name and the value are
13511 ignored as requested by the Cookie header specification (RFC6265). The cookie
13512 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
13513 well return an empty value if it is present. Use the "found" match to detect
13514 presence. Use the res.cook() variant for response cookies sent by the server.
13515
13516 ACL derivatives :
13517 cook([<name>]) : exact string match
13518 cook_beg([<name>]) : prefix match
13519 cook_dir([<name>]) : subdir match
13520 cook_dom([<name>]) : domain match
13521 cook_end([<name>]) : suffix match
13522 cook_len([<name>]) : length match
13523 cook_reg([<name>]) : regex match
13524 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +010013525
Willy Tarreau74ca5042013-06-11 23:12:07 +020013526req.cook_cnt([<name>]) : integer
13527cook_cnt([<name>]) : integer (deprecated)
13528 Returns an integer value representing the number of occurrences of the cookie
13529 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013530
Willy Tarreau74ca5042013-06-11 23:12:07 +020013531req.cook_val([<name>]) : integer
13532cook_val([<name>]) : integer (deprecated)
13533 This extracts the last occurrence of the cookie name <name> on a "Cookie"
13534 header line from the request, and converts its value to an integer which is
13535 returned. If no name is specified, the first cookie value is returned. When
13536 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +020013537
Willy Tarreau74ca5042013-06-11 23:12:07 +020013538cookie([<name>]) : string (deprecated)
13539 This extracts the last occurrence of the cookie name <name> on a "Cookie"
13540 header line from the request, or a "Set-Cookie" header from the response, and
13541 returns its value as a string. A typical use is to get multiple clients
13542 sharing a same profile use the same server. This can be similar to what
Willy Tarreau294d0f02015-08-10 19:40:12 +020013543 "appsession" did with the "request-learn" statement, but with support for
Willy Tarreau74ca5042013-06-11 23:12:07 +020013544 multi-peer synchronization and state keeping across restarts. If no name is
13545 specified, the first cookie value is returned. This fetch should not be used
13546 anymore and should be replaced by req.cook() or res.cook() instead as it
13547 ambiguously uses the direction based on the context where it is used.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013548
Willy Tarreau74ca5042013-06-11 23:12:07 +020013549hdr([<name>[,<occ>]]) : string
13550 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
13551 used on responses. Please refer to these respective fetches for more details.
13552 In case of doubt about the fetch direction, please use the explicit ones.
13553 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013554 unambiguously apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013555
Willy Tarreau74ca5042013-06-11 23:12:07 +020013556req.fhdr(<name>[,<occ>]) : string
13557 This extracts the last occurrence of header <name> in an HTTP request. When
13558 used from an ACL, all occurrences are iterated over until a match is found.
13559 Optionally, a specific occurrence might be specified as a position number.
13560 Positive values indicate a position from the first occurrence, with 1 being
13561 the first one. Negative values indicate positions relative to the last one,
13562 with -1 being the last one. It differs from req.hdr() in that any commas
13563 present in the value are returned and are not used as delimiters. This is
13564 sometimes useful with headers such as User-Agent.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013565
Willy Tarreau74ca5042013-06-11 23:12:07 +020013566req.fhdr_cnt([<name>]) : integer
13567 Returns an integer value representing the number of occurrences of request
13568 header field name <name>, or the total number of header fields if <name> is
13569 not specified. Contrary to its req.hdr_cnt() cousin, this function returns
13570 the number of full line headers and does not stop on commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013571
Willy Tarreau74ca5042013-06-11 23:12:07 +020013572req.hdr([<name>[,<occ>]]) : string
13573 This extracts the last occurrence of header <name> in an HTTP request. When
13574 used from an ACL, all occurrences are iterated over until a match is found.
13575 Optionally, a specific occurrence might be specified as a position number.
13576 Positive values indicate a position from the first occurrence, with 1 being
13577 the first one. Negative values indicate positions relative to the last one,
13578 with -1 being the last one. A typical use is with the X-Forwarded-For header
13579 once converted to IP, associated with an IP stick-table. The function
13580 considers any comma as a delimiter for distinct values. If full-line headers
13581 are desired instead, use req.fhdr(). Please carefully check RFC2616 to know
13582 how certain headers are supposed to be parsed. Also, some of them are case
13583 insensitive (eg: Connection).
Willy Tarreaud63335a2010-02-26 12:56:52 +010013584
Willy Tarreau74ca5042013-06-11 23:12:07 +020013585 ACL derivatives :
13586 hdr([<name>[,<occ>]]) : exact string match
13587 hdr_beg([<name>[,<occ>]]) : prefix match
13588 hdr_dir([<name>[,<occ>]]) : subdir match
13589 hdr_dom([<name>[,<occ>]]) : domain match
13590 hdr_end([<name>[,<occ>]]) : suffix match
13591 hdr_len([<name>[,<occ>]]) : length match
13592 hdr_reg([<name>[,<occ>]]) : regex match
13593 hdr_sub([<name>[,<occ>]]) : substring match
13594
13595req.hdr_cnt([<name>]) : integer
13596hdr_cnt([<header>]) : integer (deprecated)
13597 Returns an integer value representing the number of occurrences of request
13598 header field name <name>, or the total number of header field values if
13599 <name> is not specified. It is important to remember that one header line may
13600 count as several headers if it has several values. The function considers any
13601 comma as a delimiter for distinct values. If full-line headers are desired
13602 instead, req.fhdr_cnt() should be used instead. With ACLs, it can be used to
13603 detect presence, absence or abuse of a specific header, as well as to block
13604 request smuggling attacks by rejecting requests which contain more than one
13605 of certain headers. See "req.hdr" for more information on header matching.
13606
13607req.hdr_ip([<name>[,<occ>]]) : ip
13608hdr_ip([<name>[,<occ>]]) : ip (deprecated)
13609 This extracts the last occurrence of header <name> in an HTTP request,
13610 converts it to an IPv4 or IPv6 address and returns this address. When used
13611 with ACLs, all occurrences are checked, and if <name> is omitted, every value
13612 of every header is checked. Optionally, a specific occurrence might be
13613 specified as a position number. Positive values indicate a position from the
13614 first occurrence, with 1 being the first one. Negative values indicate
13615 positions relative to the last one, with -1 being the last one. A typical use
13616 is with the X-Forwarded-For and X-Client-IP headers.
13617
13618req.hdr_val([<name>[,<occ>]]) : integer
13619hdr_val([<name>[,<occ>]]) : integer (deprecated)
13620 This extracts the last occurrence of header <name> in an HTTP request, and
13621 converts it to an integer value. When used with ACLs, all occurrences are
13622 checked, and if <name> is omitted, every value of every header is checked.
13623 Optionally, a specific occurrence might be specified as a position number.
13624 Positive values indicate a position from the first occurrence, with 1 being
13625 the first one. Negative values indicate positions relative to the last one,
13626 with -1 being the last one. A typical use is with the X-Forwarded-For header.
13627
13628http_auth(<userlist>) : boolean
13629 Returns a boolean indicating whether the authentication data received from
13630 the client match a username & password stored in the specified userlist. This
13631 fetch function is not really useful outside of ACLs. Currently only http
13632 basic auth is supported.
13633
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010013634http_auth_group(<userlist>) : string
13635 Returns a string corresponding to the user name found in the authentication
13636 data received from the client if both the user name and password are valid
13637 according to the specified userlist. The main purpose is to use it in ACLs
13638 where it is then checked whether the user belongs to any group within a list.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013639 This fetch function is not really useful outside of ACLs. Currently only http
13640 basic auth is supported.
13641
13642 ACL derivatives :
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010013643 http_auth_group(<userlist>) : group ...
13644 Returns true when the user extracted from the request and whose password is
13645 valid according to the specified userlist belongs to at least one of the
13646 groups.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013647
13648http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020013649 Returns true when the request being processed is the first one of the
13650 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020013651 from some requests when a request is not the first one, or to help grouping
13652 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020013653
Willy Tarreau74ca5042013-06-11 23:12:07 +020013654method : integer + string
13655 Returns an integer value corresponding to the method in the HTTP request. For
13656 example, "GET" equals 1 (check sources to establish the matching). Value 9
13657 means "other method" and may be converted to a string extracted from the
13658 stream. This should not be used directly as a sample, this is only meant to
13659 be used from ACLs, which transparently convert methods from patterns to these
13660 integer + string values. Some predefined ACL already check for most common
13661 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013662
Willy Tarreau74ca5042013-06-11 23:12:07 +020013663 ACL derivatives :
13664 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020013665
Willy Tarreau74ca5042013-06-11 23:12:07 +020013666 Example :
13667 # only accept GET and HEAD requests
13668 acl valid_method method GET HEAD
13669 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020013670
Willy Tarreau74ca5042013-06-11 23:12:07 +020013671path : string
13672 This extracts the request's URL path, which starts at the first slash and
13673 ends before the question mark (without the host part). A typical use is with
13674 prefetch-capable caches, and with portals which need to aggregate multiple
13675 information from databases and keep them in caches. Note that with outgoing
13676 caches, it would be wiser to use "url" instead. With ACLs, it's typically
13677 used to match exact file names (eg: "/login.php"), or directory parts using
13678 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013679
Willy Tarreau74ca5042013-06-11 23:12:07 +020013680 ACL derivatives :
13681 path : exact string match
13682 path_beg : prefix match
13683 path_dir : subdir match
13684 path_dom : domain match
13685 path_end : suffix match
13686 path_len : length match
13687 path_reg : regex match
13688 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020013689
Willy Tarreau49ad95c2015-01-19 15:06:26 +010013690query : string
13691 This extracts the request's query string, which starts after the first
13692 question mark. If no question mark is present, this fetch returns nothing. If
13693 a question mark is present but nothing follows, it returns an empty string.
13694 This means it's possible to easily know whether a query string is present
13695 using the "found" matching method. This fetch is the completemnt of "path"
13696 which stops before the question mark.
13697
Willy Tarreaueb27ec72015-02-20 13:55:29 +010013698req.hdr_names([<delim>]) : string
13699 This builds a string made from the concatenation of all header names as they
13700 appear in the request when the rule is evaluated. The default delimiter is
13701 the comma (',') but it may be overridden as an optional argument <delim>. In
13702 this case, only the first character of <delim> is considered.
13703
Willy Tarreau74ca5042013-06-11 23:12:07 +020013704req.ver : string
13705req_ver : string (deprecated)
13706 Returns the version string from the HTTP request, for example "1.1". This can
13707 be useful for logs, but is mostly there for ACL. Some predefined ACL already
13708 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013709
Willy Tarreau74ca5042013-06-11 23:12:07 +020013710 ACL derivatives :
13711 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020013712
Willy Tarreau74ca5042013-06-11 23:12:07 +020013713res.comp : boolean
13714 Returns the boolean "true" value if the response has been compressed by
13715 HAProxy, otherwise returns boolean "false". This may be used to add
13716 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013717
Willy Tarreau74ca5042013-06-11 23:12:07 +020013718res.comp_algo : string
13719 Returns a string containing the name of the algorithm used if the response
13720 was compressed by HAProxy, for example : "deflate". This may be used to add
13721 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013722
Willy Tarreau74ca5042013-06-11 23:12:07 +020013723res.cook([<name>]) : string
13724scook([<name>]) : string (deprecated)
13725 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
13726 header line from the response, and returns its value as string. If no name is
13727 specified, the first cookie value is returned.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020013728
Willy Tarreau74ca5042013-06-11 23:12:07 +020013729 ACL derivatives :
13730 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020013731
Willy Tarreau74ca5042013-06-11 23:12:07 +020013732res.cook_cnt([<name>]) : integer
13733scook_cnt([<name>]) : integer (deprecated)
13734 Returns an integer value representing the number of occurrences of the cookie
13735 <name> in the response, or all cookies if <name> is not specified. This is
13736 mostly useful when combined with ACLs to detect suspicious responses.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013737
Willy Tarreau74ca5042013-06-11 23:12:07 +020013738res.cook_val([<name>]) : integer
13739scook_val([<name>]) : integer (deprecated)
13740 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
13741 header line from the response, and converts its value to an integer which is
13742 returned. If no name is specified, the first cookie value is returned.
Willy Tarreaud63335a2010-02-26 12:56:52 +010013743
Willy Tarreau74ca5042013-06-11 23:12:07 +020013744res.fhdr([<name>[,<occ>]]) : string
13745 This extracts the last occurrence of header <name> in an HTTP response, or of
13746 the last header if no <name> is specified. Optionally, a specific occurrence
13747 might be specified as a position number. Positive values indicate a position
13748 from the first occurrence, with 1 being the first one. Negative values
13749 indicate positions relative to the last one, with -1 being the last one. It
13750 differs from res.hdr() in that any commas present in the value are returned
13751 and are not used as delimiters. If this is not desired, the res.hdr() fetch
13752 should be used instead. This is sometimes useful with headers such as Date or
13753 Expires.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013754
Willy Tarreau74ca5042013-06-11 23:12:07 +020013755res.fhdr_cnt([<name>]) : integer
13756 Returns an integer value representing the number of occurrences of response
13757 header field name <name>, or the total number of header fields if <name> is
13758 not specified. Contrary to its res.hdr_cnt() cousin, this function returns
13759 the number of full line headers and does not stop on commas. If this is not
13760 desired, the res.hdr_cnt() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013761
Willy Tarreau74ca5042013-06-11 23:12:07 +020013762res.hdr([<name>[,<occ>]]) : string
13763shdr([<name>[,<occ>]]) : string (deprecated)
13764 This extracts the last occurrence of header <name> in an HTTP response, or of
13765 the last header if no <name> is specified. Optionally, a specific occurrence
13766 might be specified as a position number. Positive values indicate a position
13767 from the first occurrence, with 1 being the first one. Negative values
13768 indicate positions relative to the last one, with -1 being the last one. This
13769 can be useful to learn some data into a stick-table. The function considers
13770 any comma as a delimiter for distinct values. If this is not desired, the
13771 res.fhdr() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013772
Willy Tarreau74ca5042013-06-11 23:12:07 +020013773 ACL derivatives :
13774 shdr([<name>[,<occ>]]) : exact string match
13775 shdr_beg([<name>[,<occ>]]) : prefix match
13776 shdr_dir([<name>[,<occ>]]) : subdir match
13777 shdr_dom([<name>[,<occ>]]) : domain match
13778 shdr_end([<name>[,<occ>]]) : suffix match
13779 shdr_len([<name>[,<occ>]]) : length match
13780 shdr_reg([<name>[,<occ>]]) : regex match
13781 shdr_sub([<name>[,<occ>]]) : substring match
13782
13783res.hdr_cnt([<name>]) : integer
13784shdr_cnt([<name>]) : integer (deprecated)
13785 Returns an integer value representing the number of occurrences of response
13786 header field name <name>, or the total number of header fields if <name> is
13787 not specified. The function considers any comma as a delimiter for distinct
13788 values. If this is not desired, the res.fhdr_cnt() fetch should be used
13789 instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013790
Willy Tarreau74ca5042013-06-11 23:12:07 +020013791res.hdr_ip([<name>[,<occ>]]) : ip
13792shdr_ip([<name>[,<occ>]]) : ip (deprecated)
13793 This extracts the last occurrence of header <name> in an HTTP response,
13794 convert it to an IPv4 or IPv6 address and returns this address. Optionally, a
13795 specific occurrence might be specified as a position number. Positive values
13796 indicate a position from the first occurrence, with 1 being the first one.
13797 Negative values indicate positions relative to the last one, with -1 being
13798 the last one. This can be useful to learn some data into a stick table.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013799
Willy Tarreaueb27ec72015-02-20 13:55:29 +010013800res.hdr_names([<delim>]) : string
13801 This builds a string made from the concatenation of all header names as they
13802 appear in the response when the rule is evaluated. The default delimiter is
13803 the comma (',') but it may be overridden as an optional argument <delim>. In
13804 this case, only the first character of <delim> is considered.
13805
Willy Tarreau74ca5042013-06-11 23:12:07 +020013806res.hdr_val([<name>[,<occ>]]) : integer
13807shdr_val([<name>[,<occ>]]) : integer (deprecated)
13808 This extracts the last occurrence of header <name> in an HTTP response, and
13809 converts it to an integer value. Optionally, a specific occurrence might be
13810 specified as a position number. Positive values indicate a position from the
13811 first occurrence, with 1 being the first one. Negative values indicate
13812 positions relative to the last one, with -1 being the last one. This can be
13813 useful to learn some data into a stick table.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010013814
Willy Tarreau74ca5042013-06-11 23:12:07 +020013815res.ver : string
13816resp_ver : string (deprecated)
13817 Returns the version string from the HTTP response, for example "1.1". This
13818 can be useful for logs, but is mostly there for ACL.
Willy Tarreau0e698542011-09-16 08:32:32 +020013819
Willy Tarreau74ca5042013-06-11 23:12:07 +020013820 ACL derivatives :
13821 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010013822
Willy Tarreau74ca5042013-06-11 23:12:07 +020013823set-cookie([<name>]) : string (deprecated)
13824 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
13825 header line from the response and uses the corresponding value to match. This
Willy Tarreau294d0f02015-08-10 19:40:12 +020013826 can be comparable to what "appsession" did with default options, but with
Willy Tarreau74ca5042013-06-11 23:12:07 +020013827 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010013828
Willy Tarreau74ca5042013-06-11 23:12:07 +020013829 This fetch function is deprecated and has been superseded by the "res.cook"
13830 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010013831
Willy Tarreau74ca5042013-06-11 23:12:07 +020013832status : integer
13833 Returns an integer containing the HTTP status code in the HTTP response, for
13834 example, 302. It is mostly used within ACLs and integer ranges, for example,
13835 to remove any Location header if the response is not a 3xx.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013836
Willy Tarreau74ca5042013-06-11 23:12:07 +020013837url : string
13838 This extracts the request's URL as presented in the request. A typical use is
13839 with prefetch-capable caches, and with portals which need to aggregate
13840 multiple information from databases and keep them in caches. With ACLs, using
13841 "path" is preferred over using "url", because clients may send a full URL as
13842 is normally done with proxies. The only real use is to match "*" which does
13843 not match in "path", and for which there is already a predefined ACL. See
13844 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013845
Willy Tarreau74ca5042013-06-11 23:12:07 +020013846 ACL derivatives :
13847 url : exact string match
13848 url_beg : prefix match
13849 url_dir : subdir match
13850 url_dom : domain match
13851 url_end : suffix match
13852 url_len : length match
13853 url_reg : regex match
13854 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013855
Willy Tarreau74ca5042013-06-11 23:12:07 +020013856url_ip : ip
13857 This extracts the IP address from the request's URL when the host part is
13858 presented as an IP address. Its use is very limited. For instance, a
13859 monitoring system might use this field as an alternative for the source IP in
13860 order to test what path a given source address would follow, or to force an
13861 entry in a table for a given source address. With ACLs it can be used to
13862 restrict access to certain systems through a proxy, for example when combined
13863 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013864
Willy Tarreau74ca5042013-06-11 23:12:07 +020013865url_port : integer
13866 This extracts the port part from the request's URL. Note that if the port is
13867 not specified in the request, port 80 is assumed. With ACLs it can be used to
13868 restrict access to certain systems through a proxy, for example when combined
13869 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013870
Willy Tarreau1ede1da2015-05-07 16:06:18 +020013871urlp([<name>[,<delim>]]) : string
13872url_param([<name>[,<delim>]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020013873 This extracts the first occurrence of the parameter <name> in the query
13874 string, which begins after either '?' or <delim>, and which ends before '&',
Willy Tarreau1ede1da2015-05-07 16:06:18 +020013875 ';' or <delim>. The parameter name is case-sensitive. If no name is given,
13876 any parameter will match, and the first one will be returned. The result is
13877 a string corresponding to the value of the parameter <name> as presented in
13878 the request (no URL decoding is performed). This can be used for session
Willy Tarreau74ca5042013-06-11 23:12:07 +020013879 stickiness based on a client ID, to extract an application cookie passed as a
13880 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
Willy Tarreau1ede1da2015-05-07 16:06:18 +020013881 this fetch iterates over multiple parameters and will iteratively report all
13882 parameters values if no name is given
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013883
Willy Tarreau74ca5042013-06-11 23:12:07 +020013884 ACL derivatives :
13885 urlp(<name>[,<delim>]) : exact string match
13886 urlp_beg(<name>[,<delim>]) : prefix match
13887 urlp_dir(<name>[,<delim>]) : subdir match
13888 urlp_dom(<name>[,<delim>]) : domain match
13889 urlp_end(<name>[,<delim>]) : suffix match
13890 urlp_len(<name>[,<delim>]) : length match
13891 urlp_reg(<name>[,<delim>]) : regex match
13892 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013893
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013894
Willy Tarreau74ca5042013-06-11 23:12:07 +020013895 Example :
13896 # match http://example.com/foo?PHPSESSIONID=some_id
13897 stick on urlp(PHPSESSIONID)
13898 # match http://example.com/foo;JSESSIONID=some_id
13899 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020013900
Willy Tarreau1ede1da2015-05-07 16:06:18 +020013901urlp_val([<name>[,<delim>])] : integer
Willy Tarreau74ca5042013-06-11 23:12:07 +020013902 See "urlp" above. This one extracts the URL parameter <name> in the request
13903 and converts it to an integer value. This can be used for session stickiness
13904 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020013905
Willy Tarreau198a7442008-01-17 12:05:32 +010013906
Willy Tarreau74ca5042013-06-11 23:12:07 +0200139077.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013908---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010013909
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013910Some predefined ACLs are hard-coded so that they do not have to be declared in
13911every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020013912order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010013913
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013914ACL name Equivalent to Usage
13915---------------+-----------------------------+---------------------------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013916FALSE always_false never match
Willy Tarreau2492d5b2009-07-11 00:06:00 +020013917HTTP req_proto_http match if protocol is valid HTTP
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013918HTTP_1.0 req_ver 1.0 match HTTP version 1.0
13919HTTP_1.1 req_ver 1.1 match HTTP version 1.1
Willy Tarreaud63335a2010-02-26 12:56:52 +010013920HTTP_CONTENT hdr_val(content-length) gt 0 match an existing content-length
13921HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
13922HTTP_URL_SLASH url_beg / match URL beginning with "/"
13923HTTP_URL_STAR url * match URL equal to "*"
13924LOCALHOST src 127.0.0.1/8 match connection from local host
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013925METH_CONNECT method CONNECT match HTTP CONNECT method
13926METH_GET method GET HEAD match HTTP GET or HEAD method
13927METH_HEAD method HEAD match HTTP HEAD method
13928METH_OPTIONS method OPTIONS match HTTP OPTIONS method
13929METH_POST method POST match HTTP POST method
13930METH_TRACE method TRACE match HTTP TRACE method
Emeric Brunbede3d02009-06-30 17:54:00 +020013931RDP_COOKIE req_rdp_cookie_cnt gt 0 match presence of an RDP cookie
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013932REQ_CONTENT req_len gt 0 match data in the request buffer
Willy Tarreaud63335a2010-02-26 12:56:52 +010013933TRUE always_true always match
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013934WAIT_END wait_end wait for end of content analysis
13935---------------+-----------------------------+---------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010013936
Willy Tarreaub937b7e2010-01-12 15:27:54 +010013937
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200139388. Logging
13939----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010013940
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013941One of HAProxy's strong points certainly lies is its precise logs. It probably
13942provides the finest level of information available for such a product, which is
13943very important for troubleshooting complex environments. Standard information
13944provided in logs include client ports, TCP/HTTP state timers, precise session
13945state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010013946to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013947headers.
13948
13949In order to improve administrators reactivity, it offers a great transparency
13950about encountered problems, both internal and external, and it is possible to
13951send logs to different sources at the same time with different level filters :
13952
13953 - global process-level logs (system errors, start/stop, etc..)
13954 - per-instance system and internal errors (lack of resource, bugs, ...)
13955 - per-instance external troubles (servers up/down, max connections)
13956 - per-instance activity (client connections), either at the establishment or
13957 at the termination.
Jim Freeman9e8714b2015-05-26 09:16:34 -060013958 - per-request control of log-level, eg:
13959 http-request set-log-level silent if sensitive_request
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013960
13961The ability to distribute different levels of logs to different log servers
13962allow several production teams to interact and to fix their problems as soon
13963as possible. For example, the system team might monitor system-wide errors,
13964while the application team might be monitoring the up/down for their servers in
13965real time, and the security team might analyze the activity logs with one hour
13966delay.
13967
13968
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200139698.1. Log levels
13970---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013971
Simon Hormandf791f52011-05-29 15:01:10 +090013972TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013973source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090013974HTTP request, HTTP return code, number of bytes transmitted, conditions
13975in which the session ended, and even exchanged cookies values. For example
13976track a particular user's problems. All messages may be sent to up to two
13977syslog servers. Check the "log" keyword in section 4.2 for more information
13978about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013979
13980
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200139818.2. Log formats
13982----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013983
William Lallemand48940402012-01-30 16:47:22 +010013984HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090013985and will be detailed in the following sections. A few of them may vary
13986slightly with the configuration, due to indicators specific to certain
13987options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013988
13989 - the default format, which is very basic and very rarely used. It only
13990 provides very basic information about the incoming connection at the moment
13991 it is accepted : source IP:port, destination IP:port, and frontend-name.
13992 This mode will eventually disappear so it will not be described to great
13993 extents.
13994
13995 - the TCP format, which is more advanced. This format is enabled when "option
13996 tcplog" is set on the frontend. HAProxy will then usually wait for the
13997 connection to terminate before logging. This format provides much richer
13998 information, such as timers, connection counts, queue size, etc... This
13999 format is recommended for pure TCP proxies.
14000
14001 - the HTTP format, which is the most advanced for HTTP proxying. This format
14002 is enabled when "option httplog" is set on the frontend. It provides the
14003 same information as the TCP format with some HTTP-specific fields such as
14004 the request, the status code, and captures of headers and cookies. This
14005 format is recommended for HTTP proxies.
14006
Emeric Brun3a058f32009-06-30 18:26:00 +020014007 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
14008 fields arranged in the same order as the CLF format. In this mode, all
14009 timers, captures, flags, etc... appear one per field after the end of the
14010 common fields, in the same order they appear in the standard HTTP format.
14011
William Lallemand48940402012-01-30 16:47:22 +010014012 - the custom log format, allows you to make your own log line.
14013
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014014Next sections will go deeper into details for each of these formats. Format
14015specification will be performed on a "field" basis. Unless stated otherwise, a
14016field is a portion of text delimited by any number of spaces. Since syslog
14017servers are susceptible of inserting fields at the beginning of a line, it is
14018always assumed that the first field is the one containing the process name and
14019identifier.
14020
14021Note : Since log lines may be quite long, the log examples in sections below
14022 might be broken into multiple lines. The example log lines will be
14023 prefixed with 3 closing angle brackets ('>>>') and each time a log is
14024 broken into multiple lines, each non-final line will end with a
14025 backslash ('\') and the next line will start indented by two characters.
14026
14027
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200140288.2.1. Default log format
14029-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014030
14031This format is used when no specific option is set. The log is emitted as soon
14032as the connection is accepted. One should note that this currently is the only
14033format which logs the request's destination IP and ports.
14034
14035 Example :
14036 listen www
14037 mode http
14038 log global
14039 server srv1 127.0.0.1:8000
14040
14041 >>> Feb 6 12:12:09 localhost \
14042 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
14043 (www/HTTP)
14044
14045 Field Format Extract from the example above
14046 1 process_name '[' pid ']:' haproxy[14385]:
14047 2 'Connect from' Connect from
14048 3 source_ip ':' source_port 10.0.1.2:33312
14049 4 'to' to
14050 5 destination_ip ':' destination_port 10.0.3.31:8012
14051 6 '(' frontend_name '/' mode ')' (www/HTTP)
14052
14053Detailed fields description :
14054 - "source_ip" is the IP address of the client which initiated the connection.
14055 - "source_port" is the TCP port of the client which initiated the connection.
14056 - "destination_ip" is the IP address the client connected to.
14057 - "destination_port" is the TCP port the client connected to.
14058 - "frontend_name" is the name of the frontend (or listener) which received
14059 and processed the connection.
14060 - "mode is the mode the frontend is operating (TCP or HTTP).
14061
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014062In case of a UNIX socket, the source and destination addresses are marked as
14063"unix:" and the ports reflect the internal ID of the socket which accepted the
14064connection (the same ID as reported in the stats).
14065
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014066It is advised not to use this deprecated format for newer installations as it
14067will eventually disappear.
14068
14069
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200140708.2.2. TCP log format
14071---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014072
14073The TCP format is used when "option tcplog" is specified in the frontend, and
14074is the recommended format for pure TCP proxies. It provides a lot of precious
14075information for troubleshooting. Since this format includes timers and byte
14076counts, the log is normally emitted at the end of the session. It can be
14077emitted earlier if "option logasap" is specified, which makes sense in most
14078environments with long sessions such as remote terminals. Sessions which match
14079the "monitor" rules are never logged. It is also possible not to emit logs for
14080sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020014081specifying "option dontlognull" in the frontend. Successful connections will
14082not be logged if "option dontlog-normal" is specified in the frontend. A few
14083fields may slightly vary depending on some configuration options, those are
14084marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014085
14086 Example :
14087 frontend fnt
14088 mode tcp
14089 option tcplog
14090 log global
14091 default_backend bck
14092
14093 backend bck
14094 server srv1 127.0.0.1:8000
14095
14096 >>> Feb 6 12:12:56 localhost \
14097 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
14098 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
14099
14100 Field Format Extract from the example above
14101 1 process_name '[' pid ']:' haproxy[14387]:
14102 2 client_ip ':' client_port 10.0.1.2:33313
14103 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
14104 4 frontend_name fnt
14105 5 backend_name '/' server_name bck/srv1
14106 6 Tw '/' Tc '/' Tt* 0/0/5007
14107 7 bytes_read* 212
14108 8 termination_state --
14109 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
14110 10 srv_queue '/' backend_queue 0/0
14111
14112Detailed fields description :
14113 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014114 connection to haproxy. If the connection was accepted on a UNIX socket
14115 instead, the IP address would be replaced with the word "unix". Note that
14116 when the connection is accepted on a socket configured with "accept-proxy"
14117 and the PROXY protocol is correctly used, then the logs will reflect the
14118 forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014119
14120 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014121 If the connection was accepted on a UNIX socket instead, the port would be
14122 replaced with the ID of the accepting socket, which is also reported in the
14123 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014124
14125 - "accept_date" is the exact date when the connection was received by haproxy
14126 (which might be very slightly different from the date observed on the
14127 network if there was some queuing in the system's backlog). This is usually
14128 the same date which may appear in any upstream firewall's log.
14129
14130 - "frontend_name" is the name of the frontend (or listener) which received
14131 and processed the connection.
14132
14133 - "backend_name" is the name of the backend (or listener) which was selected
14134 to manage the connection to the server. This will be the same as the
14135 frontend if no switching rule has been applied, which is common for TCP
14136 applications.
14137
14138 - "server_name" is the name of the last server to which the connection was
14139 sent, which might differ from the first one if there were connection errors
14140 and a redispatch occurred. Note that this server belongs to the backend
14141 which processed the request. If the connection was aborted before reaching
14142 a server, "<NOSRV>" is indicated instead of a server name.
14143
14144 - "Tw" is the total time in milliseconds spent waiting in the various queues.
14145 It can be "-1" if the connection was aborted before reaching the queue.
14146 See "Timers" below for more details.
14147
14148 - "Tc" is the total time in milliseconds spent waiting for the connection to
14149 establish to the final server, including retries. It can be "-1" if the
14150 connection was aborted before a connection could be established. See
14151 "Timers" below for more details.
14152
14153 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014154 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014155 "option logasap" was specified, then the time counting stops at the moment
14156 the log is emitted. In this case, a '+' sign is prepended before the value,
14157 indicating that the final one will be larger. See "Timers" below for more
14158 details.
14159
14160 - "bytes_read" is the total number of bytes transmitted from the server to
14161 the client when the log is emitted. If "option logasap" is specified, the
14162 this value will be prefixed with a '+' sign indicating that the final one
14163 may be larger. Please note that this value is a 64-bit counter, so log
14164 analysis tools must be able to handle it without overflowing.
14165
14166 - "termination_state" is the condition the session was in when the session
14167 ended. This indicates the session state, which side caused the end of
14168 session to happen, and for what reason (timeout, error, ...). The normal
14169 flags should be "--", indicating the session was closed by either end with
14170 no data remaining in buffers. See below "Session state at disconnection"
14171 for more details.
14172
14173 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014174 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014175 limits have been reached. For instance, if actconn is close to 512 when
14176 multiple connection errors occur, chances are high that the system limits
14177 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014178 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014179
14180 - "feconn" is the total number of concurrent connections on the frontend when
14181 the session was logged. It is useful to estimate the amount of resource
14182 required to sustain high loads, and to detect when the frontend's "maxconn"
14183 has been reached. Most often when this value increases by huge jumps, it is
14184 because there is congestion on the backend servers, but sometimes it can be
14185 caused by a denial of service attack.
14186
14187 - "beconn" is the total number of concurrent connections handled by the
14188 backend when the session was logged. It includes the total number of
14189 concurrent connections active on servers as well as the number of
14190 connections pending in queues. It is useful to estimate the amount of
14191 additional servers needed to support high loads for a given application.
14192 Most often when this value increases by huge jumps, it is because there is
14193 congestion on the backend servers, but sometimes it can be caused by a
14194 denial of service attack.
14195
14196 - "srv_conn" is the total number of concurrent connections still active on
14197 the server when the session was logged. It can never exceed the server's
14198 configured "maxconn" parameter. If this value is very often close or equal
14199 to the server's "maxconn", it means that traffic regulation is involved a
14200 lot, meaning that either the server's maxconn value is too low, or that
14201 there aren't enough servers to process the load with an optimal response
14202 time. When only one of the server's "srv_conn" is high, it usually means
14203 that this server has some trouble causing the connections to take longer to
14204 be processed than on other servers.
14205
14206 - "retries" is the number of connection retries experienced by this session
14207 when trying to connect to the server. It must normally be zero, unless a
14208 server is being stopped at the same moment the connection was attempted.
14209 Frequent retries generally indicate either a network problem between
14210 haproxy and the server, or a misconfigured system backlog on the server
14211 preventing new connections from being queued. This field may optionally be
14212 prefixed with a '+' sign, indicating that the session has experienced a
14213 redispatch after the maximal retry count has been reached on the initial
14214 server. In this case, the server name appearing in the log is the one the
14215 connection was redispatched to, and not the first one, though both may
14216 sometimes be the same in case of hashing for instance. So as a general rule
14217 of thumb, when a '+' is present in front of the retry count, this count
14218 should not be attributed to the logged server.
14219
14220 - "srv_queue" is the total number of requests which were processed before
14221 this one in the server queue. It is zero when the request has not gone
14222 through the server queue. It makes it possible to estimate the approximate
14223 server's response time by dividing the time spent in queue by the number of
14224 requests in the queue. It is worth noting that if a session experiences a
14225 redispatch and passes through two server queues, their positions will be
14226 cumulated. A request should not pass through both the server queue and the
14227 backend queue unless a redispatch occurs.
14228
14229 - "backend_queue" is the total number of requests which were processed before
14230 this one in the backend's global queue. It is zero when the request has not
14231 gone through the global queue. It makes it possible to estimate the average
14232 queue length, which easily translates into a number of missing servers when
14233 divided by a server's "maxconn" parameter. It is worth noting that if a
14234 session experiences a redispatch, it may pass twice in the backend's queue,
14235 and then both positions will be cumulated. A request should not pass
14236 through both the server queue and the backend queue unless a redispatch
14237 occurs.
14238
14239
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200142408.2.3. HTTP log format
14241----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014242
14243The HTTP format is the most complete and the best suited for HTTP proxies. It
14244is enabled by when "option httplog" is specified in the frontend. It provides
14245the same level of information as the TCP format with additional features which
14246are specific to the HTTP protocol. Just like the TCP format, the log is usually
14247emitted at the end of the session, unless "option logasap" is specified, which
14248generally only makes sense for download sites. A session which matches the
14249"monitor" rules will never logged. It is also possible not to log sessions for
14250which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020014251frontend. Successful connections will not be logged if "option dontlog-normal"
14252is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014253
14254Most fields are shared with the TCP log, some being different. A few fields may
14255slightly vary depending on some configuration options. Those ones are marked
14256with a star ('*') after the field name below.
14257
14258 Example :
14259 frontend http-in
14260 mode http
14261 option httplog
14262 log global
14263 default_backend bck
14264
14265 backend static
14266 server srv1 127.0.0.1:8000
14267
14268 >>> Feb 6 12:14:14 localhost \
14269 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
14270 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 +010014271 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014272
14273 Field Format Extract from the example above
14274 1 process_name '[' pid ']:' haproxy[14389]:
14275 2 client_ip ':' client_port 10.0.1.2:33317
14276 3 '[' accept_date ']' [06/Feb/2009:12:14:14.655]
14277 4 frontend_name http-in
14278 5 backend_name '/' server_name static/srv1
14279 6 Tq '/' Tw '/' Tc '/' Tr '/' Tt* 10/0/30/69/109
14280 7 status_code 200
14281 8 bytes_read* 2750
14282 9 captured_request_cookie -
14283 10 captured_response_cookie -
14284 11 termination_state ----
14285 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
14286 13 srv_queue '/' backend_queue 0/0
14287 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
14288 15 '{' captured_response_headers* '}' {}
14289 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010014290
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014291
14292Detailed fields description :
14293 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014294 connection to haproxy. If the connection was accepted on a UNIX socket
14295 instead, the IP address would be replaced with the word "unix". Note that
14296 when the connection is accepted on a socket configured with "accept-proxy"
14297 and the PROXY protocol is correctly used, then the logs will reflect the
14298 forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014299
14300 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010014301 If the connection was accepted on a UNIX socket instead, the port would be
14302 replaced with the ID of the accepting socket, which is also reported in the
14303 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014304
14305 - "accept_date" is the exact date when the TCP connection was received by
14306 haproxy (which might be very slightly different from the date observed on
14307 the network if there was some queuing in the system's backlog). This is
14308 usually the same date which may appear in any upstream firewall's log. This
14309 does not depend on the fact that the client has sent the request or not.
14310
14311 - "frontend_name" is the name of the frontend (or listener) which received
14312 and processed the connection.
14313
14314 - "backend_name" is the name of the backend (or listener) which was selected
14315 to manage the connection to the server. This will be the same as the
14316 frontend if no switching rule has been applied.
14317
14318 - "server_name" is the name of the last server to which the connection was
14319 sent, which might differ from the first one if there were connection errors
14320 and a redispatch occurred. Note that this server belongs to the backend
14321 which processed the request. If the request was aborted before reaching a
14322 server, "<NOSRV>" is indicated instead of a server name. If the request was
14323 intercepted by the stats subsystem, "<STATS>" is indicated instead.
14324
14325 - "Tq" is the total time in milliseconds spent waiting for the client to send
14326 a full HTTP request, not counting data. It can be "-1" if the connection
14327 was aborted before a complete request could be received. It should always
14328 be very small because a request generally fits in one single packet. Large
14329 times here generally indicate network trouble between the client and
14330 haproxy. See "Timers" below for more details.
14331
14332 - "Tw" is the total time in milliseconds spent waiting in the various queues.
14333 It can be "-1" if the connection was aborted before reaching the queue.
14334 See "Timers" below for more details.
14335
14336 - "Tc" is the total time in milliseconds spent waiting for the connection to
14337 establish to the final server, including retries. It can be "-1" if the
14338 request was aborted before a connection could be established. See "Timers"
14339 below for more details.
14340
14341 - "Tr" is the total time in milliseconds spent waiting for the server to send
14342 a full HTTP response, not counting data. It can be "-1" if the request was
14343 aborted before a complete response could be received. It generally matches
14344 the server's processing time for the request, though it may be altered by
14345 the amount of data sent by the client to the server. Large times here on
14346 "GET" requests generally indicate an overloaded server. See "Timers" below
14347 for more details.
14348
14349 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014350 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014351 "option logasap" was specified, then the time counting stops at the moment
14352 the log is emitted. In this case, a '+' sign is prepended before the value,
14353 indicating that the final one will be larger. See "Timers" below for more
14354 details.
14355
14356 - "status_code" is the HTTP status code returned to the client. This status
14357 is generally set by the server, but it might also be set by haproxy when
14358 the server cannot be reached or when its response is blocked by haproxy.
14359
14360 - "bytes_read" is the total number of bytes transmitted to the client when
14361 the log is emitted. This does include HTTP headers. If "option logasap" is
14362 specified, the this value will be prefixed with a '+' sign indicating that
14363 the final one may be larger. Please note that this value is a 64-bit
14364 counter, so log analysis tools must be able to handle it without
14365 overflowing.
14366
14367 - "captured_request_cookie" is an optional "name=value" entry indicating that
14368 the client had this cookie in the request. The cookie name and its maximum
14369 length are defined by the "capture cookie" statement in the frontend
14370 configuration. The field is a single dash ('-') when the option is not
14371 set. Only one cookie may be captured, it is generally used to track session
14372 ID exchanges between a client and a server to detect session crossing
14373 between clients due to application bugs. For more details, please consult
14374 the section "Capturing HTTP headers and cookies" below.
14375
14376 - "captured_response_cookie" is an optional "name=value" entry indicating
14377 that the server has returned a cookie with its response. The cookie name
14378 and its maximum length are defined by the "capture cookie" statement in the
14379 frontend configuration. The field is a single dash ('-') when the option is
14380 not set. Only one cookie may be captured, it is generally used to track
14381 session ID exchanges between a client and a server to detect session
14382 crossing between clients due to application bugs. For more details, please
14383 consult the section "Capturing HTTP headers and cookies" below.
14384
14385 - "termination_state" is the condition the session was in when the session
14386 ended. This indicates the session state, which side caused the end of
14387 session to happen, for what reason (timeout, error, ...), just like in TCP
14388 logs, and information about persistence operations on cookies in the last
14389 two characters. The normal flags should begin with "--", indicating the
14390 session was closed by either end with no data remaining in buffers. See
14391 below "Session state at disconnection" for more details.
14392
14393 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014394 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014395 limits have been reached. For instance, if actconn is close to 512 or 1024
14396 when multiple connection errors occur, chances are high that the system
14397 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014398 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014399 system.
14400
14401 - "feconn" is the total number of concurrent connections on the frontend when
14402 the session was logged. It is useful to estimate the amount of resource
14403 required to sustain high loads, and to detect when the frontend's "maxconn"
14404 has been reached. Most often when this value increases by huge jumps, it is
14405 because there is congestion on the backend servers, but sometimes it can be
14406 caused by a denial of service attack.
14407
14408 - "beconn" is the total number of concurrent connections handled by the
14409 backend when the session was logged. It includes the total number of
14410 concurrent connections active on servers as well as the number of
14411 connections pending in queues. It is useful to estimate the amount of
14412 additional servers needed to support high loads for a given application.
14413 Most often when this value increases by huge jumps, it is because there is
14414 congestion on the backend servers, but sometimes it can be caused by a
14415 denial of service attack.
14416
14417 - "srv_conn" is the total number of concurrent connections still active on
14418 the server when the session was logged. It can never exceed the server's
14419 configured "maxconn" parameter. If this value is very often close or equal
14420 to the server's "maxconn", it means that traffic regulation is involved a
14421 lot, meaning that either the server's maxconn value is too low, or that
14422 there aren't enough servers to process the load with an optimal response
14423 time. When only one of the server's "srv_conn" is high, it usually means
14424 that this server has some trouble causing the requests to take longer to be
14425 processed than on other servers.
14426
14427 - "retries" is the number of connection retries experienced by this session
14428 when trying to connect to the server. It must normally be zero, unless a
14429 server is being stopped at the same moment the connection was attempted.
14430 Frequent retries generally indicate either a network problem between
14431 haproxy and the server, or a misconfigured system backlog on the server
14432 preventing new connections from being queued. This field may optionally be
14433 prefixed with a '+' sign, indicating that the session has experienced a
14434 redispatch after the maximal retry count has been reached on the initial
14435 server. In this case, the server name appearing in the log is the one the
14436 connection was redispatched to, and not the first one, though both may
14437 sometimes be the same in case of hashing for instance. So as a general rule
14438 of thumb, when a '+' is present in front of the retry count, this count
14439 should not be attributed to the logged server.
14440
14441 - "srv_queue" is the total number of requests which were processed before
14442 this one in the server queue. It is zero when the request has not gone
14443 through the server queue. It makes it possible to estimate the approximate
14444 server's response time by dividing the time spent in queue by the number of
14445 requests in the queue. It is worth noting that if a session experiences a
14446 redispatch and passes through two server queues, their positions will be
14447 cumulated. A request should not pass through both the server queue and the
14448 backend queue unless a redispatch occurs.
14449
14450 - "backend_queue" is the total number of requests which were processed before
14451 this one in the backend's global queue. It is zero when the request has not
14452 gone through the global queue. It makes it possible to estimate the average
14453 queue length, which easily translates into a number of missing servers when
14454 divided by a server's "maxconn" parameter. It is worth noting that if a
14455 session experiences a redispatch, it may pass twice in the backend's queue,
14456 and then both positions will be cumulated. A request should not pass
14457 through both the server queue and the backend queue unless a redispatch
14458 occurs.
14459
14460 - "captured_request_headers" is a list of headers captured in the request due
14461 to the presence of the "capture request header" statement in the frontend.
14462 Multiple headers can be captured, they will be delimited by a vertical bar
14463 ('|'). When no capture is enabled, the braces do not appear, causing a
14464 shift of remaining fields. It is important to note that this field may
14465 contain spaces, and that using it requires a smarter log parser than when
14466 it's not used. Please consult the section "Capturing HTTP headers and
14467 cookies" below for more details.
14468
14469 - "captured_response_headers" is a list of headers captured in the response
14470 due to the presence of the "capture response header" statement in the
14471 frontend. Multiple headers can be captured, they will be delimited by a
14472 vertical bar ('|'). When no capture is enabled, the braces do not appear,
14473 causing a shift of remaining fields. It is important to note that this
14474 field may contain spaces, and that using it requires a smarter log parser
14475 than when it's not used. Please consult the section "Capturing HTTP headers
14476 and cookies" below for more details.
14477
14478 - "http_request" is the complete HTTP request line, including the method,
14479 request and HTTP version string. Non-printable characters are encoded (see
14480 below the section "Non-printable characters"). This is always the last
14481 field, and it is always delimited by quotes and is the only one which can
14482 contain quotes. If new fields are added to the log format, they will be
14483 added before this field. This field might be truncated if the request is
14484 huge and does not fit in the standard syslog buffer (1024 characters). This
14485 is the reason why this field must always remain the last one.
14486
14487
Cyril Bontédc4d9032012-04-08 21:57:39 +0200144888.2.4. Custom log format
14489------------------------
William Lallemand48940402012-01-30 16:47:22 +010014490
Willy Tarreau2beef582012-12-20 17:22:52 +010014491The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010014492mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010014493
14494HAproxy understands some log format variables. % precedes log format variables.
14495Variables can take arguments using braces ('{}'), and multiple arguments are
14496separated by commas within the braces. Flags may be added or removed by
14497prefixing them with a '+' or '-' sign.
14498
14499Special variable "%o" may be used to propagate its flags to all other
14500variables on the same format string. This is particularly handy with quoted
14501string formats ("Q").
14502
Willy Tarreauc8368452012-12-21 00:09:23 +010014503If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreaube722a22014-06-13 16:31:59 +020014504as a sample expression rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010014505less common information such as the client's SSL certificate's DN, or to log
14506the key that would be used to store an entry into a stick table.
14507
William Lallemand48940402012-01-30 16:47:22 +010014508Note: spaces must be escaped. A space character is considered as a separator.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014509In order to emit a verbatim '%', it must be preceded by another '%' resulting
Willy Tarreau06d97f92013-12-02 17:45:48 +010014510in '%%'. HAProxy will automatically merge consecutive separators.
William Lallemand48940402012-01-30 16:47:22 +010014511
14512Flags are :
14513 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014514 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
William Lallemand48940402012-01-30 16:47:22 +010014515
14516 Example:
14517
14518 log-format %T\ %t\ Some\ Text
14519 log-format %{+Q}o\ %t\ %s\ %{-Q}r
14520
14521At the moment, the default HTTP format is defined this way :
14522
Willy Tarreau2beef582012-12-20 17:22:52 +010014523 log-format %ci:%cp\ [%t]\ %ft\ %b/%s\ %Tq/%Tw/%Tc/%Tr/%Tt\ %ST\ %B\ %CC\ \
14524 %CS\ %tsc\ %ac/%fc/%bc/%sc/%rc\ %sq/%bq\ %hr\ %hs\ %{+Q}r
William Lallemand48940402012-01-30 16:47:22 +010014525
William Lallemandbddd4fd2012-02-27 11:23:10 +010014526the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010014527
Willy Tarreau2beef582012-12-20 17:22:52 +010014528 log-format %{+Q}o\ %{-Q}ci\ -\ -\ [%T]\ %r\ %ST\ %B\ \"\"\ \"\"\ %cp\ \
Willy Tarreau773d65f2012-10-12 14:56:11 +020014529 %ms\ %ft\ %b\ %s\ \%Tq\ %Tw\ %Tc\ %Tr\ %Tt\ %tsc\ %ac\ %fc\ \
Willy Tarreau2beef582012-12-20 17:22:52 +010014530 %bc\ %sc\ %rc\ %sq\ %bq\ %CC\ %CS\ \%hrl\ %hsl
William Lallemand48940402012-01-30 16:47:22 +010014531
William Lallemandbddd4fd2012-02-27 11:23:10 +010014532and the default TCP format is defined this way :
14533
Willy Tarreau2beef582012-12-20 17:22:52 +010014534 log-format %ci:%cp\ [%t]\ %ft\ %b/%s\ %Tw/%Tc/%Tt\ %B\ %ts\ \
William Lallemandbddd4fd2012-02-27 11:23:10 +010014535 %ac/%fc/%bc/%sc/%rc\ %sq/%bq
14536
William Lallemand48940402012-01-30 16:47:22 +010014537Please refer to the table below for currently defined variables :
14538
William Lallemandbddd4fd2012-02-27 11:23:10 +010014539 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014540 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014541 +---+------+-----------------------------------------------+-------------+
14542 | | %o | special variable, apply flags on all next var | |
14543 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010014544 | | %B | bytes_read (from server to client) | numeric |
14545 | H | %CC | captured_request_cookie | string |
14546 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020014547 | | %H | hostname | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000014548 | H | %HM | HTTP method (ex: POST) | string |
14549 | H | %HP | HTTP request URI without query string (path) | string |
Andrew Hayworthe63ac872015-07-31 16:14:16 +000014550 | H | %HQ | HTTP request URI query string (ex: ?bar=baz) | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000014551 | H | %HU | HTTP request URI (ex: /foo?bar=baz) | string |
14552 | H | %HV | HTTP version (ex: HTTP/1.0) | string |
William Lallemanda73203e2012-03-12 12:48:57 +010014553 | | %ID | unique-id | string |
Willy Tarreau4bf99632014-06-13 12:21:40 +020014554 | | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020014555 | | %T | gmt_date_time | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014556 | | %Tc | Tc | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080014557 | | %Tl | local_date_time | date |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014558 | H | %Tq | Tq | numeric |
14559 | H | %Tr | Tr | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020014560 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014561 | | %Tt | Tt | numeric |
14562 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010014563 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014564 | | %ac | actconn | numeric |
14565 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010014566 | | %bc | beconn (backend concurrent connections) | numeric |
14567 | | %bi | backend_source_ip (connecting address) | IP |
14568 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014569 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010014570 | | %ci | client_ip (accepted address) | IP |
14571 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014572 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010014573 | | %fc | feconn (frontend concurrent connections) | numeric |
14574 | | %fi | frontend_ip (accepting address) | IP |
14575 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020014576 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreau7346acb2014-08-28 15:03:15 +020014577 | | %lc | frontend_log_counter | numeric |
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020014578 | | %hr | captured_request_headers default style | string |
14579 | | %hrl | captured_request_headers CLF style | string list |
14580 | | %hs | captured_response_headers default style | string |
14581 | | %hsl | captured_response_headers CLF style | string list |
Willy Tarreau812c88e2015-08-09 10:56:35 +020014582 | | %ms | accept date milliseconds (left-padded with 0) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020014583 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014584 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014585 | | %rc | retries | numeric |
Willy Tarreau1f0da242014-01-25 11:01:50 +010014586 | | %rt | request_counter (HTTP req or TCP session) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014587 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010014588 | | %sc | srv_conn (server concurrent connections) | numeric |
14589 | | %si | server_IP (target address) | IP |
14590 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014591 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014592 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
14593 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010014594 | | %t | date_time (with millisecond resolution) | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014595 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014596 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010014597 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010014598
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020014599 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010014600
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010014601
146028.2.5. Error log format
14603-----------------------
14604
14605When an incoming connection fails due to an SSL handshake or an invalid PROXY
14606protocol header, haproxy will log the event using a shorter, fixed line format.
14607By default, logs are emitted at the LOG_INFO level, unless the option
14608"log-separate-errors" is set in the backend, in which case the LOG_ERR level
14609will be used. Connections on which no data are exchanged (eg: probes) are not
14610logged if the "dontlognull" option is set.
14611
14612The format looks like this :
14613
14614 >>> Dec 3 18:27:14 localhost \
14615 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
14616 Connection error during SSL handshake
14617
14618 Field Format Extract from the example above
14619 1 process_name '[' pid ']:' haproxy[6103]:
14620 2 client_ip ':' client_port 127.0.0.1:56059
14621 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
14622 4 frontend_name "/" bind_name ":" frt/f1:
14623 5 message Connection error during SSL handshake
14624
14625These fields just provide minimal information to help debugging connection
14626failures.
14627
14628
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200146298.3. Advanced logging options
14630-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014631
14632Some advanced logging options are often looked for but are not easy to find out
14633just by looking at the various options. Here is an entry point for the few
14634options which can enable better logging. Please refer to the keywords reference
14635for more information about their usage.
14636
14637
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200146388.3.1. Disabling logging of external tests
14639------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014640
14641It is quite common to have some monitoring tools perform health checks on
14642haproxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
14643commercial load-balancer, and sometimes it will simply be a more complete
14644monitoring system such as Nagios. When the tests are very frequent, users often
14645ask how to disable logging for those checks. There are three possibilities :
14646
14647 - if connections come from everywhere and are just TCP probes, it is often
14648 desired to simply disable logging of connections without data exchange, by
14649 setting "option dontlognull" in the frontend. It also disables logging of
14650 port scans, which may or may not be desired.
14651
14652 - if the connection come from a known source network, use "monitor-net" to
14653 declare this network as monitoring only. Any host in this network will then
14654 only be able to perform health checks, and their requests will not be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014655 logged. This is generally appropriate to designate a list of equipment
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014656 such as other load-balancers.
14657
14658 - if the tests are performed on a known URI, use "monitor-uri" to declare
14659 this URI as dedicated to monitoring. Any host sending this request will
14660 only get the result of a health-check, and the request will not be logged.
14661
14662
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200146638.3.2. Logging before waiting for the session to terminate
14664----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014665
14666The problem with logging at end of connection is that you have no clue about
14667what is happening during very long sessions, such as remote terminal sessions
14668or large file downloads. This problem can be worked around by specifying
14669"option logasap" in the frontend. Haproxy will then log as soon as possible,
14670just before data transfer begins. This means that in case of TCP, it will still
14671log the connection status to the server, and in case of HTTP, it will log just
14672after processing the server headers. In this case, the number of bytes reported
14673is the number of header bytes sent to the client. In order to avoid confusion
14674with normal logs, the total time field and the number of bytes are prefixed
14675with a '+' sign which means that real numbers are certainly larger.
14676
14677
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200146788.3.3. Raising log level upon errors
14679------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020014680
14681Sometimes it is more convenient to separate normal traffic from errors logs,
14682for instance in order to ease error monitoring from log files. When the option
14683"log-separate-errors" is used, connections which experience errors, timeouts,
14684retries, redispatches or HTTP status codes 5xx will see their syslog level
14685raised from "info" to "err". This will help a syslog daemon store the log in
14686a separate file. It is very important to keep the errors in the normal traffic
14687file too, so that log ordering is not altered. You should also be careful if
14688you already have configured your syslog daemon to store all logs higher than
14689"notice" in an "admin" file, because the "err" level is higher than "notice".
14690
14691
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200146928.3.4. Disabling logging of successful connections
14693--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020014694
14695Although this may sound strange at first, some large sites have to deal with
14696multiple thousands of logs per second and are experiencing difficulties keeping
14697them intact for a long time or detecting errors within them. If the option
14698"dontlog-normal" is set on the frontend, all normal connections will not be
14699logged. In this regard, a normal connection is defined as one without any
14700error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
14701and a response with a status 5xx is not considered normal and will be logged
14702too. Of course, doing is is really discouraged as it will remove most of the
14703useful information from the logs. Do this only if you have no other
14704alternative.
14705
14706
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200147078.4. Timing events
14708------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014709
14710Timers provide a great help in troubleshooting network problems. All values are
14711reported in milliseconds (ms). These timers should be used in conjunction with
14712the session termination flags. In TCP mode with "option tcplog" set on the
14713frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
14714mode, 5 control points are reported under the form "Tq/Tw/Tc/Tr/Tt" :
14715
14716 - Tq: total time to get the client request (HTTP mode only). It's the time
14717 elapsed between the moment the client connection was accepted and the
14718 moment the proxy received the last HTTP header. The value "-1" indicates
14719 that the end of headers (empty line) has never been seen. This happens when
14720 the client closes prematurely or times out.
14721
14722 - Tw: total time spent in the queues waiting for a connection slot. It
14723 accounts for backend queue as well as the server queues, and depends on the
14724 queue size, and the time needed for the server to complete previous
14725 requests. The value "-1" means that the request was killed before reaching
14726 the queue, which is generally what happens with invalid or denied requests.
14727
14728 - Tc: total time to establish the TCP connection to the server. It's the time
14729 elapsed between the moment the proxy sent the connection request, and the
14730 moment it was acknowledged by the server, or between the TCP SYN packet and
14731 the matching SYN/ACK packet in return. The value "-1" means that the
14732 connection never established.
14733
14734 - Tr: server response time (HTTP mode only). It's the time elapsed between
14735 the moment the TCP connection was established to the server and the moment
14736 the server sent its complete response headers. It purely shows its request
14737 processing time, without the network overhead due to the data transmission.
14738 It is worth noting that when the client has data to send to the server, for
14739 instance during a POST request, the time already runs, and this can distort
14740 apparent response time. For this reason, it's generally wise not to trust
14741 too much this field for POST requests initiated from clients behind an
14742 untrusted network. A value of "-1" here means that the last the response
14743 header (empty line) was never seen, most likely because the server timeout
14744 stroke before the server managed to process the request.
14745
14746 - Tt: total session duration time, between the moment the proxy accepted it
14747 and the moment both ends were closed. The exception is when the "logasap"
14748 option is specified. In this case, it only equals (Tq+Tw+Tc+Tr), and is
14749 prefixed with a '+' sign. From this field, we can deduce "Td", the data
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014750 transmission time, by subtracting other timers when valid :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014751
14752 Td = Tt - (Tq + Tw + Tc + Tr)
14753
14754 Timers with "-1" values have to be excluded from this equation. In TCP
14755 mode, "Tq" and "Tr" have to be excluded too. Note that "Tt" can never be
14756 negative.
14757
14758These timers provide precious indications on trouble causes. Since the TCP
14759protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
14760that timers close to multiples of 3s are nearly always related to lost packets
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010014761due to network problems (wires, negotiation, congestion). Moreover, if "Tt" is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014762close to a timeout value specified in the configuration, it often means that a
14763session has been aborted on timeout.
14764
14765Most common cases :
14766
14767 - If "Tq" is close to 3000, a packet has probably been lost between the
14768 client and the proxy. This is very rare on local networks but might happen
14769 when clients are on far remote networks and send large requests. It may
14770 happen that values larger than usual appear here without any network cause.
14771 Sometimes, during an attack or just after a resource starvation has ended,
14772 haproxy may accept thousands of connections in a few milliseconds. The time
14773 spent accepting these connections will inevitably slightly delay processing
14774 of other connections, and it can happen that request times in the order of
14775 a few tens of milliseconds are measured after a few thousands of new
Patrick Mezard105faca2010-06-12 17:02:46 +020014776 connections have been accepted at once. Setting "option http-server-close"
14777 may display larger request times since "Tq" also measures the time spent
14778 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014779
14780 - If "Tc" is close to 3000, a packet has probably been lost between the
14781 server and the proxy during the server connection phase. This value should
14782 always be very low, such as 1 ms on local networks and less than a few tens
14783 of ms on remote networks.
14784
Willy Tarreau55165fe2009-05-10 12:02:55 +020014785 - If "Tr" is nearly always lower than 3000 except some rare values which seem
14786 to be the average majored by 3000, there are probably some packets lost
14787 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014788
14789 - If "Tt" is large even for small byte counts, it generally is because
14790 neither the client nor the server decides to close the connection, for
14791 instance because both have agreed on a keep-alive connection mode. In order
14792 to solve this issue, it will be needed to specify "option httpclose" on
14793 either the frontend or the backend. If the problem persists, it means that
14794 the server ignores the "close" connection mode and expects the client to
14795 close. Then it will be required to use "option forceclose". Having the
14796 smallest possible 'Tt' is important when connection regulation is used with
14797 the "maxconn" option on the servers, since no new connection will be sent
14798 to the server until another one is released.
14799
14800Other noticeable HTTP log cases ('xx' means any value to be ignored) :
14801
14802 Tq/Tw/Tc/Tr/+Tt The "option logasap" is present on the frontend and the log
14803 was emitted before the data phase. All the timers are valid
14804 except "Tt" which is shorter than reality.
14805
14806 -1/xx/xx/xx/Tt The client was not able to send a complete request in time
14807 or it aborted too early. Check the session termination flags
14808 then "timeout http-request" and "timeout client" settings.
14809
14810 Tq/-1/xx/xx/Tt It was not possible to process the request, maybe because
14811 servers were out of order, because the request was invalid
14812 or forbidden by ACL rules. Check the session termination
14813 flags.
14814
14815 Tq/Tw/-1/xx/Tt The connection could not establish on the server. Either it
14816 actively refused it or it timed out after Tt-(Tq+Tw) ms.
14817 Check the session termination flags, then check the
14818 "timeout connect" setting. Note that the tarpit action might
14819 return similar-looking patterns, with "Tw" equal to the time
14820 the client connection was maintained open.
14821
14822 Tq/Tw/Tc/-1/Tt The server has accepted the connection but did not return
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014823 a complete response in time, or it closed its connection
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014824 unexpectedly after Tt-(Tq+Tw+Tc) ms. Check the session
14825 termination flags, then check the "timeout server" setting.
14826
14827
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200148288.5. Session state at disconnection
14829-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014830
14831TCP and HTTP logs provide a session termination indicator in the
14832"termination_state" field, just before the number of active connections. It is
148332-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
14834each of which has a special meaning :
14835
14836 - On the first character, a code reporting the first event which caused the
14837 session to terminate :
14838
14839 C : the TCP session was unexpectedly aborted by the client.
14840
14841 S : the TCP session was unexpectedly aborted by the server, or the
14842 server explicitly refused it.
14843
14844 P : the session was prematurely aborted by the proxy, because of a
14845 connection limit enforcement, because a DENY filter was matched,
14846 because of a security check which detected and blocked a dangerous
14847 error in server response which might have caused information leak
Willy Tarreau570f2212013-06-10 16:42:09 +020014848 (eg: cacheable cookie).
14849
14850 L : the session was locally processed by haproxy and was not passed to
14851 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014852
14853 R : a resource on the proxy has been exhausted (memory, sockets, source
14854 ports, ...). Usually, this appears during the connection phase, and
14855 system logs should contain a copy of the precise error. If this
14856 happens, it must be considered as a very serious anomaly which
14857 should be fixed as soon as possible by any means.
14858
14859 I : an internal error was identified by the proxy during a self-check.
14860 This should NEVER happen, and you are encouraged to report any log
14861 containing this, because this would almost certainly be a bug. It
14862 would be wise to preventively restart the process after such an
14863 event too, in case it would be caused by memory corruption.
14864
Simon Horman752dc4a2011-06-21 14:34:59 +090014865 D : the session was killed by haproxy because the server was detected
14866 as down and was configured to kill all connections when going down.
14867
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070014868 U : the session was killed by haproxy on this backup server because an
14869 active server was detected as up and was configured to kill all
14870 backup connections when going up.
14871
Willy Tarreaua2a64e92011-09-07 23:01:56 +020014872 K : the session was actively killed by an admin operating on haproxy.
14873
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014874 c : the client-side timeout expired while waiting for the client to
14875 send or receive data.
14876
14877 s : the server-side timeout expired while waiting for the server to
14878 send or receive data.
14879
14880 - : normal session completion, both the client and the server closed
14881 with nothing left in the buffers.
14882
14883 - on the second character, the TCP or HTTP session state when it was closed :
14884
Willy Tarreauf7b30a92010-12-06 22:59:17 +010014885 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014886 (HTTP mode only). Nothing was sent to any server.
14887
14888 Q : the proxy was waiting in the QUEUE for a connection slot. This can
14889 only happen when servers have a 'maxconn' parameter set. It can
14890 also happen in the global queue after a redispatch consecutive to
14891 a failed attempt to connect to a dying server. If no redispatch is
14892 reported, then no connection attempt was made to any server.
14893
14894 C : the proxy was waiting for the CONNECTION to establish on the
14895 server. The server might at most have noticed a connection attempt.
14896
14897 H : the proxy was waiting for complete, valid response HEADERS from the
14898 server (HTTP only).
14899
14900 D : the session was in the DATA phase.
14901
14902 L : the proxy was still transmitting LAST data to the client while the
14903 server had already finished. This one is very rare as it can only
14904 happen when the client dies while receiving the last packets.
14905
14906 T : the request was tarpitted. It has been held open with the client
14907 during the whole "timeout tarpit" duration or until the client
14908 closed, both of which will be reported in the "Tw" timer.
14909
14910 - : normal session completion after end of data transfer.
14911
14912 - the third character tells whether the persistence cookie was provided by
14913 the client (only in HTTP mode) :
14914
14915 N : the client provided NO cookie. This is usually the case for new
14916 visitors, so counting the number of occurrences of this flag in the
14917 logs generally indicate a valid trend for the site frequentation.
14918
14919 I : the client provided an INVALID cookie matching no known server.
14920 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020014921 cookies between HTTP/HTTPS sites, persistence conditionally
14922 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014923
14924 D : the client provided a cookie designating a server which was DOWN,
14925 so either "option persist" was used and the client was sent to
14926 this server, or it was not set and the client was redispatched to
14927 another server.
14928
Willy Tarreau996a92c2010-10-13 19:30:47 +020014929 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014930 server.
14931
Willy Tarreau996a92c2010-10-13 19:30:47 +020014932 E : the client provided a valid cookie, but with a last date which was
14933 older than what is allowed by the "maxidle" cookie parameter, so
14934 the cookie is consider EXPIRED and is ignored. The request will be
14935 redispatched just as if there was no cookie.
14936
14937 O : the client provided a valid cookie, but with a first date which was
14938 older than what is allowed by the "maxlife" cookie parameter, so
14939 the cookie is consider too OLD and is ignored. The request will be
14940 redispatched just as if there was no cookie.
14941
Willy Tarreauc89ccb62012-04-05 21:18:22 +020014942 U : a cookie was present but was not used to select the server because
14943 some other server selection mechanism was used instead (typically a
14944 "use-server" rule).
14945
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014946 - : does not apply (no cookie set in configuration).
14947
14948 - the last character reports what operations were performed on the persistence
14949 cookie returned by the server (only in HTTP mode) :
14950
14951 N : NO cookie was provided by the server, and none was inserted either.
14952
14953 I : no cookie was provided by the server, and the proxy INSERTED one.
14954 Note that in "cookie insert" mode, if the server provides a cookie,
14955 it will still be overwritten and reported as "I" here.
14956
Willy Tarreau996a92c2010-10-13 19:30:47 +020014957 U : the proxy UPDATED the last date in the cookie that was presented by
14958 the client. This can only happen in insert mode with "maxidle". It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014959 happens every time there is activity at a different date than the
Willy Tarreau996a92c2010-10-13 19:30:47 +020014960 date indicated in the cookie. If any other change happens, such as
14961 a redispatch, then the cookie will be marked as inserted instead.
14962
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014963 P : a cookie was PROVIDED by the server and transmitted as-is.
14964
14965 R : the cookie provided by the server was REWRITTEN by the proxy, which
14966 happens in "cookie rewrite" or "cookie prefix" modes.
14967
14968 D : the cookie provided by the server was DELETED by the proxy.
14969
14970 - : does not apply (no cookie set in configuration).
14971
Willy Tarreau996a92c2010-10-13 19:30:47 +020014972The combination of the two first flags gives a lot of information about what
14973was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014974helpful to detect server saturation, network troubles, local system resource
14975starvation, attacks, etc...
14976
14977The most common termination flags combinations are indicated below. They are
14978alphabetically sorted, with the lowercase set just after the upper case for
14979easier finding and understanding.
14980
14981 Flags Reason
14982
14983 -- Normal termination.
14984
14985 CC The client aborted before the connection could be established to the
14986 server. This can happen when haproxy tries to connect to a recently
14987 dead (or unchecked) server, and the client aborts while haproxy is
14988 waiting for the server to respond or for "timeout connect" to expire.
14989
14990 CD The client unexpectedly aborted during data transfer. This can be
14991 caused by a browser crash, by an intermediate equipment between the
14992 client and haproxy which decided to actively break the connection,
14993 by network routing issues between the client and haproxy, or by a
14994 keep-alive session between the server and the client terminated first
14995 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010014996
Willy Tarreaucc6c8912009-02-22 10:53:55 +010014997 cD The client did not send nor acknowledge any data for as long as the
14998 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020014999 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015000
15001 CH The client aborted while waiting for the server to start responding.
15002 It might be the server taking too long to respond or the client
15003 clicking the 'Stop' button too fast.
15004
15005 cH The "timeout client" stroke while waiting for client data during a
15006 POST request. This is sometimes caused by too large TCP MSS values
15007 for PPPoE networks which cannot transport full-sized packets. It can
15008 also happen when client timeout is smaller than server timeout and
15009 the server takes too long to respond.
15010
15011 CQ The client aborted while its session was queued, waiting for a server
15012 with enough empty slots to accept it. It might be that either all the
15013 servers were saturated or that the assigned server was taking too
15014 long a time to respond.
15015
15016 CR The client aborted before sending a full HTTP request. Most likely
15017 the request was typed by hand using a telnet client, and aborted
15018 too early. The HTTP status code is likely a 400 here. Sometimes this
15019 might also be caused by an IDS killing the connection between haproxy
Willy Tarreau0f228a02015-05-01 15:37:53 +020015020 and the client. "option http-ignore-probes" can be used to ignore
15021 connections without any data transfer.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015022
15023 cR The "timeout http-request" stroke before the client sent a full HTTP
15024 request. This is sometimes caused by too large TCP MSS values on the
15025 client side for PPPoE networks which cannot transport full-sized
15026 packets, or by clients sending requests by hand and not typing fast
15027 enough, or forgetting to enter the empty line at the end of the
Willy Tarreau2705a612014-05-23 17:38:34 +020015028 request. The HTTP status code is likely a 408 here. Note: recently,
Willy Tarreau0f228a02015-05-01 15:37:53 +020015029 some browsers started to implement a "pre-connect" feature consisting
15030 in speculatively connecting to some recently visited web sites just
15031 in case the user would like to visit them. This results in many
15032 connections being established to web sites, which end up in 408
15033 Request Timeout if the timeout strikes first, or 400 Bad Request when
15034 the browser decides to close them first. These ones pollute the log
15035 and feed the error counters. Some versions of some browsers have even
15036 been reported to display the error code. It is possible to work
15037 around the undesirable effects of this behaviour by adding "option
15038 http-ignore-probes" in the frontend, resulting in connections with
15039 zero data transfer to be totally ignored. This will definitely hide
15040 the errors of people experiencing connectivity issues though.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015041
15042 CT The client aborted while its session was tarpitted. It is important to
15043 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020015044 wrong tarpit rules have been written. If a lot of them happen, it
15045 might make sense to lower the "timeout tarpit" value to something
15046 closer to the average reported "Tw" timer, in order not to consume
15047 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015048
Willy Tarreau570f2212013-06-10 16:42:09 +020015049 LR The request was intercepted and locally handled by haproxy. Generally
15050 it means that this was a redirect or a stats request.
15051
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010015052 SC The server or an equipment between it and haproxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015053 the TCP connection (the proxy received a TCP RST or an ICMP message
15054 in return). Under some circumstances, it can also be the network
15055 stack telling the proxy that the server is unreachable (eg: no route,
15056 or no ARP response on local network). When this happens in HTTP mode,
15057 the status code is likely a 502 or 503 here.
15058
15059 sC The "timeout connect" stroke before a connection to the server could
15060 complete. When this happens in HTTP mode, the status code is likely a
15061 503 or 504 here.
15062
15063 SD The connection to the server died with an error during the data
15064 transfer. This usually means that haproxy has received an RST from
15065 the server or an ICMP message from an intermediate equipment while
15066 exchanging data with the server. This can be caused by a server crash
15067 or by a network issue on an intermediate equipment.
15068
15069 sD The server did not send nor acknowledge any data for as long as the
15070 "timeout server" setting during the data phase. This is often caused
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010015071 by too short timeouts on L4 equipments before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015072 load-balancers, ...), as well as keep-alive sessions maintained
15073 between the client and the server expiring first on haproxy.
15074
15075 SH The server aborted before sending its full HTTP response headers, or
15076 it crashed while processing the request. Since a server aborting at
15077 this moment is very rare, it would be wise to inspect its logs to
15078 control whether it crashed and why. The logged request may indicate a
15079 small set of faulty requests, demonstrating bugs in the application.
15080 Sometimes this might also be caused by an IDS killing the connection
15081 between haproxy and the server.
15082
15083 sH The "timeout server" stroke before the server could return its
15084 response headers. This is the most common anomaly, indicating too
15085 long transactions, probably caused by server or database saturation.
15086 The immediate workaround consists in increasing the "timeout server"
15087 setting, but it is important to keep in mind that the user experience
15088 will suffer from these long response times. The only long term
15089 solution is to fix the application.
15090
15091 sQ The session spent too much time in queue and has been expired. See
15092 the "timeout queue" and "timeout connect" settings to find out how to
15093 fix this if it happens too often. If it often happens massively in
15094 short periods, it may indicate general problems on the affected
15095 servers due to I/O or database congestion, or saturation caused by
15096 external attacks.
15097
15098 PC The proxy refused to establish a connection to the server because the
15099 process' socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020015100 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015101 so that it does not happen anymore. This status is very rare and
15102 might happen when the global "ulimit-n" parameter is forced by hand.
15103
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010015104 PD The proxy blocked an incorrectly formatted chunked encoded message in
15105 a request or a response, after the server has emitted its headers. In
15106 most cases, this will indicate an invalid message from the server to
Willy Tarreauf3a3e132013-08-31 08:16:26 +020015107 the client. Haproxy supports chunk sizes of up to 2GB - 1 (2147483647
15108 bytes). Any larger size will be considered as an error.
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010015109
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015110 PH The proxy blocked the server's response, because it was invalid,
15111 incomplete, dangerous (cache control), or matched a security filter.
15112 In any case, an HTTP 502 error is sent to the client. One possible
15113 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010015114 containing unauthorized characters. It is also possible but quite
15115 rare, that the proxy blocked a chunked-encoding request from the
15116 client due to an invalid syntax, before the server responded. In this
15117 case, an HTTP 400 error is sent to the client and reported in the
15118 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015119
15120 PR The proxy blocked the client's HTTP request, either because of an
15121 invalid HTTP syntax, in which case it returned an HTTP 400 error to
15122 the client, or because a deny filter matched, in which case it
15123 returned an HTTP 403 error.
15124
15125 PT The proxy blocked the client's request and has tarpitted its
15126 connection before returning it a 500 server error. Nothing was sent
15127 to the server. The connection was maintained open for as long as
15128 reported by the "Tw" timer field.
15129
15130 RC A local resource has been exhausted (memory, sockets, source ports)
15131 preventing the connection to the server from establishing. The error
15132 logs will tell precisely what was missing. This is very rare and can
15133 only be solved by proper system tuning.
15134
Willy Tarreau996a92c2010-10-13 19:30:47 +020015135The combination of the two last flags gives a lot of information about how
15136persistence was handled by the client, the server and by haproxy. This is very
15137important to troubleshoot disconnections, when users complain they have to
15138re-authenticate. The commonly encountered flags are :
15139
15140 -- Persistence cookie is not enabled.
15141
15142 NN No cookie was provided by the client, none was inserted in the
15143 response. For instance, this can be in insert mode with "postonly"
15144 set on a GET request.
15145
15146 II A cookie designating an invalid server was provided by the client,
15147 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015148 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020015149 value can be presented by a client when no other server knows it.
15150
15151 NI No cookie was provided by the client, one was inserted in the
15152 response. This typically happens for first requests from every user
15153 in "insert" mode, which makes it an easy way to count real users.
15154
15155 VN A cookie was provided by the client, none was inserted in the
15156 response. This happens for most responses for which the client has
15157 already got a cookie.
15158
15159 VU A cookie was provided by the client, with a last visit date which is
15160 not completely up-to-date, so an updated cookie was provided in
15161 response. This can also happen if there was no date at all, or if
15162 there was a date but the "maxidle" parameter was not set, so that the
15163 cookie can be switched to unlimited time.
15164
15165 EI A cookie was provided by the client, with a last visit date which is
15166 too old for the "maxidle" parameter, so the cookie was ignored and a
15167 new cookie was inserted in the response.
15168
15169 OI A cookie was provided by the client, with a first visit date which is
15170 too old for the "maxlife" parameter, so the cookie was ignored and a
15171 new cookie was inserted in the response.
15172
15173 DI The server designated by the cookie was down, a new server was
15174 selected and a new cookie was emitted in the response.
15175
15176 VI The server designated by the cookie was not marked dead but could not
15177 be reached. A redispatch happened and selected another one, which was
15178 then advertised in the response.
15179
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015180
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200151818.6. Non-printable characters
15182-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015183
15184In order not to cause trouble to log analysis tools or terminals during log
15185consulting, non-printable characters are not sent as-is into log files, but are
15186converted to the two-digits hexadecimal representation of their ASCII code,
15187prefixed by the character '#'. The only characters that can be logged without
15188being escaped are comprised between 32 and 126 (inclusive). Obviously, the
15189escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
15190is the same for the character '"' which becomes "#22", as well as '{', '|' and
15191'}' when logging headers.
15192
15193Note that the space character (' ') is not encoded in headers, which can cause
15194issues for tools relying on space count to locate fields. A typical header
15195containing spaces is "User-Agent".
15196
15197Last, it has been observed that some syslog daemons such as syslog-ng escape
15198the quote ('"') with a backslash ('\'). The reverse operation can safely be
15199performed since no quote may appear anywhere else in the logs.
15200
15201
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200152028.7. Capturing HTTP cookies
15203---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015204
15205Cookie capture simplifies the tracking a complete user session. This can be
15206achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015207section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015208cookie will simultaneously be checked in the request ("Cookie:" header) and in
15209the response ("Set-Cookie:" header). The respective values will be reported in
15210the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015211locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015212not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
15213user switches to a new session for example, because the server will reassign it
15214a new cookie. It is also possible to detect if a server unexpectedly sets a
15215wrong cookie to a client, leading to session crossing.
15216
15217 Examples :
15218 # capture the first cookie whose name starts with "ASPSESSION"
15219 capture cookie ASPSESSION len 32
15220
15221 # capture the first cookie whose name is exactly "vgnvisitor"
15222 capture cookie vgnvisitor= len 32
15223
15224
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200152258.8. Capturing HTTP headers
15226---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015227
15228Header captures are useful to track unique request identifiers set by an upper
15229proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
15230the response, one can search for information about the response length, how the
15231server asked the cache to behave, or an object location during a redirection.
15232
15233Header captures are performed using the "capture request header" and "capture
15234response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015235section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015236
15237It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010015238time. Non-existent headers are logged as empty strings, and if one header
15239appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015240are grouped within braces '{' and '}' in the same order as they were declared,
15241and delimited with a vertical bar '|' without any space. Response headers
15242follow the same representation, but are displayed after a space following the
15243request headers block. These blocks are displayed just before the HTTP request
15244in the logs.
15245
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020015246As a special case, it is possible to specify an HTTP header capture in a TCP
15247frontend. The purpose is to enable logging of headers which will be parsed in
15248an HTTP backend if the request is then switched to this HTTP backend.
15249
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015250 Example :
15251 # This instance chains to the outgoing proxy
15252 listen proxy-out
15253 mode http
15254 option httplog
15255 option logasap
15256 log global
15257 server cache1 192.168.1.1:3128
15258
15259 # log the name of the virtual server
15260 capture request header Host len 20
15261
15262 # log the amount of data uploaded during a POST
15263 capture request header Content-Length len 10
15264
15265 # log the beginning of the referrer
15266 capture request header Referer len 20
15267
15268 # server name (useful for outgoing proxies only)
15269 capture response header Server len 20
15270
15271 # logging the content-length is useful with "option logasap"
15272 capture response header Content-Length len 10
15273
15274 # log the expected cache behaviour on the response
15275 capture response header Cache-Control len 8
15276
15277 # the Via header will report the next proxy's name
15278 capture response header Via len 20
15279
15280 # log the URL location during a redirection
15281 capture response header Location len 20
15282
15283 >>> Aug 9 20:26:09 localhost \
15284 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
15285 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
15286 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
15287 "GET http://fr.adserver.yahoo.com/"
15288
15289 >>> Aug 9 20:30:46 localhost \
15290 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
15291 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
15292 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010015293 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015294
15295 >>> Aug 9 20:30:46 localhost \
15296 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
15297 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
15298 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
15299 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010015300 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015301
15302
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200153038.9. Examples of logs
15304---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015305
15306These are real-world examples of logs accompanied with an explanation. Some of
15307them have been made up by hand. The syslog part has been removed for better
15308reading. Their sole purpose is to explain how to decipher them.
15309
15310 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
15311 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
15312 "HEAD / HTTP/1.0"
15313
15314 => long request (6.5s) entered by hand through 'telnet'. The server replied
15315 in 147 ms, and the session ended normally ('----')
15316
15317 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
15318 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
15319 0/9 "HEAD / HTTP/1.0"
15320
15321 => Idem, but the request was queued in the global queue behind 9 other
15322 requests, and waited there for 1230 ms.
15323
15324 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
15325 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
15326 "GET /image.iso HTTP/1.0"
15327
15328 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010015329 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015330 14 ms, 243 bytes of headers were sent to the client, and total time from
15331 accept to first data byte is 30 ms.
15332
15333 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
15334 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
15335 "GET /cgi-bin/bug.cgi? HTTP/1.0"
15336
15337 => the proxy blocked a server response either because of an "rspdeny" or
15338 "rspideny" filter, or because the response was improperly formatted and
Willy Tarreau3c92c5f2011-08-28 09:45:47 +020015339 not HTTP-compliant, or because it blocked sensitive information which
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015340 risked being cached. In this case, the response is replaced with a "502
15341 bad gateway". The flags ("PH--") tell us that it was haproxy who decided
15342 to return the 502 and not the server.
15343
15344 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010015345 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 +010015346
15347 => the client never completed its request and aborted itself ("C---") after
15348 8.5s, while the proxy was waiting for the request headers ("-R--").
15349 Nothing was sent to any server.
15350
15351 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
15352 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
15353
15354 => The client never completed its request, which was aborted by the
15355 time-out ("c---") after 50s, while the proxy was waiting for the request
15356 headers ("-R--"). Nothing was sent to any server, but the proxy could
15357 send a 408 return code to the client.
15358
15359 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
15360 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
15361
15362 => This log was produced with "option tcplog". The client timed out after
15363 5 seconds ("c----").
15364
15365 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
15366 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010015367 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015368
15369 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015370 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010015371 (config says 'retries 3'), and no redispatch (otherwise we would have
15372 seen "/+3"). Status code 503 was returned to the client. There were 115
15373 connections on this server, 202 connections on this proxy, and 205 on
15374 the global process. It is possible that the server refused the
15375 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010015376
Willy Tarreau52b2d222011-09-07 23:48:48 +020015377
Willy Tarreau0ba27502007-12-24 16:55:16 +010015378/*
15379 * Local variables:
15380 * fill-column: 79
15381 * End:
15382 */